CSDR: Using Predictive Analytics to Prevent Fails

Predict and prevent– that’s what back-office operations managers at buy and sell-side firms preparing for Europe’s Central Securities Depositories Regulation (CSDR) settlement discipline regime are starting to think about so they won’t have to pay hefty financial penalties or endure buy-in requirements if they fail to settle their European trades on time.

Instead of bemoaning the cost of a failed settlement, some firms are relying on statistical models based on historical data to calculate the probability a trade will fail to settle on time. They want to fix the problem before it costs them a penalty or forces them to execute a buy-in. “Knowing the costs for a failed trade have increased, financial firms are starting to consider predictive analytics tools that will allow them to avoid a settlement fail,” says Alexander Duggan, head of platform brokerage fees and billing in London for financial technology firm Cognizant. In recently acquiring Meritsoft, Cognizant added its CSDR predictive fail calculation engine to Meritsoft’s penalty calculation engine.

Middle and back-office operations managers at about a dozen fund management firms, custodian banks, and broker-dealers contacted by FinOps Report over the past month agree that their firms are still spending far more time worried about how they will calculate their penalties and perform buy-ins than how they will prevent settlement fails. However, they expect that stance will change within a few weeks once their trading desks start complaining about how much money they stand to lose from penalties and mandatory buy-ins. The increase in trading volumes coupled with working remotely have raised the potential for settlement fails, leaving back-offices to eventually scramble to implement either internal or third-party preventative solutions by the time the CSDR’s settlement fails penalty rules kick in during February 2021. “It’s going to be a rude awakening, if fail rates start to climb,” one European fund management operations manager tells FinOps Report.

The CSDR’s reach extends far beyond strictly European firms. It doesn’t matter who executed the order or where it was executed. As long as the trade involves a European listed firm and is settled in a national European securities depository, it falls under the CSDR’s rules which harmonize the methodology used by each European securities depository to penalize bank or broker-dealer members failing to settle their trades within two days after they were executed. The amount of the fine depends on the value of the affected transaction and the liquidity of the financial instrument involved. As a rule of thumb, transactions involving government bonds will incur the lowest fines while those involving corporate bonds will incur the highest.

Every bank or broker-dealer member of each European securities depository will end up at the end of the month receiving a net debit or credit based on the extent to which it was responsible for trades failing to settle on time or is the wronged party. Those debits or credits will ultimately be passed along to fund manager clients. Even worse than a penalty for a failed settlement, say operations managers, is the new buy-in requirement if the failed settlement isn’t fixed within either four days for liquid assets, such as equities. or seven days for illiquid assets such as corporate bonds. The buyer executing the buy-in will be compensated by the seller for the difference between the value of the shares at the time the sale was originally struck and the new price at the time of the buy-in. The compensation only applies if the share price has increased. In the case of a transaction processed through a central clearinghouse, the clearinghouse is responsible for the buy-in and for compensating the buyer. The UK’s regulatory watchdog Financial Conduct Authority has said that the UK will not be following the CSDR’s buy-in rules, but that decision only applies to transactions settled through the UK’s central securities depository; financial penalties must still be imposed for failed settlements of UK transactions.

Until now, custodian banks, broker-dealers and fund managers have focused their compliance efforts regarding CSDR with having the right systems in place to calculate penalties and to know when a buy-in must be executed. Members of international securities depository Euroclear Bank as well as Euroclear’s national depositories in France, the Netherlands and Belgium can rely using a platform from Taskize to interact with the depositories more efficiently so they can reduce the time it takes to correct failed trades. However, the platform does not predict the probability a trade will fail to settle on time.

There are no official figures on the percentage of failed trades by European market or asset class, but the pan-European securities regulatory agency European Securities and Markets Association says that failed trades account for about three percent of the value of trades in corporate bonds and sovereign debt markets and six percent in equities markets. The London-based trade group International Capital Markets Association (ICMA) has predicted that a European equity trade with an average commission of 2.5 basis points could face a fine of one basis point per day for each day the trade fails to settle on time which means that a delay of two days could wipe out almost all profits. The ICMA has also warned that the  CSDR’s settlement fails penalty regime will result in European bond spreads widening, particularly for high-yield bonds; for every E1 trillion in annual trade volume investors could end up paying E1.4 billion in wider spreads says the trade group.

Based on FinOps Report’s conversations with settlement managers at over a dozen European financial services firms taking proactive steps to address the CSDR’s settlement fails penalty phase, the approach appears to be for back-office operations manager eyeballing the historic number of trades that failed to settle on time, determining the reasons why, and figuring out who was to blame. Should they decide that a particular counterparty is causing the most fails, they will take the carrot and stick approach. Middle-office managers will start off gently persuading the counterparty that fixing it back-office systems to ensure trade information is matched quickly will benefit the relationship in some fashion. If that talk doesn’t work, the next step will be to reduce the number of trades done with the counterparty. In a worst-case scenario, if all else fails, the relationship will end. Yet another means of lowering the number of trades that fail to settle on time is to ensure that trade confirmations and allocations are done on an automated basis as close to trade date as possible. Lastly, financial firms can always focus on verifying the accuracy of their standing settlement instructions.

However, even the most diligent preventative steps won’t be enough to make a sizeable dent in the percentage of settlement fails if they rely on manual guesswork. Popular with trading desks and collateral management units, the discipline of predictive analytics is now popping up in the back office. Meritsoft’s CSDR platform measures the likelihood a trade will fail to settle on time through a percentage of one to one hundred. “The methodology used by Meritsoft is based on logistic regression in which historical data on settlement fails is fed into an machine-learning algorithm to determine the probability that a trade in a particular security will fail to settle on a particular settlement date,” explains Cognizant’s Duggan. The age of the historical data depends on the client’s requirements, but is ideally between twelve and twenty-four months old; the data includes information on trade booking and enrichment mismatches, issuer corporate actions, and availability of assets. The percentage probability of a settlement fail is then compared to the industry-wide fail rate for that particular security. “Not all probabilities for failed settlements will be investigated,” says Duggan. “Instead, the firm is likely to focus on those which are higher than the industry-wide rate.” He won’t say how many firms are using the predictive analytics module, but insists that interest is growing.

Global professional services firm Accenture is offering a predictive analytics tool to clients of its post-trade back-office operations practice Accenture Post-Trade Processing (APTP). Using artificial intelligence, machine learning and historical data, the tool allows settlement managers to determine the probability the trade will fail to settle on time, the reasons, and steps to avoid the fail. “The probability of failed settlement is coupled with information on the potential penalty or buy-in risks which will allow back-office managers to prioritize which trades they must take action on first,” says the Paris-based Thierry Weidenmann, Accenture’s managing director and lead for APTP. He would not specify the number of firms using Accenture’s APTP, but says it services more than 70 markets supported by centers in the UK, France and Asia.

Some custodian banks are also jumping on the predictive analytics bandwagon leveraging their relationships as middle office outsourcing agents. In 2018 BNP Paribas Securities Services launched a predictive analytics platform called Smart Chaser which relies on a “Random Forest” algorithm to identify the subset of trades that might be in danger of failing, to suggest the reasons why, and to propose the actions needed to correct the problem. The “Random Forest” algorithm, which uses 100 different factors, incorporates multiple decision-making trees to generate its predictions. The model is updated daily with data from the previous three months to adjust the weights assigned to different factors to increase the accuracy of its predictions. The factors include the broker’s history, the time the trade was executed, the value of the trade, location of the counterparties, whether the trade is part of a block trade or single allocation, and the geographic location of the counterparties.

“The Random Forest algorithm gives the asset manager user in the middle office a prioritization of the trades that will fail to match or settle on time, calculating a percentage of fail probability.” says Valerie Bolea-Waeterloos, head of global middle office product for BNP Paribas Securities Services in Paris. “An e-mail will also automatically be sent from the asset manager’s middle office to the executing broker-dealer’s middle office indicating there is a potential the trade will fail to match or settle on time.” The platform does not predict the potential settlement fail penalty if no remedial action is taken. About 80 asset manager clients of BNP Paribas Securities Services’ middle-office services unit across the globe completing a total of about four million trades annually will benefit from using Smart Chaser, says Bolea-Waeterloos citing an accuracy rate of 98 percent. BNP Paribas Securities Services ultimately hopes to measure its impact on clients by calculating the percentage of the reduction in settlement fails.

Regardless of the methodology financial firms use to predict settlement fails, European back-office operations managers who spoke with FinOps Report acknowledge they will need to start taking drastic action quickly. Just as important as implementing the right forecasting technology, they add, is using the right technology to work from home. Fail rates in some Asian countries rose sharply in March and April due to higher trading volumes and the temporary shutdowns of offshore processing hubs with no back-up sites immediately available. None of the predictive analytics tools, acknowledge their corporate operators, can incorporate office closures due to the global pandemic because it is a statistical outlier. They agree that no one can afford such a wild card causing additional settlement fails. “Hopefully, financial firms will have learned their lesson,” says Cognizant’s Duggan.