Don't Be A Turkey: Optimizing Usage For Optimal Returns

Alpha Theory makes it easier for process-oriented investment managers to do their work, from uploading price targets through Excel Connect, to monitoring Optimal Position Size in the application. Previous analyses have shown a strong positive relationship between increased usage of the app and increased ROIC. Those results showed that for usage, like turkey, if some is good, then more is better. In that spirit, we decided to investigate different forms of usage to see what gives the greatest return on investment.

Two common ways in which we measure client usage are coverage (how much of the portfolio has research?) and freshness (how recently were your price targets updated?). In the same way we help our clients optimize exposure to their positions, we also want to make sure they are optimizing how they spend time engaging with Alpha Theory.

We ask:

1. Which element of usage has the strongest relationship with returns?

2. How much is enough?‍

‍
‍Coverage

There are several ways to think about coverage. The first way we will examine is to think of it as the percentage of names in the book which has research. Simple enough. But this method can provide a false sense of security. Say, for example, you have a five-position portfolio weighted at 80%, 5%, 5%, 5%, 5%. You have price targets for your smaller positions, but not the big ones, giving you coverage of 80% under this definition. But are you really covered?
‍

The second method of measuring coverage is by calculating the ratio of the gross exposure of the portfolio which has research, to the gross exposure of the portfolio as a whole. Both methods have their place, but we were interested to know which was more helpful as a metric to track and optimize.
‍

An Experiment

In order to determine which measure of coverage is more helpful, we regressed both forms of coverage against one day forward returns for a sample of Alpha Theory client funds. The results for the first method are in figure 1, and the results for the second method are in figure 2.

Figure 1 (% of Positions with Price Targets):

Figure 2 (% of Exposure with Price Targets):

The results for the exposure-based method (figure 2) show a higher coefficient, larger t-statistic, and lower p-value than method 1, giving us a high degree of confidence there is a stronger relationship between one day forward returns and the exposure-based method (p-value < 0.05 is statistically significant for our purposes). This measure of exposure is what we will use going forward.

Freshness

‍We should ask the same questions about freshness that we have for coverage. The first way we can think about freshness is to take the ratio of positions which have had their price targets updated in the last 90 days to all positions with price targets. The second is to take the ratio of gross exposure of positions with research updated less than 90 days to gross exposure of the portfolio.
‍

Figure 3 (% of Positions with Price Targets that are Fresh):

Figure 4 (% of Exposure with Price Targets that are Fresh):

The exposure-based method is more predictive here as well, with a higher t-stat and lower p-value. These results make sense – small positions can be placeholders and get less attention. The positions with large exposure have targets subjected to greater diligence.
‍

Comparing the Variables

We can now compare the results of regression against one day forward returns for the two variables under consideration.

Figure 5: Coverage

Figure 6: Freshness

In figures 5 and 6, we see that both coverage and freshness have t-statistics and p-values which are statistically significant, and both variables have positive coefficients. Of the elements of usage examined, higher coverage is most strongly related to higher one day forward returns, followed by freshness.

How much is enough?

Unlike at Thanksgiving dinner, we can estimate how much is enough of each. Focusing on Coverage and Freshness, we examined various thresholds to see how the average client performed above and below it. The 85% and 95% thresholds resulted in the greatest difference in annualized ROIC for clients above and below. More clients have been able to reach the upper echelons of coverage than freshness, which makes sense – over time, and given the reasonably low turnover, it isn’t unfathomable that you can make forecasts on > 95% of the exposure of your book.

Keeping everything fresh, however, requires strict discipline, and an emphasis on process. But it appears as if the hard work pays off: the four funds with freshness over 95% have average annualized ROIC 6.6% higher than the average for all the other funds under consideration.
‍

What is interesting is the persistent relationship for Coverage where higher Coverage always results in better returns but Freshness reverses course at the 70% level. This phenomenon is counterintuitive given the statistical relationship discovered in the regression. It is clear that more research is required and in next month’s blog post we’ll investigate this discovery and explore what can be learned from examining on a by-position basis instead of by-firm.
‍

‍