When your analysis doesn’t work

Let’s be frank. During the past three or four months if you’ve caught any of the presidential election debates, you’ve probably yelled at your television at least once. This year’s crop of candidates, perhaps more than usual, have presented a broad spectrum of strategies to tackle national issues, occasionally drawing cries of frustration, regardless of where you stand politically.

All of the candidates posture about what they would fix if elected. How their plan would benefit America and its citizens, and how their well-informed and experienced stances about issues of national importance are the right ones, delivered in well-practiced speeches outlining action points and priorities.

I don’t know about you, but when I watch these I sometimes wish they would just be inconclusive about something. Just say, “I need more information”, or “You know what, what we’re doing right now is pretty good, let’s keep with that.”

To me, the ability to be certain about what isn’t clear when asked to provide advice is an important skill and something that as a former investment banker I’ve learned to appreciate.

Case in point, recently a client wanted my advice on the wording of a share repurchase announcement. Eventually, the responsibility of crafting the exact copy would be passed to the communications department, but he wanted my thoughts, and as a trusted advisor, I felt it my responsibility to at least investigate.

Now in addition to the opportunity to yell at the TV for something other than sports, the election offers new ideas for analysis and visualizations. It’s a time rich in data, from poll numbers, to state races, to the history of each candidate, there is a lot of information to sift through and a good opportunity to steal ideas for pitchbooks.

One idea I wanted to try out was a word cloud as a form of text analysis. A word cloud, in case you’re not familiar, is a visualization consisting of words from a particular text or subject, where the size of each word indicates its frequency or importance, such as this one from Wolfram Alpha visualizing the remarks made by Bernie Sanders in the first two Democratic debates.


Image credit: Wolfram Alpha http://blog.wolfram.com/data/uploads/2015/10/Sanders_large.png

In politics, word clouds bubble up insights that may otherwise go unnoticed and can be just as revealing as polling data as they show speech patterns, phrases, and which ideas are receiving more air time, such as Sanders’ frequent use of “country” and “people”.

So, I decided this client request would be a good excuse to take a look at how word clouds could be effective in banking. The ten best-received share repurchase authorizations of 20151: were pulled together, the announcement copy sorted, parsed, and organized into a word cloud.


Nothing immediately jumped out. “Stock”“Company”“million”“program” are all words I would expect in a share repurchase announcement. So I performed the same exercise but for the most poorly received share repurchase announcements of 2015 and hoped it would reveal something insightful.


It did.

It revealed that pretty much the same words are used in positively and negatively received share repurchase announcements, essentially, the analysis was completely inconclusive.

You may consider this to be a failure and that, as a trusted advisor, I should scrape together something to sound smart and informed.

But, sometimes the smartest thing you can say is, “The data doesn’t really say anything. We need to look elsewhere for insight.”

The most interesting thing about this exercise? Nothing to do with the data, but with my relationship with the client. By going back to him with something inconclusive I not only demonstrated that the effort was made to answer his question but more importantly it reaffirmed his trust in my advice, as sometimes the best advice is no advice at all.

This isn’t to say word clouds will always be inconclusive. Fed announcements, quarterly calls, shareholder meeting minutes, even tweets about a company, are all rich sources of data that by performing a textual analysis could illuminate something interesting.

In this case, however, it turned out the best course of action was to take an undecided position and turn to other sources for information, something I hope whoever becomes elected President will be able to do, too.

Have you ever used a word cloud for a client? Or been faced with the challenge of offering no advice? Email me at adrian.s.crockett@gmail.com.

  1. As measured one-day beta-adjusted excess returns to the S&P 500.