With the influx of data — such as eval prices, dealer indicative levels and block transaction levels — pricing plays an important role for the buyside, as the 'build versus buy' debate continues to play out as technology accelerates.
Some shops have opted to build their own pricing systems, which provides customization and competitive advantages, but carry hidden costs. Others have decided to buy programs, which offer quicker implementation, lower upfront costs and best built-in practices, with the downside being possible heavy reliance on the vendor.
Shops like AllianceBernstein and BlackRock have built their own systems. For the former, it built its own pricing system internally, believing it can unify all needed information, said Susan Joyce, head of municipals trading and fixed income market structure at AllianceBernstein, at a panel at the Fixed Income Leader Summit on Tuesday.
The buyside has the best information in the muni market, seeing trace prints, in addition to dealer runs and customer bids, Joyce said.
"We see all of this stuff … so being able to put all that together is crucial, and I think, on the buy side, you really should be looking at building something internal using all of this great data that's out there," she noted.
Data has never been better, and there have never been more tools to help market participants analyze it, thanks to the rise of artificial intelligence, Joyce said.
"It all starts with price. If you don't price, you don't have a way to do your no-touch trading, your ability to scale our trading. It all comes down to having that really solid price," she said.
BlackRock has also built its own in-house, real-time pricing mechanism, Alessio Muscara, director and head of SMA implementation at the firm, said during the same panel.
With the several sources of data the firm receives, the question becomes how to create a model with multiple variables that translate what a trader would do on a daily basis, he said.
And based on the information market participants receive, they can decide, for example, whether that bond is being offered at a reasonable price, Muscara said.
However, "if you're trying to systematize and build a skill around it, that has to be something that happens automatically," he said.
Therefore, "coming up with your own model that forecasts, or at least gives an indication where it does, should be super important," Muscara noted.
However, just because shops decide to build their own systems, it doesn't mean third-party pricing vendors are lacking, Joyce said.
Some firms, for instance, prefer vendors even if they have the means to build.
Market data platform SOLVE uses AI to build its own predictive pricing data set. The suite of products has been so helpful, one large buyside shop said, it believes 70% of its trades are more profitable after using SOLVE's AI-generated price than its own internal pricing, Tim Stevens, chief product officer at SOLVE, said at a separate panel at the conference.
Using AI, SOLVE can generate live signals from the data and predict where a bond will trade, he said.
The model is also designed to minimize the prediction error between where the firm believes a trade will occur and where it actually occurs, Stevens said.
The shop's clients, in turn, use that data either through the front-end that SOLVE provides, such as Perform, or via feeds or application programming interfaces, he said.
The data is then used in various ways, such as providing clients with a signal or level to help them determine the best trading price for a particular bond, serving as input to their own internal pricing models, and providing guardrails for those models, some of which are AI-driven, Stevens said.
