AllianceBernstein digs into its own data, looking for alpha

Firm combs through information about its portfolio managers for signs of bias and bad habits


Andrew Chin, head of quantitative research at AllianceBernstein, views its in-house data as “uncharted territory”. Chin, who is also Alliance’s chief risk officer, is looking to scry through so-called alternative data sets beyond the typical market, company and economic fundamentals. And he has a ready hoard of numbers – the firm’s internal and proprietary data – to distil for alpha.

Over the years, AllianceBernstein has gathered data on its portfolio managers’ decisions, from how they trade – volume, frequency, pricing and other factors – to what stocks they focus on.

Yearly post-mortems evaluate “what’s worked, what didn’t work and what we should do to improve our decision-making”, Chin says. “This is part of our DNA.”

As part of a broader data science initiative, AllianceBernstein has hired a team of six data scientists and engineers, and has already completed a dozen projects spanning the firm’s investment and distribution teams. With the recent additions, the firm has the ability to sift more internal data with more sophisticated techniques, tracking things like how research is conducted, how positions are sized and how trades are executed – all in granular detail.

The analysis could help uncover slants portfolio managers have on what stocks they like or don’t like, behaviour that could translate into a missed investment, or blindness to risk in a market favourite.

“Are there biases in how they do that that they might not have noticed?” Chin says of portfolio managers. “Humans are limited by what we’ve seen and know. We can uncover a lot about how we make decisions. We can learn from past mistakes and successes.”

Grasping that data science would be crucial to the firm’s success, Chin began the data-science effort two years ago.

“I realised that if we’re not there in five years, we’ll be challenged from all sides,” he says.

Humans are limited by what we’ve seen and know. We can uncover a lot about how we make decisions. We can learn from past mistakes and successes

Andrew Chin

AllianceBernstein’s push mirrors similar efforts at other asset managers, JP Morgan Asset Management, Schroders and UBS Asset Management among them, to give their portfolio managers access to quant expertise and tools, and to alternative data.

Like them, AllianceBernstein puts investment decisions through its quantitative lens. Recently, it was able to better predict demand for the 2018 Subaru Ascent, a seven-seat sports utility vehicle, using natural-language processing, a form of machine learning that analyses written or spoken words.

In the past, analysts might have talked to dealers and surveyed customers to assess demand. No more. Using a combination of web-scraping and natural-language processing, AllianceBernstein analysed thousands of reviews of seven-seaters on car review websites, gleaning the features that were most important to buyers. Those features were then mapped onto the Ascent to get an idea of whether people would buy it.

“We know what the sales were, so we can use machine learning across all cars in the past to find out what features drove returns and build own forecast model for the Ascent,” Chin adds.

Chin’s dual role in overseeing quant technology and risk management has positioned him nicely to know what tools or techniques could not only help in investment – but also spot and manage risk, he says.

As part of its risk management programme, AllianceBernstein will use quant tools to better see portfolio and market risks, and understand the connections between market events and portfolio positions. On the operational side, the tools would help oversee the firm’s processes, catch vulnerabilities and provide a dashboard of the key risks.

“I’d like to wake up in the morning and the computer to send me a notification to focus on particular portfolios, because overnight someone put a sanction on Russian companies and we hold them,” Chin says. “That should not be hard to do.”

One obstacle AllianceBernstein faces that established quant shops do not is less firepower; the bigger operations have large teams of data scientists mining all day long. For the firm, this means answering narrower, specific questions.

For instance, recently the firm began analysing foot traffic to measure ecommerce’s part in killing shopping malls. The study found the best- and lowest-rated malls were performing well, but the middle tier was suffering. The insight helped asset managers decide what to invest in and what to sell.

The drawback in looking at foot traffic is the lack of historical data, Chin notes, but even so, the exercise allowed the firm to work with new data sets and uncover their strengths and biases. For example, foot traffic might not reflect purchases, or the data might be regional, and therefore not useful as a national indicator.

From our perspective, we have to get better at using the data because the data will get better

Andrew Chin

“From our perspective, we have to get better at using the data because the data will get better,” Chin adds.

AllianceBernstein has also used quant tools on elections. The firm has built a political dashboard for individual countries with text-mining of news articles and social media, as well as sentiment scoring, a type of analysis that aims to take the emotional temperature of language. Alongside company and market data, it then creates an alternative to conventional national surveys.

“Generally, what we find is that traditional polls are different from online chatter and the final result of the election is somewhere between the two of them,” Chin says.

Another application is predicting inflation risk. Using natural-language processing, the firm tries to analyse sentiment in corporate filings on inflationary pressures on supply costs, as a possible precursor to actual inflation.

“We didn’t know it even existed, but we asked economists to help predict inflation better and they told us there is talk about this at the transcript level,” Chin adds, referring to the write-outs of earnings calls.

The world of data analysis is leaping ahead in a way that is hard to keep a grip on – even by those in the field. In fact, Chin’s motto these days is: “We don’t know what we don’t know.”

“Today, an equity analyst covering oil stocks can track the movement of oil transporters, but they don’t know what the data means and how to use it,” he says. “All this alternative data is out there, and I don’t think quants or fundamental analysts really know what’s out there yet.”

Only users who have a paid subscription or are part of a corporate subscription are able to print or copy content.

To access these options, along with all other subscription benefits, please contact or view our subscription options here:

You are currently unable to copy this content. Please contact to find out more.

You need to sign in to use this feature. If you don’t have a account, please register for a trial.

Sign in
You are currently on corporate access.

To use this feature you will need an individual account. If you have one already please sign in.

Sign in.

Alternatively you can request an individual account here