Blog article
See all stories »

Un-’block’ your business 1.0: Combining Machine Learning and Blockchain for Capital Efficiency

This is the first of my series of blogs on how businesses can use the power of Blockchain technology to un-‘block’ the inefficiencies in their current models and practices. Almost every paper or blog you will read on the technology will start with Satoshi Nakamoto and explain what Blockchain technology basics are. I am going to skip the basics and dive directly into its applications. If you are a Blockchain junkie in the financial industry like I am, you are sure to have read a number of papers on how one of its greatest application is in Securities Clearing and Settlement (yes of course cross border payments and trade finance as well) and how many FinTech companies have already sprouted to solve this exact problem statement and will dis-intermediate clearing houses. Even the big banks have joined hands to do exactly this (see recent announcement Six global banks join forces to create digital currency). Equally, you will find enough objections and constraints raised by the organizations that will get dis-intermediated (or not) with the implementation of Blockchain – for e.g. the clearing houses. So, as part my 15,000 words long MBA dissertation, I went about researching and interviewing people to understand both sides of the story.

The opportunity

A little bit of basics for the uninitiated on securities clearing use case on Blockchain (experts can skip this paragraph). Securities trading involves buying and selling of stocks, bonds and other assets using exchanges and broker/dealers. A securities trade is initiated by an investor who is interested in buying a security (an equity, bond or derivative). This is usually done via a broker/dealer who has special licences and permissions to operate as one. The broker/dealers then submit this request to the various exchanges dealing in this security. Similar process happens when an investor needs to sell a security. The exchanges then match a seller and a buyer automatically based on algorithms (matching price, product etc.) and a trade is executed. This is the Trade execution part of securities trading and involves broker/dealers and exchanges and is usually called a Front Office function. Where Blockchain plays a role is in the back office process of the trade settlement and payment. Once a trade is executed, the exchange works closely with the clearing houses that guarantees the transactions for the buyer and seller. These effectively become the central counterparty (CCP) for the transaction and bear the risk of the transaction until it is settled by taking legal ownership of both legs of the transaction and this is called novation. This guarantee is provided using collaterals and margins that the broker/dealers need to maintain with the clearing houses. At the end of the day, the clearing house sends all the net positions of all transactions done throughout the day to the counterparty’s custodian or Nostro agent. The typical settlement cycle is T+2 or T+3 (trading date + 2 or 3 days) and once settlement is done, the depositories change the ownership of the asset from the seller to the buyer.

So, if the broker/dealers were all on a ‘double permissioned’ Blockchain network ( See r3 Corda or Hyperledger Fabric) that records all transactions in an immutable timestamped manner, there should be no need for a clearing house to become a CCP and take on any counterparty risks (as these will be eliminated). This is because all transactions are near real time. Double permissioned Blockchain implementations also address the concerns raised on bi-lateral and multi-lateral transaction privacy in an open transparent Blockchain infrastructure and also goes a long way in addressing the scalability issues raised with the technology’s maturity. There are also many studies out there on what this would mean in terms of savings for the industry overall – Goldman Sachs estimate this could mean $11-12 Bn. savings per year. Ok, so by now I hope many of you are convinced that there is definite potential here for Blockchain to disrupt the clearing and settlement space and save truckloads of money for all involved.

The objections

Now one of the key objections that many of my interviewees raised (apart from uncertainty on regulations and technology readiness) in implementing Blockchain for securities clearing was that the current process and infrastructure of banks and broker/dealers are equipped to the current way of working (obviously!) and this means that their capital planning is done for the two or three day settlement cycle or on a netted basis. So the problem is more process related rather than technology related. If clearing and settlement move to Blockchain, they need to be able to settle immediately to get maximum benefit from the technology (this is where all the costs savings come from – reduced need to reconcile, less capital locked up etc.). While Blockchain actually helps with releasing capital locked up in margins and collaterals, it also means need for immediate liquidity. In the current scenario, this risk is taken over by the clearing houses and managed through collateral management, margin requirements and netting services. Several attempts and studies have been done in the past to weigh the costs vs. benefits for moving to a faster settlement cycle and some were abandoned due to the prohibitive costs of keeping the various ledgers in synch . A BCG study commissioned by DTCC estimated the costs to move to a T+2 clearing at $550 million and $1.8 billion to move to a T+1 settlement. And the payback period based on operational costs savings is about 3 years for T+2 and about 10 years to move to a T+1 settlement. Some have even suggested that it is better to have netted settlement of transactions over a period of time rather than real time settlements. Well of course, these studies were done with the view to reconcile various ledgers across each of the market participants. A similar study using Blockchain infrastructure would be very useful too.

The potential solution

Ok, let us assume clearing and settlement does move onto a Blockchain network where all broker/dealers settle all their trades using consensus based algorithms. All transactions and positions are on the Blockchain and is visible to the ‘machine’ while only permissioned positions are visible to the various participants. The transactions and data stored on the various nodes can be analysed using big data analytics and machine learning techniques. We could either perform siloed machine learning or model chains that perform learning across the entire chain. One potential application then is to use big data analytics and machine learning model chains to help with the issue of the right settlement cycle. Data analytics and machine learning can analyse all outstanding transactions and positions and propose the most optimum settlement sequence to maximize capital utilization for all participants. This proposal then can be propagated via dApps onto the network for the participants to approve via consensus mechanisms. So instead of validating and verifying individual transactions, the participants will provide consensus based approvals for the optimum sequence of settlement. In essence, by combining the power of Blockchain, smart contracts and machine learning, we could create a new clearing and settlement infrastructure that has the benefit of near real time settlements while also offering a sort of netting services on the go – all decentralized and consensus based! And as the machine learns with each sequence of transaction after transaction, it could also provide key insights on the patterns of trading and propose future trading strategy for the individual broker / dealer on their respective nodes. By combining two powerful technology advancements, we can start to tackle some of the concerns raised by the experts I interviewed and could lead to a very efficient capital market infrastructure. 


Comments: (3)

Ramdas Narayanan
Ramdas Narayanan - Bank of America - Charlotte 27 September, 2017, 14:40Be the first to give this comment the thumbs up 0 likes

Hi Priya,

Nice article exploring Machine Learning and Blockchain. I have been hearing about block chain a lot, what is true value add of blockchain, still trying to comprehend that.

Priya Lakshmi
Priya Lakshmi - Wipro - London 27 September, 2017, 17:511 like 1 like


Hi Ramdas - thanks for your comment. Please read this paper by Bank Of England Staff to see the benefits of Blockchain in the settlement space :

Basically Blockchain provides you with an immutable , timestamped , distrubuted record of transactions that are agreed on the basis of consensus and removes the need for central party , therefore saving costs in reconciliation, operations and locked up capital

Hope this helps


Ramdas Narayanan
Ramdas Narayanan - Bank of America - Charlotte 27 September, 2017, 17:58Be the first to give this comment the thumbs up 0 likes

Thank you Priya for the link, will check it out.