Churning earnings with a robot

I was recently asked to consult with a company that converts bulk data into "narratives". For example, the company can take a baseball scorecard and it's robot will convert this into a news story about the baseball game.

A financial services client has engaged this company to use the robot to automate sell side earnings reports. The aim is to create reports that sound as if they were written by an equity research analyst, presumably so the client could market analysis on every company on the stock exchange. The company is looking for someone to help make the earnings reports sound more genuine.

All of this raises the question of the value of earnings reports.

Earnings season consumes an inordinate amount of time for sell side analysts. Companies report, sometimes several on the same day, and analysts  race to be the first to publish their notes, much of which was already presented in the press release, then call their clients telling them their views and perspectives. Earnings days are long and noisy and the reports feed the information needs of institutional and individual investors as well as the business / media complex that both reports on these things and reinforces their importance.

In the week leading up to earnings, the equity research associates prepare models and write templates for the reports. It is a busy time. So an automated note - to an associate - is something of a holy grail that would allow more time to improve a stock screen, manage a personal account, study for the CFA or buy crap online.

Ahead of one earnings season, I created an automated earnings note that did a reasonable job pulling numbers, rates of change and keywords from an excel file to create simple paragraphs like "Revenues were $180M, up 10% y/y and above our expectations. Revenues were driven by [fill in something management said on the conference call]. Gross profit was 10%, 180 bps above our expectations and up 40bps y/y. Profit was driven by [fill in something management said on the conference call]."

The system worked, but raised the question; if I can automate it, what value does it actually have?

We live in a time when capitalism, consumerism and technology have all converged to rapidly convert anything into a commodity. This goes equally for things like electronic goods, healthcare and information. But what retains value over time, will  - as always - be things that are rare, unusual and hard to replicate. In the case of the equity analyst, this includes hard to find information and perspective, on a company, an industry, a management or a strategy.

I haven't decided if I'm going to work with this company or not. On one hand, it would allow dumb banks to convert their research staffs into robots, further shed staff and reduce comp, but still provide "equity research". On the other hand, maybe it would allow analysts to focus on value added research like independent sources, useful and unique surveys, interesting correlations and industry perspective, all in a highly regulated environment that adds value and reduces risk for clients.

Either way, I would remind the robot that just as owning a knife doesn't turn anyone into a chef, so automated earnings reports would not turn it into a sell side analyst.