But, as far as we’ve come even these “next-generation” algos are still mired in solving problems only at the point and time of execution. What’s really needed are holistic process algorithms. These algos represent hybrid technologies that combine several stages of pre-trade, in-flight and post-trade logic to optimize critical tasks and decision making throughout the investment-trading process.
Holistic algos would combine execution logic with transaction cost analysis (TCA). They would aggregate and analyze complex sets of microstructure data and produce real-time forecasts to maximize the likelihood of best execution. They would integrate trade-order management with performance accounting and would automate regulatory and compliance procedures.
Buyside demand
Their time has come and buy-side demand for them is growing. When we consider the existing paradigm of algorithmic technology, we quickly recognize that much of the functionality resides on the other side of the fiduciary fence with the sell-side.
That is by design of course – sell-side institutions seek to maximize profits by capturing share of wallet and lifting volumes. The algos they design are intended to captivate clients and add proprietary and competitive value.
But when one deploys algos nested solely within the agency/principal environment, the ability to optimize critical pre and post trade decision making that must occur in-house becomes limited. These decisions drive performance just as much as, if not more than, slicing notional and pricing liquidity.
The fact is that the majority of systematic analysis and critical decision making that goes into the eventual deployment of an algo actually occurs upstream in the investment process and well in advance of selecting an algo strategy.
Day in the life
Consider the typical “day in the life” of a buy-side execution desk. At a large mutual fund, multiple portfolio managers running multiple mandates across different international asset classes submit their required re-balancing, subscription/redemption and hedge transactions on a daily basis.
From these transactions, the execution desk must (often by hand or ad hoc routine) aggregate and net the buy and sell transactions across multiple currency pairs, segregating spot transactions from forwards and swaps. Once overall notional amounts are determined, the execution desk must then consider when and with whom all or parts of these transactions should be conducted. Liquidity must be considered across time zones and cross currency pairs.
An assessment of market conditions must be conducted along with the timing of transactions. The implications of market impact, information leakage and other costs of implementation, like agency fees, must be considered and then the appropriate trading strategies must be selected.
At this point an avalanche of questions must be answered:
- Do I request quotes from a bank for risk-transfer pricing, aggressively sweep liquidity from a platform, or run an algo?
- If algo, which type?
- What slicing size/slicing interval should one choose?
- Should I be passive or aggressive?
- Why?
- What are the risks?
- What are the benefits?
- What are the expected costs?
- Of all of these choices, which approach will get me done as quickly as possible, with greatest probability of completion, while still providing best execution?
- And by the way, is there any way to systematically or opportunistically generate some alpha out of all this?
These procedures are all closely related to the algorithmic logic that must be eventually implemented. But the answers needed, the data and analyses required to arrive at the answers and the interface by which all of the information may be organized and manipulated must all be available before even a single slice is executed.
Is the answer to simply place more user controls and inputs on a sell-side algo GUI? Perhaps, but to accommodate all of the possible bespoke scenarios that could arise such an interface will begin to rival the dashboard of the space shuttle.
Common algo design protocols
What’s really needed are solutions that would better integrate with buy-side technologies, both vendor and proprietary solutions. Even better, the industry needs to be moving towards common algorithmic design protocols that would allow seamless integration of disparate systems to achieve a desired goal. The FiX protocol attempts to do this with messaging, but we need to move past relying only on this antiquated technology.
The holistic algo is not a single stand-alone solution. It’s more of a component driven, distributed system whose parts can communicate each other seamlessly and which can easily scale up or down as modifications are needed.
In the same way that self-driving cars are revolutionizing the safety and efficiency of getting from point A to point B, holistic algos will revolutionize investment process trajectory of handling transactions. The goal should not be to simply slice and dice, the goal should be add value, achieve tailored best execution and drive performance at all levels of the investment-trading decision making process.
