Banks, Payments and Blockchain: Right Idea, Wrong Tool
SEE LAST PAGE OF THIS REPORT Paul Sagawa / Tejas Raut Dessai
FOR IMPORTANT DISCLOSURES 203.901.1633/.901.1634
psagawa@ / firstname.lastname@example.org
July 20, 2018
Banks, Payments and Blockchain: Right Idea, Wrong Tool
The banking industry has a big opportunity to use technology to streamline $495T in annual interbank transaction flows (which generate $182B in fees) but blockchain (and its distributed ledger architecture) is impractical because of unresolved technical limitations and structural impediments. Specifically, scaling issues make blockchain unwieldy for payments due to processing requirements and transaction delays. The full participation of players in a market is vital to success but will be very difficult to marshal. Compliance demands transparency and oversight that negate some of the benefits of blockchain. Instead, alternative solutions may offer many of the benefits (e.g. cost, speed, security) without some of the downsides (e.g. processing overhead, network complexity, lack of transparency), particularly if participation is limited to regulated banks and/or supervised by a trustworthy party. These solutions use some of the same elements as blockchain – e.g. public key cryptography, verifiable data structures, and fault tolerance – but may be developed and deployed more quickly, obviating the usefulness of blockchain for many applications. We see more commercial opportunity for pure blockchain in supply chains and other applications where a single party can force participation. Still, we see continued investment by banks to prototype blockchains and alternative solutions, making IT consultants the most fruitful way to invest in FinTech for transactions.
- Payments transactions are a big business. In the US, banks handle more than $495T in payment transactions, generating $182B in fees and $51B in related costs. These transactions are cleared and settled with processes that often require multiple labor-intensive steps to authenticate, authorize and execute, adding time (and thus risk exposure and client enmity) and cost. Automated approaches could reduce costs and friction for transactions – blockchain is one possibility.
- Blockchain technology has real limitations. Blockchain was designed to facilitate transactions amongst untrustworthy parties. As such, security and recordkeeping are crowdsourced – distributed to all members of the network. This adds substantial overhead – mechanisms to establish membership, to validate transactions, maintain records and to adjudicate disputes are computationally complex and time consuming, with transaction latency and cost increasing geometrically with the scale of the network. In systems that need to process millions of transactions (e.g. consumer payments) with expectations of authorization in seconds rather than hours, blockchain is unlikely to be practical for many years. Moreover, every member of the network must keep a full record of all transactions, demanding many times the storage capacity (and cost) of a centrally maintained database.
- Blockchains require full participation. The overhead costs necessary to settle and clear transactions in traditional banking processes cannot be significantly reduced unless most activity has transitioned to the new solution. Blockchains – which distribute the processing and storage burdens to all participants, require the sharing of transaction data, need agreement on governance/procedure, and may call for IP licensing agreements, etc. – will have an uphill battle. With many competing development projects underway, rallying support to a single network will be even more difficult. Moreover, with their role, and thus financial opportunity, diminished by blockchain, intermediaries in existing transaction processes will work to discourage adoption.
- Regulation demands transparency. Blockchain was built to support anonymity. Bank transactions are subject to scrutiny for AML and other compliance issues. This oversight requires a master ledger and visibility into the details of each transaction – including the identities of all parties and the amount of each transaction. This negates some of the primary intended benefits of blockchain without eliminating the computational overhead inherent in its structure. It may also slow regulatory approval for processes automated by blockchain.
- There are alternatives to blockchain. While blockchain researchers continue to work to counter the scaling and processing overhead issues with the technology, other technical approaches may be better suited to the needs of the banking industry. Given high levels of trust between highly regulated banks and the existence of 3rd parties already tasked with managing transaction networks – a distributed ledger, with all participants recording and validating every transaction is overkill. Public key encryption and verifiable data structures (hash trees) combined with regular audits could provide strong protection against fraud and clear resolution for disputes. Redundant backup databases could protect against malfeasance by a central party without requiring each market participant to have an up-to-date ledger.
- Centralized leadership key for near term progress. Transaction networks that can be coordinated by a trusted central party will have an easier route to adoption. A single dominant enterprise might force participation, such as Walmart deploying a blockchain-based solution for its supply chain or Maersk using blockchain to track shipping containers. An existing 3rdparty – like a securities exchange, a central bank, or a banking consortium – might marshal cooperation from an existing membership. Even so, it is likely that the most successful solutions will not incorporate a fully distributed ledger, arguably the defining characteristic of blockchain technology.
- Blockchain or not, banks and partners will invest in next-gen transaction processes.We are skeptical of financial blockchain startups and consortia given the drawbacks of technology, the barriers to adoption and the existence of viable alternative technical approaches. For transactions amongst banks and their well vetted commercial customers, automated processes that rely on immutable receipts, regular audits, and redundant databases without the overhead of a fully distributed ledger and crowdsourced settlement, seem far more practical. We believe that these will be bespoke solutions fit to each specific market. While the largest banks have invested in these technologies and have much to contribute, we believe more neutral parties will be needed to drive broader participation. From a tech industry perspective, we favor the top IT consultants to the banking industry – IBM and ACN. Longer term, we expect next-gen payments networks to migrate to public cloud hosting, a future boon for IBM, MSFT, GOOGL and AMZN.
A Solution Looking for a Problem
Blockchain technology was first described by the pseudonymous Satoshi Nakamoto in a 2008 paper and first employed in the launch of Bitcoin a year later. The concept combined several known techniques – i.e. public key encryption, Merkle (Hash) trees, and distributed ledgers – to solve the “double spending problem” that had plagued all previous attempts at digital currency. With blockchain, digital transactions could be conducted amongst anonymous and untrustworthy parties with confidence, as transactions were immutable and scrutinized by every member of the market. It had the added benefit of eliminating the need for costly and time consuming human oversight, as the system was essentially “cheat proof”.
US banks handle more than $495T in payment transactions each year, generating $182B in fees and $51B in related costs. Blockchain, with its potential to automate much of that, has been viewed by many as a gamechanger. Startups, like Ripple, and banking consortia, like R3, Hyperledger and UTC, have invested heavily in “proof-of-concept” networks to demonstrate the usefulness of the technology in a banking context. Every big bank, exchange and payments facilitator has been funding their own development efforts, often with the support of outside consultants, to figure out the best way to use blockchain in their businesses. Still, more than 10 years after Nakamoto’s paper and an explosion of cryptocurrencies, there are no live full-scale blockchains in the banking world.
There are important reasons for this. First, banks and their customers are not anonymous, and, with strict government scrutiny, are considerably more trustworthy than the Bitcoin community. Because of this, a fully distributed ledger – the defining element of blockchain – is overkill in a banking context, scaling poorly and adding burdensome computational and storage overhead that slows down transactions and adds infrastructure costs for every member of the network. This highlights another obstacle: blockchains require full participation, requiring each member to bear their full share of infrastructure costs and to buy in to the design and operating principles of the network. In a fiercely competitive world of rival projects, this will be very difficult to achieve. Finally, blockchain limits transparency to the value of an asset and its chain of custody across anonymous accounts. Governments – tracking money laundering and other financial crimes – will demand far more scrutiny, either opening the details of all transactions to all participants or requiring a single central database.
Banks are likely to look at alternative approaches. Public key encryption and Merkle trees are not unique to blockchain and can provide similar assurances for the integrity of transaction records without the overhead associated with distributed ledgers. A central database can be replicated and regularly audited by a small subset of network members – the risk of sanctions for non-compliance and the likelihood of discovery would be sufficient to make consensus review of each transaction unnecessary.
We are bullish on bespoke projects to automate banking transaction processes – we just don’t expect the solutions to be blockchains, per se. This means that the IT consulting gravy train should continue – IBM and ACN are the obvious beneficiaries. We are not enthusiastic for the banking blockchain startups – Ripple is unlikely to displace existing entities, like SWIFT or The Clearinghouse, and their internally developed transaction automation solutions. We believe distributed ledgers will be more applicable to trading exchanges, supply chains, complex asset sales, and other non-banking commercial applications.
The Trouble with Payments
After providing credit, facilitating payments is likely the banking industry’s next biggest and most lucrative business (Exhibit 1). US banks earn some $182B in fees for handling more than $495T in payments, generating $51B in related costs. Customers – consumers, enterprises, other banks – get frustrated when payments are delayed, tying up funds and holding up transactions, and would rather that fees were lower. Errors happen, and disputes can be difficult to adjudicate, creating more delays and costs. Competitive advantage in this arena, via lower cost, more efficient, accurate, and faster processes, would be obviously valuable.
Exh 1: Roll up of Payment Transaction Fees Earned by Commercial Banks
In this context, blockchain is intriguing. For the few uninitiated, blockchain is the technology underlying Bitcoin, detailed in the seminal 2008 paper published by the pseudonymous Satoshi Nakamoto. “Nakamoto” laid out a schema to eliminate the risk of fraud in purely digital transactions. With blockchain, each participant in a market keeps a full ledger of all the assets in the system with the complete history of prior transactions. The assets are represented as irreplaceable digital tokens that can be presumed as the property of the holder, as the system maintains a commonly held record of the chain of custody of that token. When a new transaction is attempted, the community collectively approves it by assuring that the asset holders have legitimate claim to their holdings, that the transaction is allowed by the terms of the network, and that the transfer of assets is logged in the distributed ledgers (Exhibit 2). In this system, the parties to a transaction can be anonymous and untrustworthy, but the validity of the transfers assured.
Thus, transactions can be automated without fear of fraud, eliminating intermediaries – e.g. title insurers, escrow holders, inspectors, brokers, etc. – and with them, cost and delay. Banking transactions, which currently have rosters of humans to check, double-check and triple-check anything even a bit out of the ordinary, could benefit from some of this. It is not surprising that banks have broken out their checkbooks,
Exh 2: Blockchain Transaction Confirmation Workflow
funding proof-of-concept networks to explore the usefulness of blockchains for their business and the issues that will need to be resolved along the way.
The R3 Consortium was launched in 2015 by 9 banks led by Goldman Sachs and JP Morgan (both of whom later bowed out to pursue their own independent development), with a 40-bank trial of a debt-trading blockchain conducted in February 2016 (Exhibit 3). It has believed to have been funded with more than $100M in investment by members. Later in 2016, the Hyperledger Consortium was launched, with a roster of financial entities and tech firms. Hyperledger has built an opensource blockchain platform based on technology contributed by IBM. Meanwhile, the biggest financial entities – Goldman, JP Morgan, Morgan Stanley, Visa, Swift, and others – have substantial internal programs to work on blockchain solutions.
Exh 3: Blockchain Ecosystem Development Timeline
The Bigger the Chain, the Heavier the Block
As a fully distributed ledger, blockchain requires each member to review each transaction and to maintain a full database of the history of all assets on the system (Exhibit 4). This generates a lot of overhead, as large transaction networks can generate many thousands of transactions per second to be reviewed by many thousands of members – this takes time. The Bitcoin network, which has roughly 13K nodes, has an average approval time of more than 20 minutes, with the delays increasing as the community grows and spikes where delays can stretch to hours (Exhibit 5). While this performance would be an improvement for the least automated bank transactions – some international transfers can take days to clear – for most it would be an untenable delay.
Blockchain also requires ample computing and storage resources duplicated for every node on the system to maintain that fully distributed ledger. In comparison to a centrally administered system, where transactions are executed once and stored in a single location, distributed ledgers repeat the process many times over, multiplying the infrastructure requirements along the way.
While these costs and delays are a necessary evil for facilitating transactions amongst anonymous parties with an omnipresent risk of malfeasance, they would be a substantial burden for banks. Banking transactions are scrutinized by regulators and compliance failures are prosecuted and punished, making them largely trustworthy agents. Any implementation of blockchain in this context would need to address the unnecessary overhead. Researchers are at work to define alternative versions of blockchain that loosen the requirement for a fully distributed ledger and unanimous approvals, but these are not yet available to commercial applications.
Exh 4: Bitcoin Network Average Block Size Historically Since Inception
Exh 5: Bitcoin Network Average Block Confirmation Times in Minutes
If a blockchain is to replace an existing process for banking transactions, all the important parties to those transactions will have to agree to participate. This will be difficult to accomplish. First, blockchain requires that the details of transactions be shared with all nodes on the network. Participants will have to agree on the specific details to be shared and be comfortable with rivals having visibility into their activities. Traditional systems may have restricted this access to a trusted central party, and it is easy to imagine resistance to a more open protocol.
Second, the various players will have to agree on the design of the network itself – does it use proprietary IPR? How does it handle clearance and settlement? How will it be audited? And so on. There are a lot of “proof-of-concept” implementations out there, with different banks championing different solutions. Getting everyone on the same page will be difficult. Third, blockchain is going to require all participants to invest in computing and storage to deal with their responsibilities in approving transactions and keeping their copy of the ledger. In a large, busy network this could be quite costly, and some participants would likely balk. Finally, there are influential players whose role in intermediating transactions would be diminished or even eliminated by blockchain. These players could challenge a developing blockchain through legal or business channels.
The evolving membership of the big banking blockchain consortia is a preview of this jockeying for position (Exhibit 6). R3 lost founding members Goldman Sachs and JP Morgan, along with Morgan Stanley and Santander, as all decided to focus on their own internal work. Hyperledger has the support of many leading tech companies, and has had success in launching numerous prototype networks, but its banking membership is still somewhat thin (Exhibit 7). Ripple claims to have dozens of banks supporting it, but comments from the banks sound much less enthusiastic. Until a critical mass coalesces around a common solution to a specific transaction market, it is difficult to be bullish.
Exh 6: Key Elements Hindering Meaningful Adoption of Blockchain in Banking