The technology is available to empower better transaction analytics and increased execution transparency. So why aren’t brokers offering it to their clients? And why aren’t buy-side firms demanding it?
After reading the August 15 TabbFORUM commentary by TABB Group CEO Larry Tabb concerning transparency in the U.S. markets (“Executing in the Dark (Both Literally and Figuratively)” in which he made seven excellent, mostly technology-specific recommendations for brokers to provide better transaction analytics, I felt compelled to offer a few timely insights as to why I believe they haven’t been widely implemented.
For the sake of transparency of a different stripe, know that I’m co-owner of an emerging tech firm that implements infrastructure solutions for trading firms, asset managers, banks and exchanges. Much of what we do is closely related to Larry’s suggestions. Drawing on that vendor experience and my own background in high-frequency trading, I’d like to share reasons we’ve heard as to why firms are not demanding these changes. Even though the following addresses some of the roadblocks to transparency, I believe these hurdles will be less significant once the right set of products and services are available.
All of Tabb’s recommendations – from required collocation and requiring better data feeds, to high-precision timestamps in multiple locations, regular infrastructure audits and detailed trading records – are technically feasible. The additional difficulty and “unknown” unknowns are oftentimes only a part of the reason why brokers are not taking the leap to offer better transaction analytics to their clients.
For the purpose of this response, I’ll break down critiques to three key issues: client demand, cost, and technical challenges to overcome to make this new infrastructure work.
#1 Client Demand
Many asset managers are simply not that worried right now about getting better fills if it’s hard to do. I needn’t remind anyone that the buy side is a large, diverse space populated by some rather tech-savvy buy-side people – but from experience, it’s far from the norm. In order to demand many of these improvements, you’d need fairly sophisticated clients to describe the features and utilize them, which is why I believe the majority of buy-side clients are not making these demands of their brokers.
Many portfolio managers and traders would certainly like better fills, but once the conversation goes into the fine-grained details of where things are timestamped and minute routing details, many lose interest. The amount of time it takes to analyze and understand these issues isn’t seen as being worthwhile, given their many responsibilities. A sentiment I’ve heard is: “I don’t have time to worry about pennies, nickels, dimes or quarters a share; I need to be finding things that move $5, $10 and $25 dollars.”
As many brokerage agreements involve research credits and related offerings, portfolio managers tend to value these above any execution improvements. True, when a firm trades large blocks, a fraction of a dollar per share adds up, but higher quality research is still a much bigger win.
Another question that is somewhat hypothetical at this point is what would happen if brokers did provide this additional timestamp and detailed routing information?
And here’s one more: How do portfolio managers make sense of all this data?
Analysis of algorithms is difficult. I suspect that many buy-side firms know that even if they had all this data, it wouldn’t be easy to get all of the value that’s contained inside of it. For example, let’s assume a broker made a routing decision the moment an order arrived. You might suspect that they made a poor decision, but how could you prove it? You’d need precise datasets and complex software to rebuild the market’s state as it existed at exact moments in the past and then overlay that with what you know about the algorithm.
A second reason many brokers haven’t implemented better transaction analytics is cost. There are different types of brokers: agency-only brokers may have very different budgets on technology than a broker that’s also a principal trader.
For a group that does make the decision to build transaction analytics on top of its existing infrastructure, how much more would collocation, better data feeds, new routing, switch and timestamping infrastructures at each exchange, and high-precision reporting cost? It would be worthwhile to assume that a brokerage can trade on all of the US equity and equity options exchanges. Given the distributed nature of trading, these brokers would probably need to be able to connect to every lit venue for assets they trade.
There are a lot of variables involved in upgrading and building a new system. These are just a starting point for the major additional estimated costs:
- Collocation at Mahwah, Carteret, NY4/NY5: assuming a single cabinet per colocation site at $10k per month per cabinet = $30k per month.
- Exchange Cross-connects: redundant 10Gbe cross-connects at each collocation venue, $20k per venue per month = $60k.
- Data charges: for equities: NYSE, Amex, Arca, NASDAQ, BX, PSX, Bats x, bats z, edga, edge x, CHX = $65k per month; if you want options + $75k (CBOE, ISE, MIAX, BOX, BOM, BX, PHLX) = $140k per month.
- Fiber, or fiber wireless hybrid, networks between venues = $50k per month.
- There are the additional software and hardware costs, harder to estimate because it depends a great deal on a firm’s current level of technical sophistication. Let’s assume you would need to add at least two developers at $250k each.
- Non-recurring hardware costs: redundant routers, switches and 20 servers, hardware-capture equipment for 3 venues x $500k each venue.
The total for the first year is $5.3M with ~$4M recurring each year after that – a rather large chunk of money to invest without overwhelming client demand. Yes, there’s business risk in all of this and like any IT project, brokers may decide to invest but not gain results, benefits and profits expected. Worth noting, brokers may not be able to quantify how much value this additional spend improves executions.
#3 Technical Complexity
This brings me to the third point – technical complexity. Assuming a brokerage group wants to move ahead and give its customers a better technology platform, just how difficult is this to accomplish technically? From what we’ve seen, the major challenge is evolving the existing infrastructure to support a broad set of new functionality without impacting clients. As we are all aware, every software change introduces the risk of failure. Introducing high-precision timestamps and logging of detailed order information into the foundational layers of an existing infrastructure may involve changing data structures in many different places. I doubt the people who originally wrote many of these systems could have ever imagined they’d need nanosecond precision on anything.
Another key question: How many firms have the testing infrastructure and team required to support large-scale data structure changes? The new platform will have to deal with the initial hurdle of ingesting all of this new data without dropping it, often with a much wider set of rules for interacting with feeds than the SIP-based ones they are currently using. Once this higher-quality data is available, any algorithms would have to be changed to make use of it. That could require research, which may not be available. For reporting to clients, a good approach is to use high-precision loss less capture at the WAN links, then split that data out for many uses. Warning: Converting the data to something useful to the business is a laborious task.
The obstacles of cost and technical difficulty can be surmounted, but buy-side clients are simply not demanding transaction analytics enough – and transparency is suffering. If customers did get all of this new data, it wouldn’t be easy to use the results; there are many legacy infrastructures within brokers and those aren’t easy to change; and it’s expensive to build the appropriate infrastructure.
Over time, I expect buy-side attitudes will change. The tools used to service them will improve, and quite possibly regulations could play a positive role by making rules that allow for a better SIP without the large costs of the alternative.
But for now, if you’re on the buy side and want better transaction analytics, as Larry Tabb suggests, it’s time for brokers to come out of the dark, literally and figuratively.
To that I add: Demand your rights and speak up to see more.
Sources: Tabb Forum