Startup helps buy-side firms retain ‘control’ over analytics

“There aren’t really many people who can help you set up your system, provide meaningful tooling that you can run while still—and most importantly—retaining control.” So goes the lament of an executive at a large asset manager. It is why buy-side firms that have the budget prefer to build their own systems—it’s about control, they say.

But even well-heeled institutions can struggle when building analytics systems on their own.

“It’s kind of a very strange halfway house because most people—once you’ve got a product or a system—they dictate the rules,” the executive tells WatersTechnology. “But how do I get the best of both worlds where we have full control over the model and also get somebody who can help build it out? Typically, most people want to build a product and just sell it, meaning you’re back in that previous world where the problem is, ‘How do I put my world into this?’”  

Then the executive met Cat Turley.

Turley has spent over 20 years of her career in finance building greenfield analytics platforms. She started out working with KX Systems, a software company known for its time series database kdb+ and the programming language q, and became an expert on using these tools to analyze data and build analytics platforms.

If you try and get your data to fit into somebody else’s boxes, you’re going to lose the nuance in your data—and that nuance is where all the insight is

Cat Turley, ExeQution Analytics

Turley then took this knowledge and applied it at the other firms she worked for—including JP Morgan, RBS, CIMB Securities, and Haitong Internal Securities Group. She was both a builder and a user of these systems, which gave her an alternative view of all the decisions she made from a design angle when developing a platform.

“After a number of years across a different few brokers and several greenfield builds, I feel like I’d refined an approach to analytics that was efficient to build but also efficient to use, because I had both perspectives,” she says.

This also took place during a time when she felt the market was demanding to do more with data. It led Turley to launch ExeQution Analytics three years ago with the aim of bridging the gap between having technological expertise and understanding trading and quantitative requirements.

At its core, ExeQution Analytics provides consulting services and tech resources while collaborating with both buy- and sell-side firms to design solutions that enable efficient data analysis at scale and speed. It has produced a suite of tools—building blocks, if you will—that are commonly used for quantitative research purposes, trading research, TCA, best execution, and surveillance.

ExeQution Analytics does not have a platform. Instead, it has what Turley refers to as “Lego blocks” that allow users to develop either greenfield analytics platforms or build on top of what they already have to identify patterns in trade data. She says the value lies in understanding the client’s process and its data, and respecting that data rather than trying to shoehorn it into a pre-existing solution.

“I think if you try and get your data to fit into somebody else’s boxes, you’re going to lose the nuance in your data—and that nuance is where all the insight is,” she says.

The executive at the large asset manager agrees.

Nuance in the data

Today, the executive is a client of ExeQution Analytics. They believe that the starting point of optimizing trading is all about understanding: whether it’s your historical trading costs and patterns, what you are trading, or how market data feeds into your trading processes. Just as important, though, is understanding market structure at a granular level and then feeding that knowledge into an analytics engine.

A lot of the time, what was perceived as a problem was just a difference in how trades are handled between an asset manager and a broker.

“An off-the-shelf platform would have lots of data but that might not be enough to understand trading; they only know my world from when I send the order,” says the buy-side executive. “The reality is they might have the data only for the trading, and they have the result of what the brokers did, but then they may not have the detail of how the brokers approach something.”

At their firm, the goal was to build a system that was controlled internally. But the firm needed some expertise—not just in coding, but in subject matter. This is where Turley’s ExeQution Analytics entered the picture.

“There isn’t anybody else that’s come across our door—I don’t know another company that can meet that middle ground right now,” says the executive. “That’s where she’s completely differentiated, because that’s her own personal belief system, which we share. I need real tooling that’s going to help you improve, help you evolve, understand, and do it in a way that’s fast and scalable. Others are either tech or product, but not many have the subject matter expertise.”

Turley points out that buy-side firms can certainly do this themselves—for example, by aggregating TCA reports from their brokers. That process might be painful, though, as there will likely be no consistency from one broker’s report to another.

“It can be as basic as: Do you represent cost as a ‘+’ or ‘-’? If I’m +3 basis points, does that mean I beat the benchmark? Or is it a cost against the benchmark? There’s no consistency on that, and that’s the most basic: plus or minus,” she says.

But again, echoing the buy-side source, those reports will only include data from when the order was sent.

The buy side possesses information about the order that the sell side will never have, such as where it came from, which fund manager it was with, whether it’s part of a bigger order, and over what horizon the alpha will play out.

“Sell sides are never going to have that information—and they shouldn’t,” Turley says. “But if you do your TCA without taking that information into context, you’re limiting yourself. Buy-side firms that rely exclusively on their sell-side brokers to provide their TCA can only see what their sell-side brokers can see. That holistic view is only capable if you’re doing it in-house on the buy side.”

Deeper understanding

ExeQution Analytics analyzes the execution of how the fund managers’ decision was delivered. That falls into two categories: instructions given to the broker, and how well those instructions were carried out.

Turley says TCA has to always serve two purposes. “It’s very easy to game a benchmark and to outperform VWAP or to outperform the close or the open. But understanding the purpose of the analysis, and then being able to choose different benchmarks and different metrics to give context around the analysis, enables you to divide that performance into, ‘How well did my broker execute on the instructions that I gave them, and then, secondly, how good were my instructions?’”

That enables firms to have an understanding of say, Morgan Stanley’s or JP Morgan’s VWAP to make better decisions when they want to trade. But the other side of it is, should they choose a VWAP algo in the first place, or should it be something else?

Previously, firms would look at TCA from a pre- and post-trade perspective, but that is changing as they think about the insights it can provide to the trading desk in real-time to augment decision-making. 

The analysis’s result will allow firms to gain a deeper understanding of the costs associated with their current strategies and have a more collaborative discussion of their brokers’ strengths and weaknesses, thereby improving their entire trading process.

Turley says there is a huge amount of development going into quantitative models to improve algo execution, and this is why customization of analytics is important. It’s about linking up what the model predicts and how the algo reacts because of that prediction, but also how good is that model at predicting? 

An off-the-shelf platform would have lots of data but that might not be enough to understand trading; they only know my world from when I send the order

Executive at a large asset manager

This work is usually done offline by quants before the model is put into production. But it’s the second piece of the analysis—when the algo reacts—that is more important. Is the reaction correct across all different markets and stock characteristics? Or is it overreacting in some circumstances and undoing the value of its prediction because the reaction isn’t tailored to the market or stock?

Turley says being able to put analytics around questions like that brings model development full circle. It connects the quants developing the models with the quants doing the execution analysis on where the algos work and where they don’t.

Some of the building blocks ExeQution Analytics has worked on involve understanding market data. Exchanges, for instance, might publish different descriptions of trade condition codes, which some data vendors might have limited knowledge about.

“How do you identify auctions on Hong Kong versus Taiwan versus Japan, for instance? It’s not easy to find, and if you go to the vendor and ask those questions, it can still be difficult to get those answers,” she says.

ExeQution Analytics documented everything it could understand about trade condition codes, such as how exchanges identify blocks or auctions. “It’s really boring work, but if you don’t do it right, then you’re not going to get interesting results. Now, that’s one of the utilities that we can bring with us to a client,” she says.

If a quant wants to build a volume forecast model and to focus on block trading, for example, they would want only that volume to feed into the analytics model. But if it’s TCA analysis, they might want to look at addressable volume and volume they could have theoretically participated in.

“Being able to define across all of the exchanges globally, that is the purpose of this analysis, and this is the type of trade filtering you need to do. It gives everybody an immediate sprint start in analysis and in getting results, because they don’t have to start with raw data and do all of the data curation themselves,” Turley says.

In November 2023, ExeQution Analytics partnered with KX to develop an advanced trading analytics framework built on the kdb+ time series database. And in July this year, it partnered with specialist global data and technology consultancy Data Intellect, which will allow both firms to provide end-to-end solutions from data capture and quantitative research to trade monitoring and surveillance.