Financial Markets Industry
The New York Chapter of PRMIA hosted "Regulatory, Compliance, and Risk Data Technology Challenges" at Credit Suisse's offices in New York, last Thursday 10th April. Abraham Thomas introduce the panelists, and Don Wesnofske started off by setting the scene for the evening's event.
Don outlined how in reaction to the 2008 Crisis the regulators now require data retention for up to 10 years or more. Don cited one particular example where data must be reconstructed within 24 to 48 hours for any date up to 7 years back, and said that this kind of "forensic" investigation capability was an important consideration for many financial institutions. He took us through a good presentation slide of his view on data management/risk architecture, and outlined how operational risk is comprised of people, process, technology and events. Don ended his presentation by taking us through Wikipedia's definition of "Big Data", and in particular talked about how data has a life cycle going through:
Don handed then handed over to Luigi Mercone of Credit Suisse who is a Director of Engineering Strategy & Architecture at Credit Suisse. Luigi started by saying that to the business at CS, he is technical support which involves asking "What is on fire today? And whats going to be on fire tomorrow?" Luigi described how some time back CS had regulatory enquiry around their equities business which required them to reconstruct data from 2 years back.
The project to do this took around 4-5 months of database adminstrators time to reconstruct the world as at that point in time (I guess because tape storage was being used, and this needed restoring to disk/database). This was for an equity order management system that had doubled in size every year for the past 17 years, and at that point CS was only retaining data going back 2 years. Luigi said that it was then thought that with new regulations requiring the ability to produce forensice evidence at any point in time would potentially swamp CS's resources unless it was addressed head on and strategically.
Luigi described the original architecture that they were using being based on an in-memory database for intraday workloads, then standard Sybase (probably ASE I guess) and then Sybase IQ for longer term archiving, taking advantage of the column-store capabilities of Sybase IQ and the resulting data compression possible. He added that the data storage requirements of the system had grown from 150TB to 1.2PB in 4 years.
Luigi then offered a comparison of this original architecture with what he found by implementing RainStor, in the original architecture the Sybase IQ database compressed data down into 160TB, whereas this was improved by a further factor of 10 down to 14TB using RainStor. He said that the RainStor was self-service providing a standard SQL interface, eliminated the need for tape storage, reduced the system "footprint" by 90% at CS, was 1/5 of the cost and the performance was good. (I guess here I would like to caveat that I know nothing of the original architecture other than the summary Luigi provided, and as such it is hard to judge whether the original architecture was optimal for the data growth experienced, and hence whether this was overall an objective comparison of Sybase IQ's capabilities with RainStor.) Luigi closed by saying that whilst RainStor was a great archive database, its original origins were in in-memory databases and he would encourage RainStor to re-enter that market too, given his experience so far.
John Bantleman CEO of RainStor took over and described how RainStor had been designed specifically for the needs of data archiving (I guess talking more about what it does now rather than its origins outlined by Luigi above). He said that RainStor offers a 20-40x storage footprint reduction over traditional database technology and operates efficiently even at the PetaByte (PB) scale, based around RainStor proprietary database technology making use of columnar storage and being capable of storing data in both relational-style tabular format and also in more "document" style using XML and JSON formats using Key-Value access. John mention that in terms of being able to store data that not only could RainStor retrieve data at a point in time, but it could retrieve the schema being used at that point in time for a more complete view of the state of the world at that point. This echos a couple of past articles that I have penned, one for IRD and one for Wilmott Magazine on bitemporal regulatory requirements.
John said that regulation was driving the need for data archiving capabilities, with 1400 regulations added since 2008 (not sure of source, but believable) and the comment from a Chief Data Officer (CDO) at one financial markets client that if a project wasn't driven by regulatory compliance then the project isn't going to get done (certainly sounds like regulatory overload). John's opening remarks were really around how regulatory cost, complexity and compliance were driving forces behind the growth of RainStor in financial services technology, and whilst regulation is the driver, firms should look at archiving of data as an opportunity too, in order to create value from corporate memory, and to be proactive in addressing future reporting and analysis needs.
John illustrated the regulatory need for data archiving through the Consolidated Audit Trail (CAT) regulation with data retention over 7 years will generate 100PB of data. He also mentioned SEC Rule 17a-4 for broker dealers as another example of "data retention" regulation, with particular reference to storage of records in on-rewriteable, non-erasable format. John termed this WORM storage, meaning Write Once, Read Many. John seemed to imply that both the software (RainStor) and the hardware it runs on (e.g. EMC or Teradata etc) need to be WORM compliant. One of the audience members asked John about BCBS 239, to which John said that he didn't know that particular regulation (fair enough that John didn't know in my opinion, RainStor's tech is general about "data" and is applicable across many industries, whereas BCBS 239 is obviously about banks specifically and is more about data aggregation and reporting than data retention/archiving to my understanding, and this seems to be confirmed with a quick doc scan for "archive" or "retention".)
To finish off the main part of the event (before the drinks and food began) there was a panel discussion. Luigi said that it was best to "prepare for all time, not just specifics" with respect to data retention and that there were dangers in rolling up data (effectively aggregating and loosing granularity to reduce storage needs). John added that his definition of "Big Data" was "All information, for ever". Luigi added that implementing RainStor had allowed CS to spend more time on interesting questions rather than on database restoration. John proposed that version 1 of Big Data involved the retention of web data, and as such loosing a data point here and their didn't matter. Version 2 of Big Data is concerned more with enterprise data where all data has value and needs to be retained i.e. lots of high value data. He added that this was an opportunity for risk and compliance to become an asset.
Abraham (second from left), Don (center) and John (second from right)
Overall it was a good event which I found very interesting (but I have to admit to a certain geeky interest in this kind of tech). The event would have benefitted from say another competitive or complementary technology vendor involved maybe, plus maybe an academic to give a different slant on data retention and on what the regulators hope to gain from this kind of mandated data retention. Not that the regulators have been that good at managing data themselves recently.
Networking afterwards courtesy of Credit Suisse and RainStor
Posted by Brian Sentance | 17 April 2014 | 3:06 pm
Very pleased to announce that Xenomorph will be hosting an event, "Financial Markets Data and Analytics. Everywhere You Need Them.", at Microsoft's Times Square New York offices on May 9th.
This breakfast briefing includes Sang Lee of the analyst firm Aite Group offering some great insights from financial institutions into their adoption of cloud technology, applying it to address risk management, data management and regulatory reporting challenges.
Microsoft will be showing how their new Power BI can radically change and accelerate the integration of data for business and IT staff alike, regardless of what kind of data it is, what format it is stored in or where it is located.
And Xenomorph will be introducing the TimeScape MarketPlace, our new cloud-based data mashup service for publishing and consuming financial markets data and analytics. More background and updates on MarketPlace in coming weeks.
In the meantime, please take a look at the event and register if you can come along, it would be great to see you there.
Posted by Brian Sentance | 15 April 2014 | 3:57 pm
Good article from Tim Harford (he of the enjoyable "Undercover Economist" books) in the FT last week called "Big data: are we making a big mistake". Tim injects some healthy realism into the hype of Big Data without dismissing its importance and potential benefits. The article talks about the four claims often made when talking about Big Data:
- Data analysis often produces uncannily accurate results
- Make statistical samplying obsolete by capturing all the data
- Statistical correlation is all you need - no need to understand causation
- Enough data means that scientific or statistical models aren't needed
Now models can have their own problems, but I can see where he is coming from, for instance 3. and 4. above seem to be in direct contradiction. I particularly like the comment later in the article that "causality won't be discarded, but it is being knocked off its pedestal as the primary fountain of meaning."
Also I liked the definition by one of the academics mentioned of a big data set being one where "N = All", and that you have "all" the data is an incorrect assumption behind some Big Data analysis put forward. Large data sets can mean that sample error is low, but sample bias is still a potentially big problem - for example everyone on Twitter is probably not representative of the population of the human race in general.
So I will now press save on this blog post, publish in Twitter and help re-enforce the impression that Big Data is a hot topic...which it is, but not for everyone I guess is the point.
Posted by Brian Sentance | 8 April 2014 | 10:21 pm
Melissa Sexton of Morgan Stanley introduced the agenda, saying that the evening would focus on three aspects of liquidity risk management:
- industry practice
LiquidityMetrics by MSCI - Carlo Acerbi of MSCI then took over with his presentation on "LiquidityMetrics". Carlo said that he was pleased to be involved with MSCI (and RiskMetrics, aquired by MSCI) in that it had helped to establish and define standards for risk management that were used across the industry. He said that liquidity risk management was difficult because:
- Clarity of Definition - Carlo suggest that if he asked the audience to define liquidity risk he would receive 70 differing definitions. Put another way, he suggested that liquidity risk was "a strange animal with many faces".
- Data Availability - Carlo said that there were aspects of the market that we unobservable and hence data was scarce/non-existent and as such this was a limit on the validity of the models that could be applied to liquidity risk.
Carlo went on to clarify that liquidity risk was different depending upon the organization type/context being considered, with banks obviously focusing on funding. He said that LiquidityMetrics was focused on asset liquidity risk, and as such was more applicable to the needs of asset managers and hedge funds given recent regulation such as UCITS/AIFMD/FormPF. The methodology is aimed at bringing traditional equity market impact models out from the trading floor across into risk management and across other asset classes.
Liquidity Surfaces - LiquidityMetrics measures the expected price impact for an order of a given size, and as such has dimensions in:
- order size
- liquidity time horizon
- transaction costs
The representation shown by Carlo was of a "liquidity surface" with x dimension of order size (both bid and ask around 0), y dimension of time horizon for liquidation and z (vertical) dimension of transaction cost. The surface shown had a U-shaped cross section around zero order size, at which the transaction cost was half the bid-ask spread (this link illustrates my attempt at verbal visualization). The U-shape cross section indicates "Market Impact", its shape over time "Market Elasticity" and the limits for what it is observable "Market Depth".
Carlo then moved to consider a portfolio of instruments, and how obligations on an investment fund (a portfolio) can be translated into the estimated transaction costs of meeting this obligations, so as to quantify the hidden costs of redemption in a fund. He mentioned that LiquidityMetrics could be used to quantify the costs of regulations such as UCITS/AIFMD/FormPF. There was some audience questioning about portfolios of foreign assets, such as holding Russian Bonds (maybe currently topical for an audience member maybe?). Carlo said that you would use both the liquidity surfaces for both the bond itself and the FX transaction (and in FX, there is much data available). He was however keen to emphasize that LiquidityMetrics was not intended to be used to predict "regime change" i.e. it is concerned with transaction costs under normal market conditions).
Model Calibration - In terms of model calibration, then Carlo said that the established equity market impact models (see this link for some background for instance) have observable market data to work with. In equity markets, traditionally there was a "lit" central trading venue (i.e. an exchange) with a star network of participants fanning out from it. In OTC markets such as bonds, there is no star network but rather many to many linkages establised between all market participants, where each participant may have a network of connections of different size. As such there has not been enough data around to calibrate traditional market impact models for OTC markets. As a result, Carlo said that MSCI had implemented some simple models with a relatively small number of parameters.
Two characteristics of standard market impact models are:
- Permanent Effects - this is where the fair price is impacted by a large order and the order book is dragged along to follow this.
- Temporary Effects - this is where the order book is emptied but then liquidity regenerates
Carlo said that the effects were obviously related to the behavioural aspects of market participants. He said that the bright side for bonds (and OTC markets) was given that the trades are private there was no public information, and price movements were often constrained by theoretical pricing, therefore permanent effects could be ignored and the fair price is insenstive to trading (again under "normal" market conditions). Carlo then moved on to talk about some of the research his team was doing looking at the shape of the order book and the time needed to regenerate it. He talked of "Perfectly Elastic" markets that digest orders immediately and "Perfectly Plastic" markets that never regenerate, and how "Relaxation Time" measures in days how long the market takes to regenerate the order book.
Liquidity Observatory - Carlo described how the data was gathered from market participants on a monthly basis using a spreadsheet to categorize the bond/asset class type, and again using simple parameters from active "expert" traders. Take a look at this link and sign up if this is you. (This sounded to me a lot like another "market consensus" data gathering exercise which are proving increasingly popular, such as one the first I had heard of many years back in Totem - we are not quite fully ready for "crowdsourcing" in financial markets maybe, but more people are seeing sense in sharing data.).
Panel Debate - Ron Papenek of MSCI was moderator of the panel, and asked Karen Cassidy of Morgan Stanley about her experiences in liquidity risk management.
Liqudity Risk Management at Banks - Karen started by saying that in liquidity management at Morgan Stanley they look at:
- Operating Capital
- Client Behaviour
Since 2008, Karen said that liquidity management had become a lot more rigorous and formalized, being rule based and using a categorisation of assets held from highly liquid to highly illiquid. She said that Morgan Stanley undertake stress testing by market and also by idiosyncratic risk over time frames of 1 month and 1 year. As part of this they are assessing the minimum operating liquidity needed based on working capital needs.
Karen added that Morgan Stanley are expending a lot of effect currently on data collection and modelling given that their data is specific to a retail broker-dealer unit, unlike many other firms. They are also looking at metrics around financial advisors, and how many clients follow the financial advisor when he or she decides to switch firms.
Business or Regulation Driving Liquidity Risk Management - Ron asked Karen what were the drivers of their processes at Morgan Stanley. Karen said that in 2008 the focus was on fundability of assets, saying that the FED was monitoring this on a daily basis. She made the side comment that this monitoring was not unusual since "Regulators live with us anyway". Karen said that it was the responsibility of firms to come up with the controls and best practice needed to manage liquidity risk, and that is what Morgan Stanley do anyway.
Karen added that in her view the industry was over-funding and funding too long in response to regulation, and that funding would be at lower but still pragmatic levels in the absence of regulatory pressure. Like many in the industry, Karen thought the regulation had swung too far in response to the 2008 crisis and would eventually swing back to more normal levels.
Carlo added that he had written an unintentionally prescient academic paper on liquidity management in 2008 just prior to the crisis hitting, and he thought the regulators certainly arrived "after" the crisis rather than anticipating it in any way. He thought that the banks have anticipated the regulators very well with measures such as LCR and SFR already in place.
In contrast, Carlo said that the regulators were lost in dealing with liquidity risk management for asset managers and hedge funds, with regulation such as UCITS being very vague on this topic and regulators themselves seeking guidance from the industry. He recounted a meeting he had with BaFin in 2009 where he told them that certain of their regulations made no sense and he said they acknowledge this and said the asset management industry needed to tell them what to implement (sounds like the German regulator is using the same card as the UK regulators in keeping regulations vague when they are uncertain, waiting for regulated firms to implement them to see what the regulation really becomes...).
What Have We Learnt Since 2008 - Karen said that back in 2008 liquidity was not managed to term, funding basis was not rigorous and relied heavily on unsecured debt. She said that since then Morgan Stanley had been actively involved in shaping the requirements of better liquidity risk management with more rigorous analysis of counterparties and funding capacity. Karen said that stronger governance was a foundation for the creation of better policy and process. She said that regulators were receptive to new ideas and had been working with them closely.
What will be the effect of CCPs on OTC markets? Carlo said that when executing a large order, you have the choice between executing 1) multiple small orders with multiple counterparties or 2) a single large block order with one counterparty. In this regard, the equity and bond markets are very different. In lit equity venues, the best approach is 1), but in the bond markets approach 2) is taken since the trade information is not transparent to the market.
Obviously equity markets have become more fragmented, and this has resulted in improve market quality since it is harder to get all market information and hence the market is less resonant to big events/orders. Carlo added that with the increased transparency proposed for OTC markets with CCPs etc will this improve them? His answer was that this was likely to improve the counterparty risk inherent in the market but due to increased transaparency is likely to have a negative effect on transaction costs (I guess another example of the law of unintended consequencies for the regulators).
Audience Questions - there then followed some audience questions:
LiqidityMetrics extrapolation - one audience member asked about transaction cost extrapolation in Carlo's modelling. Carlo said that MSCI do not extrapolate and the liquidity surface terminates where the market terminates its liquidity. There was some extrapolation used along the time dimension however particularly in relation to the time-relaxation parameter.
LiquidityMetrics "Cross-Impact" - looking at applying LiquidityMetrics to a portfolio, one audience member wondering if an order for one asset distorted the liquidity surface for other potentially related assets. Carlo said this was a very interesting area with little research done so far. He said that this "cross-impact" had not been detected in equity markets but that they were looking at it in other markets such as fixed income where effective two assets might be proxies for duration related trading. Carlo put forward a simple model of where the two assets are analogous to two species of animal feeding from the same source of food.
Long and short position liquidity modelling - one audience member asked Carlo what the effects would be of being long or short and that in a crisis you would prefer to be short (maybe obviously?) given the sell off by those with long positions. Carlo clarified that being "short" was not merely taking the negative number on a liquidity surface for a particular asset but rather a "short" is a borrowing position with an obligation to deliver a security at some defined point, and as such is a different asset with its own liquidity surface.
Changing markets, changing participants - final question of the evening was from one member of the audience who asked if the general move out of fixed income trading by the banks over recent years was visible in Carlo's data? Carlo said that MSCI only have around two years of data so far and as such this was not yet visible but his team are looking for effects like this amongst others. He added that the August 2011 weak banks - weak sovereigns in Europe was visible with signals present in the data.
Good food and good (really good I thought) wine put on by MSCI at the event reception. Great view of Manhattan from the 48th floor of World Trade Centre 7 too.
Posted by Brian Sentance | 31 March 2014 | 2:40 pm
The second panel of the day was "Regulation and Risk as Data Management Drivers" - you can find the A-Team's write up here. Some of my thoughts/notes can be found below:
- Ian Webster of Axioma responded to a question about whether consistency was the Holy Grail of data management said that there isn't consistent view possible for data used in risk and regulation - there are many regulations with many different requirements and so unnecessary data consistency is "the hobgoblin of little minds" in delaying progress and achieving goals in data management.
- James of Lombard Risk suggest that firms should seek competitive advantage from regulatory compliance rather than just compliance alone - seeking the carrot and not just avoiding the stick.
- Ian said he thought too many firms dealt with regulatory compliance in a tactical manner and asked if regulation and risk were truly related? He suggested that risk levels might remain unchanged even if regulation demanded a great deal more reporting.
- Marcelle von Wendland said she thought that regulation added cost only, and that firms must focus on risk management and margin.
- James said that "regulatory risk" was a category of risk all in itself alongside its mainstream comtempories.
- Ian added that risk and finance think about risk differently and this didn't help in promoting consistency of ideas in discussions about risk management.
- James said that the legacy of systems in financial markets was a hindrince in complying with new regulation and mentioned the example of the relatively young energy industry where STP was much easier to implement.
- Laurent of Bloomberg said that young, emerging markets like energy were greenfield and as such easier to implement systems but that they did not have any experience or culture around data governance.
- Marcelle said that the G20 initiatives around trade reporting at least promoted some consistency and allowed issues to be identified at last.
- Ian said in response that was unconvinced about politically driven regulation, questioning its effectiveness and motivations.
- Ian raised the issues of the assumptions behind VaR and said that the current stress tests were overdone.
- Marcelle agreed that a single number for VaR or some other measure meant that other useful information has potentially been ignored/thrown away.
- General consensus across the panel that fines were not enough and that restricting business activities might be a more effective stick for the regulators.
- James reference the risk data aggregation paper from the Basel Committee and suggested that data should be capture once, cleaned once and used many times.
- Ian disagreed with James in that he thought clean once, capture once and use many times was not practically possible and this goal was one of the main causes of failure within the data management industry over the past 10 years.
- The panel ended with Ian saying that we not just solve for the last crisis, but the underlying causes of crises were similar and mostly around asset price bubbles so in order to recuce risk in the system 1) lets make data more transparent and 2) do what we can to avoid bubbles with better indices and risk measures.
Posted by Brian Sentance | 24 March 2014 | 6:09 pm
I went along to this PRMIA event on Thursday evening hosted by Credit Suisse and sponsored by Acacia Capital. Viktoria Baklanova introduced the panel with Joseph Tenaga as MC for the panel and very quickly got a plug in for her about to be released book written with Joe on money market funds. For those of us who don't know so much about money market funds, then these are a form of interest-bearing fund that invests in short term debt securities. The funds attempt to maintain a stable Net Asset Value (NAV) but to quote Wikipedia they "are widely (though not necessarily accurately) regarded as being as safe as bank deposits yet providing a higher yield." Their role in the 2008 financial crisis echos on strongly through to the present day, with controversy of their supposedly stable NAV (typically $1 in the US) and the associated phrase "Breaking the Buck".
Joe Tenaga started the panel with an (unnecessary in my view) justification of academia, asking the rhetorical question "What is the point of academia?" to which Joe answered that "knowledge is what makes the impossible possible" and he added that knowledge drives us to make things better. Joe introduced the next panelist, Matthew Fink of Oppenheimer Mutual Funds. Matt said that we would be prepared to wager that he had worked in the money market funds area the longest of anyone in the room, having started his involvement in the industry in April of 1971. Matt gave a picture of the mutual funds industry at the time, with around $60B AUM in the US with 95% invested in equities. At that time the mutual funds industry was going through a very bad time, as the economy and markets were falling and fund redemptions were rising to such an extent that they had fallen to $30B over the next few years. At the time, if redemptions had continued at this rate the industry would have vanished.
Against this background for the mutual funds industry, interest rates in the US were very high rising from 6% in 1969 to around 12% in 1974. So many people were paying very high rates on mortgage obligations whilst being limited to receiving only 4-5% on savings due to "Regulation Q". For wealthy individuals, it was possible to get around these savings limits, but only if you had $100,000 to put in a Commercial Deposit or $10,000 into a T-Bill. Ironically it was the regulation to remove one risk (it had been thought that competition on deposit rates had contributed to the bank failures of the Great Depression) that had sparked the drive to innovate to find higher returns and create the money market funds industry as a result, with the first fund being "The Reserve Fund" in 1971. (side comment - if regulation from the 1930's via the 1970 can cause problems in 2014, then I would have to defer to which ever Deity you worsphip to advise on what the longer-term consequencies will be of the current round of complexity being implemented...).
The banks saw the money pouring into money market funds such as those from Fidelity and Dreyfus, and understandably wanted to be part of the party too. Some of the worries about money market funds were firstly what if a fund got into trouble? Secondly, the bank regulators were angry that funds were flowing into this new industry and were concerned that it would increase bank failures. 1979 saw a certain Paul Volcker (ever heard of him?) complaining that money market funds were acting like checking accounts. Matt said that he spoke with Volcker and said that this was not the case, to which Volcker replied that it was true since his wife's company was paying staff wages out on checks written against money market funds.
Henry Shilling of Moody's took over from Matt and showed a few slides, firstly showing the number of funds with AAA (AAAmmf, Aaamf, AAAmf) from Fitch (49), Moody's (130) and S&P Ratings (156). Henry described how regulators have wanted to reduce the risk of funds by shortening the maturity of the debt held from 90 to 60 days, and having one and seven day liquidity windows. He showed that there is a high degree of concentration risk in the industry with the top 10 firms have 74% AUM and the top 20 covering 94% of the AUM for the industry. Similarly, looking at the assets invested in the funds, 80% are from financial institutions.
Igor Axenov of Barclays Capital then showed his slides, illustrating the composition of the funds by asset type prior to the crisis:
- ABS related - 34%
- Bank products - 23%
- Repos - 15%
- Corporate - 11%
- Unsecured - 8%
- Other - 6%
He said that the largest exposure then was to securitized products, with implicit indirect exposure to banks. Igor said that CDO issuance was rising at a rate of $300B per year through 2005/6/7 and that much of the structuring was done to ensure that the ABS products fitted the needs and regulations of money market funds. Detailing the ABS asset composition, Igor showed:
- Asset Backed (AB) commercial paper - 50%
- AB medium term notes - 24%
- Extendible AB commercial paper - 17%
- ABS Bonds - 5%
Igor said the asset backed commercial paper market (largely funded through money market funds) had grown to $1.2Trln by 2007, and has fallen precipitously since then down to around $200M now.
Looking at the current money market fund portfolio, it looks like:
- Bank products - 41%
- Repos - 18%
- CP - 15%
- US Govt + Agency debt - 10%
- Asset backed CP - 9%
- Other corporate - 4%
Terence Ma added that the Money Market Fund industry sat at 4Trln in 2008 and was now around $2.7Trln in 2014. Matthew Fink said that given his involvement in regulation that he had "never met the face of the enemy before" in Igor was the start of some lively but well-intended banter between the ex-regulator and structurer.
Terence Ma of South Street Securities described his business, which exclusively involves repurchase agreements "Repos". Terry said that in the 1990's Citi were very disciplined on balanced sheet management and in his opinion, then adhead of the market in this regard. He that the Repo business earns small spreads and as a result needs a big balance sheet. When John Read took over Citi, he decided that he did not like the Repo business since its ROE could not compete with some of the products in retail and other parts of the business. So Terry and his partners wondered whether the Repo business could be managed off balance sheet, so they formed a broker-dealer business and when Citi merged with Salomen Brothers they span off. This was December of 2003 but by 2008 they were left "sucking wind" by the crisis.
Terry was quite explicit that his firm is not part of the "shadow banking system" but are subject to the SEC. He then described a few more things about his business, starting with his definition of a Repo as "an agreement to sell and repurchase a security at a fixed date in the future", with the objectives of providing cash inventory, leverage and short cover. All borrowings are lent out, unlike Lehman Brothers in 2008. They do not finance again structured products unless guaranteed, and only accept collateral from Fannie, Freddie and the US Government.
Joe Tenaga then open out questions to the audience. Someone asked who the first MMF was (I think they missed the first part of the talk) and Matt said that the Keystone MMF filed first but the first was Reserve MMF (which got into trouble in 2008). Matt said that it was interesting that the same people like Paul Volcker were stilled involved with the same concerns about the industry many years on.
The next question was how did early MMFs keep their NAV at $1? Henry said that the "Break the Buck" definition is when there is a mark to market fall of 50bp or more. He said that historically that fund sponsors had addressed any issues with breaking the buck with purchases of the fund at par or direct equity investment in the fund - they did this since the effect on their funds and the industry would be too great to comtemplate. Hence an MMF is not a perfect product but (up until Lehmans in 2008 with a 50% NAV loss) has a near-perfect record. He added that the first funds to break the buck were from Salomon's and First Chicago.
Matt added some further history saying the need to maintain the $1 NAV was initially due to the needs of some of the early investors in the industry, who could not invest in products unless they had fixed NAV. He mentioned that one of the companies, Federated, had a long running battle with the SEC over Money Market Funds, filing for exemptions to avoid some of the restrictions that the SEC was trying to impose since the SEC regarded the MMF industry as damaging the mutual funds industry. He mentioned Rule 2-a7 which defines the accountancy procedures for keeping the NAV at $1, and some of the battles around amortization and penny-rounding policies to facilitate this. To later questions, Matt said that the SEC wants a floating NAV for institutional MMFs but currently wants to leave retail alone (seems somewhat arbitrary choice i.e. lets only change what has been problematic before, ignore anything else and not contemplate what could happen if only we understood things better). He said that the SEC was weak and FSOC is driving the SEC to change (and FSOC itself is a pawn of the Federal Reserve).
Overall an interesting panel, particularly when you have characters such as Matt Fink who know the history and stories within the industry so well.
Posted by Brian Sentance | 24 March 2014 | 12:01 pm
You can find A-Team's view on "Building a Flexible Enterprise Architecture" here. Some additional notes/thoughts:
- I thought Neil van Lint of GoldenSource's comment about "putting lipstick on a pig" with reference to legacy architectures was pretty funny and apt.
- The old Irish joke about asking for directions and receiving the response "Well I wouldn't start from here" is also amusing but too true with our industry and most large organisations.
- "Schema on read, not on write" is getting my award for phrase of the month from NoSQL proponents (quote Amir from Mark Logic).
- Agree that ETL is problematic/a big resource drain but unless starting from a greenfield site it is currently unavoidable.
- I like the idea of FIBO (and decoupling data meaning from data structure) but still left unsure what it actually (practically) covers so far and how much it is used, despite the references to it by Peter of Nordea. I guess it is all a matter of semantics.
- I knew little of TOGAF mentioned by Rupert but maybe that is because I am a techie no more (if I ever was).
- Rupert came back to his "where are we?" and data map questions and asked the audience how many of them had a good handle on where data was used in what systems - unsurprisingly not many with a Morgan Stanley guy saying that there monitoring systems were linked to the operational systems for a full inventory of data.
- I agree that the regulators need to push standards directly on the industry - Amir ended the panel suggesting the regulators need to say things like "Thou shalt use FIBO".
Posted by Brian Sentance | 21 March 2014 | 7:44 pm
Rupert Brown of UBS did the keynote at this Spring's A-Team Data Management Summit (DMS). Rupert's talk was about understanding what data there is within a financial institution and understanding where it comes from and where it goes to. Rupert started by asking the question "Where are we?" illustrating it with a map of systems and data flows for an institution - to my recollection I think he said it stretched to 7 metres in length and did not look that accessible or easy to understand. He asked what dimensions it should have as a "map" of data, wondering what dimensions are analogous to latitude, longitude, altitude and orientation? Maybe things like function, product, process, accounting or legal entity as potential candidates.
Briefly Rupert took a bit of a detour into his love of trains with a little history on the London Underground Map. He started by mentioning the role of George Dow who illustrated maps for train routes in a single line, showing just dependency and lineage (what stations are next etc) and ignoring geography and distance. This was built upon by another gentleman, Harry Beck, who took these ideas a stage further with the early ancestors of the current Undergroud map, showing both routes but interweaving all the lines together into a map that additionally was topologically sufficient (indicating broad direction - NESW).
Continuing on with this analogy of Underground to maps of data and data management, Rupert then mentioned Frank Pick who created the Underground brand. Through creating such an identifiable brand, effectively Frank got people to believe and refer to the map, and that people in data governance need and could benefit from taking a similar approach to data governance with data management. I guess it is easy to take maps we see every day for granted and particularly some of the thought that went into them, maybe ideas that initially were not intuitive (or at least not directly representative of physical reality) but that greatly improved understand and comprehension. Put another way, representing reality one for one does not necessarily get you to something that is easy to understand (sounds like a "model" to me).
Rupert then described some of his efforts using Open Street Map to map data, making use of the concepts of nodes, ways and areas. Apparently he had implemented this using a NoSQL database (Mark Logic) for performance reasons (doesn't sound like a really "big data" sized problem with several hundred apps and several thousand data transports but nevertheless he said it was needed, maybe as a result of its graph like nature?). He said that the data was crowdsourced to refine the data, with a wiki for annotations. He said he was interested in the bitemporality of data, i.e. how the map changes over time. He advised that every application should also be thought of as its own "databus" in addition to any de facto databuses might be present in the architecture.
In summary the talk was interesting, but it was demonstrable from what Rupert showed that we have long way to go in representing clearly and easily where data came from, where it goes to and how it is used. I think Rupert acknowledges this and has some academic partnerships trying to develop better ways of representing and visualizing data. Certainly data lineage and audit trail on everything is a hot topic for many of our clients currently, and something that deserves more attention. You can download Rupert's presentation here and the A-Team's take on his talk can be found here.
Posted by Brian Sentance | 20 March 2014 | 5:49 pm
Christian Nilsson of S&P CIQ followed up Richard Burtsal's talk with a presentation on data management for risk, containing many interesting questions for those considering data for risk management needs. Christian started his talk by taking a time machine back to 2006, and asking what were the issues then in Enterprise Data Management:
- There is no current crisis - we have other priorities (we now know what happened there)
- The business case is still too fuzzy (regulation took care of this issue)
- Dealing with the politics of implementation (silos are still around, but cost and regulation are weakening politics as a defence?)
- Understanding data dependencies (understanding this throughout the value chain, but still not clear today?)
- The risk of doing it wrong (there are risk you will do data management wrong given all the external parties and sources involved, but what is the risk of not doing it?)
Christian then moved on to say the current regulatory focus is on clearer roadmaps for financial institutions, citing Basel II/III, Dodd Frank/Volker Rule in the US, challenges in valuation from IASB and IFRS, fund management challenges with UCITS, AIFMD, EMIR, MiFID and MiFIR, and Solvency II in the Insurance industry. He coined the phrase that "Regulation Goes Hollywood" with multiple versions of regulation like UCITS I, II, III, IV, V, VII for example having more versions than a set of Rocky movies.
He then touched upon some of the main motivations behind the BCBS 239 document and said that regulation had three main themes at the moment:
- Higher Capital and Liquidity Ratios
- Restrictions on Trading Activities
- Structural Changes ("ring fence" retail, global operations move to being capitalized local subsidiaries)
Some further observations were on what will be the implications of the effective "loss" of globablization within financial markets, and also what now can be considered as risk free assets (do such things now exist?). Christian then gave some stats on risk as a driver of data and technology spend with over $20-50B being spent over the next 2-3 years (seems a wide range, nothing like a consensus from analysts I guess!).
The talk then moved on to what role data and data management plays within regulatory compliance, with for example:
- LEI - Legal Entity Identifiers play out throughout most regulation, as a means to enable automated processing and as a way to understand and aggregate exposures.
- Dodd-Frank - Data management plays within OTC processing and STP in general.
- Solvency II - This regulation for insurers places emphasis on data quality/data lineage and within capital reserve requirements.
- Basel III - Risk aggregation and counterparty credit risk are two areas of key focus.
Christian outlined the small budget of the regulators relative to the biggest banks (a topic discussed in previous posts, how society wants stronger, more effective regulation but then isn't prepared to pay for it directly - although I would add we all pay for it indirectly but that is another story, in part illustrated in the document this post talks about).
In addtion to the well-known term "regulatory arbitrage" dealing with different regulations in different jurisdictions, Christian also mentioned the increasingly used term "subsituted compliance" where a global company tries to optimise which jurisdictions it and its subsidiaries comply within, with the aim of avoiding compliance in more difficult regimes through compliance within others.
I think Christian outlined the "data management dichotomy" within financial markets very well :
- Regulation requires data that is complete, accurate and appropriate
- Industry standards of data management and data are poorly regulated, and there is weak industry leadership in this area.
(not sure if it was quite at this point, but certainly some of the audience questions were about whether the data vendors themselves should be regulated which was entertaining).
He also outlined the opportunity from regulation in that it could be used as a catalyst for efficiency, STP and cost base reduction.
Obviously "Big Data" (I keep telling myself to drop the quotes, but old habits die hard) is hard to avoid, and Christian mentioned that IBM say that 90% of the world's data has been created in the last 2 years. He described the opportunities of the "3 V's" of Volume, Variety, Velocity and "Dark Data" (exploiting underused data with new technology - "Dark" and "Deep" are getting more and more use of late). No mention directly in his presentation but throughout there was the implied extension of the "3 V's" to "5 V's" with Veracity (aka quality) and Value (aka we could do this, but is it worth it?). Related to the "Value" point Christian brought out the debate about what data do you capture, analyse, store but also what do you deliberately discard which is point worth more consideration that it gets (e.g. one major data vendor I know did not store its real-time tick data and now buys its tick data history from an institution who thought it would be a good idea to store the data long before the data vendor thought of it).
I will close this post taking a couple of summary lists directly from his presentation, the first being the top areas of focus for risk managers:
- Counterparty Risk
- Integrating risk into the Pre-trade process
- Risk Aggregation across the firm
- Risk Transparency
- Cross Asset Risk Reporting
- Cost Management/displacement
The second list outlines the main challenges:
- Getting complete view of risk from multiple systems
- Lack of front to back integration of systems
- Data Mapping
- Data availability of history
- Lack of Instrument coverage
- Inability to source from single vendor
- Growing volumes of data
Christian's presentation then put forward a lot of practical ideas about how best to meet these challenges (I particularly liked the risk data warehouse parts, but I am unsurprisingly biassed). In summary if you get the chance then see or take a read of Christian's presentation, I thought it was a very thoughtful document with some interesting ideas and advice put forward.
Posted by Brian Sentance | 11 March 2014 | 7:26 pm
Attended a good event at S&P Capital IQ's offices on Tuesday morning last week in London, built around the BCBS 239 document on risk aggregation and reporting (see earlier PRMIA event on this topic too). A partner vendor of S&P CIQ, Tech Mahindra, started the morning with Richard Burtsal's presentation on "Delivering an Enterprise Data Strategy". Tech Mahindra recently acquired a data management platform from UBS Asset Management and are offering a managed service data management offering based on this (see A-Team article).
Richard said that he wasn't going to "sell" in his presentation (always a worrying admission from one of us data management vendors, it usually means entirely the opposite). That small criticism aside, Richard gave a solid update on the state of the industry and obviously on what Tech Mahindra are offering, and added that:
- For every $1 spent directly on market data, the total cost of that data goes up by a factor of 6 by the time the data is actually used
- 33% of rejected trades are caused by incorrect reference data
- 60% of staff manipulate, report on or support data on a daily basis (I wonder what the other 40% actually do then? Be good to get the Tower Group report this came from to find out maybe?)
- 25% of reference data management is wasted due to duplication and inefficiences
- In their work with UBS Asset Management they had jointly shown that the cost of data management were reduced by 25-30% using a managed service (sounds worth verifying what the "before" situation was I guess, but interesting/impressive).
- Clients were pushing for much faster instrument setup and a reduction in time from the 1-2 weeks setup in some systems.
There were a few questions from the audience during Richard's talk, the first asked about the differences in doing data management with the buy-side and data management on the sell-side. Richard said that his experience was that the buy-side managed less instruments (<500,000) but with greater depth of data, and sell-side held more instruments (10M+) but with less depth of data (not sure that completely reflects my experience, but sounds worth a survey maybe).
The second question was why is the utility model for data management going to succeed right now, when previous attempts over the past 10 years had failed? Richard responded that he thought Tech Mahindra would succeed due to:
- Tech Mahindra are data-vendor agnostic (I assume aimed at Markit-Cadis and Bloomberg-PolarLake)
- Tech Mahindra own all their own IP (hmm, not really so sure this is a good reason or even a differentiator, but a I guess aimed at managed services that are not run by the firm that develops the data management system?)
I think the answers to this second question need thinking through more clearly, to be fair Richard had stated the 25% cost reduction already as one benefit, and various folks have said that the technology is ripe for these kinds of offerings now, but all the same the response need to be more fully developed to convince many I think (I remain undecided personally, it would be good to have some more evidence to back this up). One of the S&P CIQ added that what he thinks clients want is "Utility of Delivery" and not "Utility of Content" which I thought was a sensible comment and one that I will be revisiting in the coming months.
On a related note to why managed services just now, another audience member asked how client specific data was managed within a utility or managed service model, and Richard said that client specific data was often managed at the client but that they can upload and integrate client generated data into the managed service offering. I think this is a very key issue within the debate about managed services and utilities, I mean I get the point the data utility proponents make that certain datasets are simple "facts" as such are either write or wrong and hence commoditisable, but much of the data is subjective and all of the data needs validating together in the context of its intended use in my view. I guess I kind of loose myself in looping arguments about why data utility vendors aren't ultimately wanting to be the next Thomson Reuters or Bloomberg (not that that is not a laudible aim but it is not going to change the world or indeed financial markets data provision very much).
Posted by Brian Sentance | 8 March 2014 | 10:34 pm
Xenomorph is sponsoring the networking reception at the A-Team DMS event in London this week, and if you are attending then I wanted to extend a cordial invite to you to attend the drinks and networking reception at the end of day at 5:30pm on Thursday.
In preparation for Thursday’s Agenda then the blog links below are a quick reminder of some of the main highlights from last September’s DMS:
- Data Architecture: Sticks or Carrots?
- What Will Drive Data Management?
- Big Data, Cloud, In-Memory
- The Chief Data Officer Challenge
- Managed Services and the Utility Model
I will also be speaking on the 2pm panel “Reporting for the C-Suite: Data Management for Enterprise & Risk Analytics”. So if you like what you have heard during the day, come along to the drinks and firm up your understanding with further discussion with like-minded individuals. Alternatively, if you find your brain is so full by then of enterprise data architecture, managed services, analytics, risk and regulation that you can hardly speak, come along and allow your cerebellum to relax and make sense of it all with your favourite beverage in hand. Either way your you will leave the event more informed then when you went in...well that’s my excuse and I am sticking with it!
Hope to see you there!
Posted by Brian Sentance | 3 March 2014 | 6:33 pm
Very pleased that our partnering with Aqumin and their AlphaVision visual landscapes has been announced this week (see press release from Monday). Further background and visuals can be found at the following link and for those of you that like instant gratification please find a sample visual below showing some analysis of the S&P500.
Posted by Brian Sentance | 31 January 2014 | 7:04 pm
Quick plug for the New York version of F# in Finance event taking place next Wednesday December 11th, following on from the recent event in London. Don Syme of Microsoft Research will be demonstrating access to market data using F# and TimeScape. Hope to see you there!
Posted by Brian Sentance | 6 December 2013 | 7:49 am
Quick thank you to Don Syme of Microsoft Research for including a demonstration of F# connecting to TimeScape running on the Windows Azure cloud in the F# in Finance event this week in London. F# is functional language that is developing a large following in finance due to its applicability to mathematical problems, the ease of development with F# and its performance. You can find some testimonials on the language here.
Don has implemented a proof-of-concept F# type provider for TimeScape. If that doesn't mean much to you, then a practical example below will help, showing how the financial instrument data in TimeScape is exposed at runtime into the F# programming environment. I guess the key point is just how easy it looks to code with data, since effectively you get guided through what is (and is not!) available as you are coding (sorry if I sound impressed, I spent a reasonable amount of time writing mathematical C code using vi in the mid 90's - so any young uber-geeks reading this, please make allowances as I am getting old(er)...). Example steps are shown below:
Referencing the Xenomorph TimeScape type provider and creating a data context:
Connecting to a TimeScape database:
Looking at categories (classes) of financial instrument available:
Choosing an item (instrument) in a category by name:
Looking at the properties associated with an item:
The intellisense-like behaviour above is similar to what TimeScape's Query Explorer offers and it is great to see this implemented in an external run-time programming language such as F#. Don additionally made the point that each instrument only displays the data it individually has available, making it easy to understand what data you have to work with. This functionality is based on F#'s ability to make each item uniquely nameable, and to optionally to assign each item (instrument) a unique type, where all the category properties (defined at the category schema level) that are not available for the item are hidden.
The next event for F# in Finance will take place in New York on Wednesday 11th of December 2013 in New York, so hope to see you there. We are currently working on a beta program for this functionality to be available early in the New Year so please get in touch if this is of interest via email@example.com.
Posted by Brian Sentance | 27 November 2013 | 1:28 pm
An exciting departure from Xenomorph's typical focus on data management for risk in capital markets, but one of our partners, i2i Logic, has just announced the launch of their customer engagement platform for institutional and commercial banks based on Xenomorph's TimeScape. The i2i Logic team have a background in commercial banking, and have put together a platform that allows much greater interaction with a corporate client that a bank is trying to engage with.
Hosted in the cloud, and delivered to sales staff through an easy and powerful tablet app, the system enables bank sales staff to produce analysis and reports that are very specific to a particular client, based upon predictive analytics and models applied to market, fundamentals and operational data, initially supplied by S&P Capital IQ. This allows the bank and the corporate to discuss and understand where the corporate is when benchmarked against peers in a variety of metrics current across financial and operational performance, and to provide insight on where the bank's services may be able to assist in the profitability, efficiency and future growth of the corporate client.
Put another way, it sounds like the corporate customers of commercial banks are in not much better a position than us individuals dealing with retail banks, in that currently the offerings from the banks are not that engaging, generic and very hard to differentiate. Sounds like the i2i Logic team are on to something, so I wish them well in trying to move the industry's expectations of customer service and engagement, and would like to thank them for choosing TimeScape as the analytics and data management platform behind their solution.
Posted by Brian Sentance | 19 November 2013 | 2:17 pm
Another good event from PRMIA at the Harmonie Club here in NYC last week, entitled Risk Data Agregation and Risk Reporting - Progress and Challenges for Risk Management. Abraham Thomas of Citi and PRMIA introduced the evening, setting the scene by refering to the BCBS document Principles for effective risk data aggregation and risk reporting, with its 14 principles to be implemented by January 2016 for G-SIBs (Globally Systemically Important Banks) and December 2016 for D-SIBS (Domestically Systemically Important Banks).
The event was sponsored by SAP and they were represented by Dr Michael Adam on the panel, who gave a presentation around risk data management and the problems have having data siloed across many different systems. Maybe unsurprisingly Michael's presentation had a distinct "in-memory" focus to it, with Michael emphasizing the data analysis speed that is now possible using technologies such as SAP's in-memory database offering "Hana".
Following the presentation, the panel discussion started with a debate involving Dilip Krishna of Deloitte and Stephanie Losi of the Federal Reserve Bank of New York. They discussed whether the BCBS document and compliance with it should become a project in itself or part of existing initiatives to comply with data intensive regulations such as CCAR and CVA etc. Stephanie is on the board of the BCBS committee for risk data aggregation and she said that the document should be a guide and not a check list. There seemed to be general agreement on the panel that data architectures should be put together not with a view to compliance with one specific regulation but more as a framework to deal with all regulation to come, a more generalized approach.
Dilip said that whilst technology and data integration are issues, people are the biggest issue in getting a solid data architecture in place. There was an audience question about how different departments need different views of risk and how were these to be reconciled/facilitated. Stephanie said that data security and control of who can see what is an issue, and Dilip agreed and added that enterprise risk views need to be seen by many which was a security issue to be resolved.
Don Wesnofske of PRMIA and Dell said that data quality was another key issue in risk. Dilip agreed and added that the front office need to be involved in this (data management projects are not just for the back office in insolation) and that data quality was one of a number of needs that compete for resources/budget at many banks at the moment. Coming back to his people theme, Dilip also said that data quality also needed intuition to be carried out successfully.
An audience question from Dan Rodriguez (of PRMIA and Credit Suisse) asked whether regulation was granting an advantage to "Too Big To Fail" organisations in that only they have the resources to be able to cope with the ever-increasing demands of the regulators, to the detriment of the smaller financial insitutions. The panel did not completely agree with Dan's premise, arguing that smaller organizations were more agile and did not have the legacy and complexity of the larger institutions, so there was probably a sweet spot between large and small from a regulatory compliance perspective (I guess it was interesting that the panel did not deny that regulation was at least affecting the size of financial institutions in some way...)
Again focussing on where resources should be deployed, the panel debated trade-offs such as those between accuracy and consistency. The Legal Entity Identifier (LEI) initiative was thought of as a great start in establishing standards for data aggregation, and the panel encouraged regulators to look at doing more. One audience question was around the different and inconsistent treatment of gross notional and trade accounts. Dilip said that yes this was an issue, but came back to Stephanie's point that what is needed is a single risk data platform that is flexible enough to be used across multiple business and compliance projects. Don said that he suggests four "views" on risk:
- Risk Taking
- Risk Management
- Risk Measurement
- Risk Regulation
Stephanie added that organisations should focus on the measures that are most appropriate to your business activity.
The next audience question asked whether the panel thought that the projects driven by regulation had a negative return. Dilip said that his experience was yes, they do have negative returns but this was simply a cost of being in business. Unsurprisingly maybe, Stephanie took a different view advocating the benefits side coming out of some of the regulatory projects that drove improvements in data management.
The final audience question was whether the panel through the it was possible to reconcile all of the regulatory initiatives like Dodd-Frank, Basel III, EMIR etc with operational risk. Don took a data angle to this question, taking about the benefits of big data technologies applied across all relevant data sets, and that any data was now potentially valuable and could be retained. Dilip thought that the costs of data retention were continually going down as data volumes go up, but that there were costs in capturing the data need for operational risk and other applications. Dilip said that when compared globally across many industries, financial markets were way behind the data capabilities of many sectors, and that finance was more "Tiny Data" than "Big Data" and again he came back to the fact that people were getting in the way of better data management. Michael said that many banks and market data vendors are dealing with data in the 10's of TeraBytes range, whereas the amount of data in the world was around 8-900 PetaBytes (I thought we were already just over into ZetaBytes but what are a few hundred PetaBytes between friends...).
Abraham closed off the evening, firstly by asking the audience if they thought the 2016 deadline would be achieved by their organisation. Only 3 people out of around 50+ said yes. Not sure if this was simply people's reticence to put their hand up, but when Abraham asked one key concern for many was that the target would change by then - my guess is that we are probably back into the territory of the banks not implementing a regulation because it is too vague, and the regulators not being too prescriptive because they want feedback too. So a big game of chicken results, with the banks weighing up the costs/fines of non-compliance against the costs of implementing something big that they can't be sure will be acceptable to the regulators. Abraham then asked the panel for closing remarks: Don said that data architecture was key; Stephanie suggested getting the strategic aims in place but implementing iteratively towards these aims; Dilip said that deciding your goal first was vital; and Michael advised building a roadmap for data in risk.
Posted by Brian Sentance | 2 November 2013 | 1:48 pm
...Xenomorph!!! Thanks to all who voted for us in the recent A-Team Data Management Awards, it was great to win the award for Best Risk Data Management and Analytics Platform. Great that our strength in the Data Management for Risk field is being recognised, and big thanks again to clients, partners and staff who make it all possible!
Please also find below some posts for the various panel debates at the event:
- Data Architecture: Sticks or Carrots?
- What Will Drive Data Management?
- Big Data, Cloud, In-Memory
- The Chief Data Officer Challenge
- Managed Services and the Utility Model
Some photos, slides and videos from the event are now available on the A-Team site.
Posted by Brian Sentance | 24 October 2013 | 8:17 pm
Guest blog post by Qi Fu of PRMIA and Credit Suisse NYC with some notes on a model risk management event held ealier in September of this year. Big thank you to Qi for his notes and to all involved in organising the event:
The PRMIA event on Model Risk Management (MRM) was held in the evening of September 16th at Credit Suisse. The discussion was sponsored by Ernst & Young, and was organized by Cynthia Williams, Regulatory Coordinator for Americas at Credit Suisse.
As financial institutions have shifted considerable focus to model governance and independent model validation, MRM is as timely a topic as any in risk management, particularly since the Fed and OCC issued the Supervisory Guidance on Model Risk Management, also known as SR 11-7.
The event brings together a diverse range of views: the investment banks Morgan Stanley, Bank of American Merrill Lynch, and Credit Suisse are each represented, also on the panel are a consultant from E&Y and a regulator from Federal Reserve Bank of NY. The event was well attended with over 100 attendees.
Colin Love-Mason, Head of Market Risk Analytics at CS moderated the panel, and led off by discussing his 2 functions at Credit Suisse, one being traditional model validation (MV), the other being VaR development and completing gap assessment, as well as compiling model inventory. Colin made an analogy between model risk management with real estate. As in real estate, there are three golden rules in MRM, which are emphasized in SR 11-7: documentation, documentation, and documentation. Looking into the future, the continuing goals in MRM are quantification and aggregation.
Gagan Agarwala of E&Y’s Risk Advisory Practice noted that there is nothing new about many of the ideas in MRM. Most large institutions already have in place guidance on model validation and model risk management. In the past validation consisted of mostly quantitative analysis, but the trend has shifted towards establishing more mature, holistic, and sustainable risk management practices.
Karen Schneck of FRBNY’s Models and Methodology Department spoke about her role at the FRB where she is on the model validation unit for stress testing for Comprehensive Capital Analysis and Review (CCAR); thus part of her work was on MRM before SR 11-7 was written. SR 11-7 is definitely a “game changer”; since its release, there is now more formalization and organization around the oversight of MRM; rather than a rigid organization chart, the reporting structure at the FRB is much more open minded. In addition, there is an increased appreciation of the infrastructure around the models themselves and the challenges faced by practitioners, in particularly the model implementation component, which is not always immediately recognized.
Craig Wotherspoon of BAML Model Risk Management remarked on his experience in risk management, and comments that a new feature in the structure of risk governance is that model validation is turning into a component of risk management. In addition, the people involved are changing: risk professionals with the combination of a scientific mind, business sense, and writing skills will be in as high demand as ever.
Jon Hill, Head of Morgan Stanley’s Quantitative Analytics Group discussed his past experience in MRM since 90’s, when then the primary tools applied were “sniff tests”. Since then, the landscape has long been completely changed. In the past, focus had been on production, while documentation of models was an afterthought, now documentation must be detailed enough for highly qualified individual to review. In times past the focus was only around validating methodology, nowadays it is just as important to validate the implementation. There is an emphasis on stress testing, especially for complex models, in addition to internal threshold models and independent benchmarking. The definition of what a model is has also expanded to anything that takes numbers in and haves numbers as output. However, these increased demands require a substantial increase in resources; the difficulty of recruiting talent in these areas will remain a major challenge.
Colin noted a contrast in the initial comments of the panelists, on one hand some are indicating that MRM is mostly common sense; but Karen in particular emphasized the “game-changing” implications of SR 11-7, with MRM becoming more process oriented, when in the past it had been more of an intellectual exercise. With regards to recruitment, it is difficult to find candidates with all the prerequisite skill sets, one option is to split up the workload to make it easier to hire.
Craig noted the shift in the risk governance structure, the model risk control committees are defining what models are, more formally and rigorously. Gagan added that models have lifecycles, and there are inherent risks associated within that lifecycle. It is important to connect the dots to make sure everything is conceptually sound, and to ascertain that other control functions understand the lifecycles.
Karen admits that additional process requirements contain the risk of trumping value. MRM should aim to maintain high standards while not get overwhelmed by the process itself, so that some ideas become too expensive to implement. There is also the challenge of maintaining independence of the MV team.
Jon concurred with Karen on the importance of maintaining independence. A common experience is when validators find mistakes in the models, they become drawn into the development process with the modelers. He also notes differences with the US, UK, and European MV processes, and Jon asserts his view that the US is ahead of the curve and setting standards.
Colin noted the issue of the lack of an analogous PRA document to SR 11-7, that drills down into nuts and bolts of the challenges in MRM. He also concurred on the difficulty of maintaining independence, particularly in areas with no established governance. It is important to get model developers to talk to other developers about the definition and scope of the models, as well as possible expansion of scope. There is a wide gamut of models: core, pricing, risk, vendor, sensitivity, scenarios, etc. Who is responsible for validating which? Who checks on the calibration, tolerance, and weights of the models? These are important questions to address.
Craig commented further on the complexity and uncertainty of defining what a model is, and on whose job it is to determine that, amongst the different stakeholders. It also needs to be taken into consideration that model developers maybe biased towards limiting the number of models.
Gagan followed up by noting that while the generic definition of models is broad, and will need to be redefined, but analytics do not all need to have the same standards, the definition should leave some flexibility for context. Also, the highest standard should be assigned to risk models.
Karen adds that, defining and validating models used to have a narrow focus, and done in a tailor-controlled environment. It would be better to broaden the scope, and to reexamine the question on an ongoing basis (it is however important to point out that annual review does not equal annual re-validation). In addition to the primary models, some challenge models also need to be supported; developers should discuss why they’re happy with primary model, how it is different from challenger model, and how it impacts output.
Colin brought up the point of stress-testing. Jon asserts that stress-testing is more important for stochastic models, which are more likely to break under nonsensical inputs. Also any model that plugs into the risk system should require judicious decision-making, as well as annual reviews to look at changes since the previous review.
Colin also brought up the topic of change management: what are the system challenges when model developers release code, which may include experimental releases. Often discussed are concepts of annual certification and checkpoints. Jon commented that the focus should be on changes of 5% or more, with pricing model being less of a priority; and firms should move towards centralized source code depositories.
Karen also added the question of what ought to considered material change: the more conservative answer is any variation, even if a pure code change that didn’t change model usage or business application, may need to be communicated to upper management.
Colin noted that developers often have a tendency to encapsulate intentions, and have difficulty or reluctance to document changes, thus resulting in many grey areas. Gagan added that infrastructure is crucial. Especially when market conditions are rapidly changing, MRM need to have controls that are in place. Also, models are in Excel make the change management process more difficult.
The panel discussion was followed by a lively Q&A session with an engaged audience, below are some highlights.
Q: How do you distinguish between a trader whose model actually needs change, versus a trader who is only saying so because he/she has lost money?
Colin: Maintain independent price verification and control functions.
Craig: Good process for model change, and identify all stakeholders.
Karen: Focus on what model outputs are being changed, what the trader’s assumptions are, and what is driving results.
Q: How do you make sure models are used in business in a way that makes sense?
Colin: This can be difficult, front office builds the models, states what is it good for, there is no simple answer from the MV perspective; usage means get as many people in the governance process as possible, internal audit and setting up controls.
Gagan: Have coordination with other functions, holistic MRM.
Karen: Need structure, inventory a useful tool for governance function.
Q: Comments on models used in the insurance industry?
Colin: Very qualitative, possible to give indications, difficult to do exact quantitative analysis, estimates are based on a range of values. Need to be careful with inputs for very complex models, which can be based on only a few trades.
Q: What to do about big shocks in CCAR?
Jon: MV should validate for severe shocks, and if model fails may need only simple solution.
Karen: Validation tools, some backtesting data, need to benchmark, quant element of stress testing need to substantiated and supported by qualitative assessment.
Q: How to deal with vendor models?
Karen: Not acceptable just to say it’s okay as long as the vendor is reputable, want to see testing done, consider usage also compare to original intent.
Craig: New guidance makes it difficult to buy vendors models, but if vendor recognizes this, this will give them competitive advantage.
Q: How to define independence for medium and small firms?
Colin: Be flexible with resources, bring in different people, get feedback from senior management, and look for consistency.
Jon: Hire E&Y? There is never complete independence even in a big bank.
Gagan: Key is the review process.
Karen: Consultants could be cost effective; vendor validation may not be enough.
Q: At firm level, do you see practice of assessing risk models?
Jon: Large bank should appoint Model Risk Officer.
Karen: Just slapping on additional capital is not enough
Q: Who actually does MV?
Colin: First should be user, then developer, 4 eyes principle.
Q: Additional comments on change management?
Colin: Ban Excel for anything official; need controlled environment.
Posted by Brian Sentance | 24 October 2013 | 6:59 pm
Very pleased to announce our new data integration for TimeSCape with S&P Capital IQ - see the press release
Posted by Brian Sentance | 21 October 2013 | 2:25 pm
Great event from PRMIA on Tuesday evening of last week, entitled Credit Risk: The link between Loss Given Default and Default. The event was kicked off by Melissa Sexton of PRMIA, who introduced Jon Frye of the Federal Reserve Bank of Chicago. Jon seems to an acknowledged expert in the field of Loss Given Default (LGD) and credit risk modelling. I am sure that the slides will be up on the PRMIA event page above soon, but much of Jon's presentation seems to be around the following working paper. So take a look at the paper (which is good in my view) but I will stick to an overview and in particular any anecdotal comments made by Jon and other panelists.
Jon is an excellent speaker, relaxed in manner, very knowledgeable about his subject, humourous but also sensibly reserved in coming up with immediate answers to audience questions. He started by saying that his talk was not going to be long on philosophy, but very pragmatic in nature. Before going into detail, he outlined that the area of credit risk can and will be improved, but that this improvement becomes easier as more data is collected, and inevitably that this data collection process may need to run for many years and decades yet before the data becomes statistically significant.
Which Formula is Simpler? Jon showed two formulas for estimating LGD, one a relatively complex looking formula (the Vasicek distribution mentioned his working paper) and the other a simple linear model of the a + b.x. Jon said that looking at the two formulas, then many would hope that the second formula might work best given its simplicity, but he wanted to convince us that the first formula was infact simpler than the second. He said that the second formula would need to be regressed on all loans to estimate its parameters, whereas the first formula depended on two parameters that most banks should have a fairly good handle on. The two parameters were Default Rate (DR) and Expected Loss (EL). The fact that these parameters were relatively well understood seemed to be the basis for saying the first formula was simpler, despite its relative mathematical complexity. This prompted an audience question on what is the difference between Probability of Default (PD) and Default Rate (DR). Apparently it turns out PD is the expected probability of default before default happens (so ex-ante) and DR is the the realised rate of default (so ex-post).
Default and LGD over Time. Jon showed a graph (by an academic called Altman) of DR and LGD over time. When the DR was high (lots of companies failing, in a likely economic downtown) the LGD was also perhaps understandably high (so high number of companies failing, in an economic background that is both part of the causes of the failures but also not helping the loss recovery process). When DR is low, then there is a disconnect between LGD and DR. Put another way, when the number of companies failing is low, the losses incurred by those companies that do default can be high or low, there is no discernable pattern. I guess I am not sure in part whether this disconnect is due to the smaller number of companies failing meaning the sample space is much smaller and hence the outcomes are more volatile (no averaging effect), or more likely that in healthy economic times the loss given a default is much more of random variable, dependent on the defaulting company specifics rather than on general economic background.
Conclusions Beware: Data is Sparse. Jon emphasised from the graph that the Altman data went back 28 years, of which 23 years were periods of low default, with 5 years of high default levels but only across 3 separate recessions. Therefore from a statistical point of view this is very little data, so makes drawing any firm statistical conclusions about default and levels of loss given default very difficult and error-prone.
The Inherent Risk of LGD. Jon here seemed to be focussed not on the probability of default, but rather on the conditional risk that once a default has occurred then how does LGD behave and what is the risk inherent from the different losses faced. He described how LGD affects i) Economic Capital - if LGD is more variable, then you need stronger capital reserves, ii) Risk and Reward - if a loan has more LGD risk, then the lender wants more reward, and iii) Pricing/Valuation - even if the expected LGD of two loans is equal, then different loans can still default under different conditions having different LGD levels.
Models of LGD.
Jon showed a chart will LGC plotted against DR for 6 models (two of which I think he was involved in). All six models were dependent on three parameters, PD, EL and correlation, plus all six models seemed to produce almost identical results when plotted on the chart. Jon mentioned that one of his models had been validated (successfully I think, but with a lot of noise in the data) against Moody's loan data taken over the past 14 years. He added that he was surprised that all six models produced almost the same results, implying that either all models were converging around the correct solution or in total contrast that all six models were potentially subject to "group think" and were systematically all wrong in the ways the problem should be looked at.
Jon took one of his LGD models and compared it against the simple linear model, using simulated data. He showed a graph of some data points for what he called a "lucky bank" with the two models superimposed over the top. The lucky bit came in since this bank's data points for DR against LGD showed lower DR than expected for a given LGD, and lower LGD for a given DR. On this specific case, Jon said that the simple linear model fits better than his non-linear one, but when done over many data sets his LGD model fitted better overall since it seemed to be less affected by random data.
There were then a few audience questions as Jon closed his talk, one leading Jon to remind everyone of the scarcity of data in LGD modelling. In another Jon seemed to imply that he would favor using his model (maybe understandably) in the Dodd-Frank Annual Stress Tests for banks, emphasising that models should be kept simple unless a more complex model can be justified statistically.
Steve Bennet and the Data Scarcity Issue
Following Jon's talk, Steve Bennet of PECDC picked on Jon's issue of scare data within LGD modelling. Steve is based in the US, working for his organisation PECDC which is a cross border initiative to collect LGD and EAD (exposure at default) data. The basic premise seems to be that in dealing with the scarce data problem, we do not have 100 years of data yet, so in the mean time lets pool data across member banks and hence build up a more statistically significant data set - put another way: let's increase the width of the dataset if we can't control the depth.
PECDC is a consortia of around 50 organisations that pool data relating to credit events. Steve said that capture data fields per default at four "snapshot" times: orgination, 1 year prior to default, at default and at resolution. He said that every bank that had joined the organisation had managed to improve its datasets. Following an audience question, he clarified that PECDC does not predict LGD with any of its own models, but rather provides the pooled data to enable the banks to model LGD better.
Steve said that LGD turns out to be very different for different sectors of the market, particularly between SMEs and large corporations (levels of LGD for large corporations being more stable globally and less subject to regional variations). But also there is great LGD variation across specialist sectors such as aircraft finance, shipping and project finance.
Steve ended by saying that PECDC was orginally formed in Europe, and was now attempting to get more US banks involved, with 3 US banks already involved and 7 waiting to join. There was an audience question relating to whether regulators allowed pooled data to be used under Basel IRB - apparently Nordic regulators allow this due to needing more data in a smaller market, European banks use the pooled data to validate their own data in IRB but in the US banks much use their own data at the moment.
Following Steve, Til Schuermann added his thoughts on LGD. He said that LGD has a time variation and is not random, being worse in recession when DR is high. His stylized argument to support this was that in recession there are lots of defaults, leading to lots of distressed assets and that following the laws of supply and demand, then assets used in recovery would be subject to lower prices. Til mentioned that there was a large effect in the timing of recovery, with recovery following default between 1 and 10 quarters later. He offered words of warning that not all defaults and not all collateral are created equal, emphasising that debt structures and industry stress matter.
The evening closed with a few audience questions and a general summation by the panelists of the main issues of their talks, primarily around models and modelling, the scarcity of data and how to be pragmatic in the application of this kind of credit analysis.
Posted by Brian Sentance | 20 October 2013 | 7:16 pm
Andrew Delaney introduced the final panel of the day, involving Steve Cheng of Rimes, Jonathan Clark of Tech Mahindra, Tom Dalglish of UBS and Martijn Groot of Euroclear. Main points:
- Andrew started by asking the panel for their definitions of managed data services and data utilities
- Martijn said that a managed data service was usually the lifting out of a data process from in a company to be run by somebody else whereas a data utility had many users.
- Tom put it another way saying that a managed service was run for you whereas a utility was run for them. Tom suggested that there were some concerns around data utilities for the industry in terms of knowing/being transparent about data vendor affinity and any data monopoly aspects.
- When asked why past attempts at data utilities had failed, Tom said that it must be frustrating to be right but at wrong time, but in addition to the timing being right just now (costs/regulations being drivers) then the tech stack available is better and the appreciation of data usage importance is clearer.
- Steve added a great point on the tech stack, in that it now made mass customisation much easier.
- Jonathan made the point that past attempts at data utilities were built on product platforms used at clients, whereas the latest utilities were built on platforms specifically designed for use by a data utility.
- Looking at the cost savings of using a data utility, Martijn said that the industry spends around $16-20B on data, and that with his Euroclear data utility they can serve 2000 clients with a staff level that is less than any one client employs directly.
- Tom said that the savings from collapsing the data silos were primarily from more efficient/reduced usage of people and hardware to perform a specific function, and not data.
- Steve suggested that some utilities take an incremental data services and not take all data as in the old utility model, again coming back to his earlier point of mass customisation.
- Tom mentioned it was a bit like cable TV, where you can subscribe to a set of services of your choice but where certain services cost more than others.
- Martijn said that there were too many vested interests to turn data costs around quickly. He said that data utilities could go a long way however.
- Tom concluded by saying that it was about content not feeds, licensing was important as was how to segregate data.
Good panel - additionally one final audience question/discussion was around data utilities providing LEI data, and it was argued that LEI without the hierarchy is just another set of data to map and manage.
Posted by Brian Sentance | 7 October 2013 | 12:28 pm
The first panel of the afternoon touched on a hot topic at the moment, the role of the Chief Data Officer (CDO). Andrew Delaney again moderated the panel, consisting of Rupert Brown of UBS, Patrick Dewald of Diaku, Colin Hall of Credit Suisse, Nigel Matthews of Barclays and Neill Vanlint of GoldenSource. Main points:
- Colin said that the need for the CDO role is that someone needs to sit at the top table who is both nerdy about data but also can communicate a vision for data to the CEO.
- Rupert said that role of CDO was still a bit nebulous covering data conformance, storage management, security and data opportunity (new functionality and profit). He suggested this role used to be called "Data Stewardship" and that the CDO tag is really a rename.
- Colin answered that the role did use to be a junior one, but regulation and the rate of industry change demands a CDO, a point contact for everyone when anything comes up that concerns data - previously nobody knew quite who to speak to on this topic.
- Patrick suggested that a CDO needs a long-term vision for data, since the role is not just an operational one.
- Nigel pointed out that the CDO needs to cover all kinds of data and mentioned recent initiatives like BCBS with their risk data aggregation paper.
- Neil said that he had seen the use of a CDO per business line at some of his clients.
- There was some conversation around the different types of CDO and the various carrots and sticks that can be employed. Neil made the audience laugh with his quote from a client that "If the stick doesn't work, I have a five-foot carrot to hit them with!"
- Patrick said that CDO role is about business not just data.
- Colin picked up on what Patrick said and illustrated this with an example of legal contract data feeding directly into capital calculations.
- Nigel said that the CDO is a facilitator with all departments. He added that the monitoring tools from market data where needed in reference data
Overall good debate, and I guess if you were starting from scratch (if only we could!) you would have to think that the CDO is a key role given the finance industry is primarily built on the flow of data from one organisation to another.
Posted by Brian Sentance | 7 October 2013 | 12:26 pm
Andrew Delaney introduced the second panel of the day, with the long title of "The Industry Response: High Performance Technologies for Data Management - Big Data, Cloud, In-Memory, Meta Data & Big Meta Data". The panel included Rupert Brown of UBS, John Glendenning of Datastax, Stuart Grant of SAP and Pavlo Paska of Falconsoft. Andrew started the panel by asking what technology challenges the industry faced:
- Stuart said that risk data on-demand was a key challenge, that there was the related need to collapse the legacy silos of data.
- Pavlo backed up Stuart by suggesting that accuracy and consistency were needed for all live data.
- Rupert suggested that there has been a big focus on low latency and fast data, but raised a smile from the audience when he said that he was a bit frustrated by the "format fetishes" in the industry. He then brought the conversation back to some fundamentals from his viewpoint, talking about wholeness of data and namespaces/data dictionaries - Rupert said that naming data had been too stuck in the functional area and not considered more in isolation from the technology.
- John said that he thought there were too many technologies around at the moment, particularly in the area of Not Only SQL (NoSQL) databases. John seemed keen to push NoSQL, and in particular Apache Cassandra, as post relational databases. He put forward that these technologies, developed originally by the likes of Google and Yahoo, were the way forward and that in-memory databases from traditional database vendors were "papering over the cracks" of relational database weaknesses.
- Stuart countered John by saying that properly designed in-memory databases had their place but that some in-memory databases had indeed been designed to paper over the cracks and this was the wrong approach, exascerbating the problem sometimes.
- Responding to Andrew's questions around whether cloud usage was more accepted by the industry than it had been, Rupert said he thought it was although concerns remain over privacy and regulatory blockers to cloud usage, plus there was a real need for effective cloud data management. Rupert also asked the audience if we knew of any good release management tools for databases (controlling/managing schema versioning etc) because he and his group were yet to find one.
- Rupert expressed that Hadoop 2 was of more interest to him at UBS that Hadoop, and as a side note mentioned that map reduce was becoming more prevalent across NoSQL not just within the Hadoop domain. Maybe controversially, he said that UBS was using less data than it used to and as such it was not the "big data" organisation people might think it to be.
- As one example of the difficulties of dealing with silos, Stuart said that at one client it required the integration of data from 18 different system to a get an overall view of the risk exposure to one counterparty. Stuart advocated bring the analytics closer to the data, enabling more than one job to be done on one system.
- Rupert thought that Goldman Sachs and Morgan Stanley seem to do what is the right thing for their firm, laying out a long-term vision for data management. He said that a rethink was needed at many organisations since fundamentally a bank is a data flow.
- Stuart picked up on this and said that there will be those organisations that view data as an asset and those that view data as an annoyance.
- Rupert mentioned that in his view accountants and lawyers are getting in the way of better data usage in the industry.
- Rupert added that data in Excel needed to passed by reference and not passed by value. This "copy confluence" was wasting disk space and a source of operational problems for many organisations (a few past posts here and here on this topic).
- Moving on to describe some of the benefits of semantic data and triple stores, Rupert proposed that the statistical world needed to be added to the semantic world to produce "Analytical Semantics" (see past post relating to the idea of "analytics management").
Great panel, lots of great insight with particularly good contributions from Rupert Brown.
Posted by Brian Sentance | 7 October 2013 | 12:23 pm
The first panel of the day opened with an introductory talk by Chris Johnson of HSBC. Chris started his talk by proudly announcing that he drives a Skoda car, something that to him would have been unthinkable 25 years ago but with investment, process and standards things can and will change. He suggested that data management needs to go through a similar transformation, but that there remained a lot to be done.
Moving on to the current hot topics of data unitilities and managed services, he said that reduced costs of managed services only became apparent in the long term and that both types of initiative have historically faced issues with:
- Logistical Challenges and Risks
Chris made the very good point that until service providers accept liability for data quality then this means that clients must always check the data they use. He also mentioned that in relation to Solvency II (a hot topic for Chris at HSBC Security Services), that EIOPA had recently mentioned that managed services may need to be regulated. Chris mentioned the lack of time available to respond to all the various regulatory deadlines faced (a recurring theme) and that the industry still lacked some basic fundamentals such as a standard instrument identifier.
Chris then joined the panel discussion with Andrew Delaney as moderator and with other panelists including Colin Gibson (see previous post), Matt Cox of Denver Perry, Sally Hinds of Data Management Consultancy Services and Robert Hofstetter of Bank J. Safra Sarasin. The key points I took from the panel are outlined below:
- Sally said that many firms were around Level 3 in the Data Management Maturity Model, and that many were struggling particularly with data integration. Sally added that utililities were new, as was the CDO role and that implications for data management were only just playing out.
- Matt thought that reducing cost was an obvious priority in the industry at the moment, with offshoring playing its part but progress was slow. He believed that data management remains underdeveloped with much more to be done.
- Colin said that organisations remain daunted by their data management challenges and said that new challenges for data management with transactional data and derived data.
- Sally emphasised the role of the US FATCA regulation and how it touches upon some many processess and departments including KYC, AML, Legal, Tax etc.
- Matt highlighted derivatives regulation with the current activity in central clearing, Dodd-Frank, Basel III and EMIR.
- Chris picked up on this and added Solvency II into the mix (I think you can sense regulation was a key theme...). He expressed the need and desirability of a Unique Product Identifier (UPI see report) as essential for the financial markets industry and how we need not just stand still now the LEI was coming. He said that industry associations really needed to pick up their game to get more standards in place but added that the IMA had been quite proactive in this regard. He expressed his frustration at current data licensing arrangements with data vendors, with the insistence on a single point of use being the main issue (big problem if you are in security services serving your clients I guess)
- Robert added that his main issues were data costs and data quality
- Andrew then brought the topic around to risk management and its impact on data management.
- Colin suggested that more effort was needed to understand the data needs of end users within risk management. He also mentioned that products are not all standard and data complexity presents problems that need addressing in data management.
- Chris mentioned that there 30 data fields used in Solvency II calculations and that if any are wrong this would have a direct impact on the calcualated capital charge (i.e. data is important!)
- Colin got onto the topic of unstructured data and said how it needed to be tagged in some way to become useful. He suggested that there was an embrionic cross-over taking place between structured and unstructured data usage.
- Sally thought that the merging of Business Intelligence into Data Management was a key development, and that if you have clean data then use it as much as you can.
- Robert thought that increased complexity in risk management and elsewhere should drive the need for increased automation.
- Colin thought cost pressures mean that the industry simply cannot afford the old IT infrastructure and that architecture needs to be completely rethought.
- Chris said that we all need to get the basics right, with LEI but then on to UPI. He said to his knowledge data management will always be a cost centre and standardisation was a key element of reducing costs across the industry.
- Sally thought that governance and ownership of data was wooly at many organisations and needed more work. She added this needed senior sponsorship and that data management was an ongoing process, not a one-off project.
- Matt said that the "stick" was very much needed in addition to the carrot, advising that the proponents of improved data management should very much lay out the negative consequences to bring home the reality to business users who might not see the immediate benefits and costs.
Overall good panel, lots of good debate and exchanging of ideas.
Posted by Brian Sentance | 7 October 2013 | 12:17 pm
Great day on Thursday at the A-Team Data Management Summit in London (personally not least because Xenomorph won the Best Risk Data Management/Analytics Platform Award but more of that later!). The event kicked off with a brief intro from Andrew Delaney of the A-Team talking through some of the drivers behind the current activity in data management, with Andrew saying that risk and regulation were to the fore. Andrew then introduced Colin Gibson, Head of Data Architecture, Markets Division at Royal Bank of Scotland.
Data Architecture - Sticks or Carrots? Colin began by looking at the definition of "data architecture" showing how the definition on Wikipedia (now obviously the definitive source of all knowledge...) was not particularly clear in his view. He suggested himself that data architecture is composed of two related frameworks:
- Orderly Arrangement of Parts
He said that the orderly arrangement of parts is focussed on business needs and aims, covering how data is sourced, stored, referenced, accessed, moved and managed. On the discipline side, he said that this covered topics such as rules, governance, guides, best practice, modelling and tools.
Colin then put some numbers around the benefits of data management, saying that for every dollar spend on centralising data saves 20 dollars, and mentioning a resulting 80% reduction in operational costs. Related to this he said that for every dollar spent on not replicating data saved a dollar on reconcilliation tools and a further dollar saved on the use of reconcilliation tools (not sure how the two overlap but these are obviously some of the "carrots" from the title of the talk).
Despite these incentives, Colin added that getting people to actually use centralised reference data remains a big problem in most organisations. He said he thought that people find it too difficult to understand and consume what is there, and faced with a choice they do their own thing as an easier alternative. Colin then talked about a program within RBS called "GoldRush" whereby there is a standard data management library available to all new projects in RBS which contains:
- messaging standards
- standard schema
- update mechanisms
The benefit being that if the project conforms with the above standards then they have little work to do for managing reference data since all the work is done once and centrally. Colin mentioned that also there needs to be feedback from the projects back to central data management team around what is missing/needing to be improved in the library (personally I would take it one step further so that end-users and not just IT projects have easy discovery and access to centralised reference data). The lessons he took from this were that we all need to "learn to love" enterprise messaging if we are to get to the top down publish once/consume often nirvana, where consuming systems can pick up new data and functionality without significant (if any) changes (might be worth a view of this post on this topic). He also mentioned the role of metadata in automating reconcilliation where that needed to occur.
Colin then mentioned that allocation of costs of reference data to consumers is still a hot topic, one where reference data lags behind the market data permissioning/metering insisted upon by exchanges. Related to this Colin thought that the role of the Chief Data Officer to enforce policies was important, and the need for the role was being driven by regulation. He said that the true costs of a tactical, non-standard approach need to be identifiable (quantifying the size of the stick I guess) but that he had found it difficult to eliminate the tactical use of pricing data sourced for the front office. He ended by mentioning that there needs to be a coming together of market data and reference data since operations staff are not doing quantitative valuations (e.g. does the theoretical price of this new bond look ok?) and this needs to be done to ensure better data quality and increased efficiency (couldn't agree more, have a look at this article and this post for a few of my thoughts on the matter). Overall very good speaker with interesting, practical examples to back up the key points he was trying to get across.
Posted by Brian Sentance | 7 October 2013 | 12:12 pm