How to Make Data-Driven Decisions: 3 Pointers

Originally Published September 22, 2017

In the recent blog posts on data analytics, we've covered gathering data, cleaning data, structuring dataand finally analyzing data.

In those posts, we covered how building a robust litigation finance business is not a "completable" task, but an ongoing process with no definitive stopping point.

The same thing can be said of building a data-driven investment process. Too many executives and investment professionals alike think of big data and data analytics as a box that can be checked, rather than as a part of ongoing business operations.

The reality is that the most successful businesses using data are those that make analytical rigor a feature of their day-to-day operations.

Now, we'll talk about how you can use data to improve business decision-making.  

#1 Use the past to look to the future 

When it comes to interpreting data, just looking at the numbers is not enough. 

Instead, investment professionals of all stripes need to determine whether the results they are seeing, which reflect historical data, also indicate future numbers and statistics. This is especially difficult in legal funding.

One of the major critiques of date use litigation finance is the relative scarcity of data, owing to the industry's short history.  This is valid, of course; but, it also pressures investment managers to give extra consideration to whether that short time series of data accurately reflects future progression and trends.

Now, the reality is that, although data is not perfectly representative of the future (and it never is), it's the best alternative we have, and much better than subjective opinions. 

#2 Using Standard-Deviation to Make Better Predictions

In order to assess data effectively and decide how to act based on results from data analysis, we need to look at what's called the standard deviation or standard error around any data points we examine. The standard deviation tells us how much variation exists in a set of data over time.

For example, if we are looking at case duration, we might find that the average case resolution for a given case-type is 16 months, making it seem pretty attractive. But if the standard deviation is just under eight months, that tells us the 16-month figure is something of an outlier- many cases finalize in a very short time period, while others randomly drag on for three years or more. In short, that 16-month number is highly unreliable.

In fact, in a normal data distribution, we can estimate that 90% of our data points will fall within two standard deviations of the mean. So, in the example above, 90% of our cases might reach a conclusion in somewhere between one month and 31 months.

The point here is that the narrower the standard deviation, the more useful any predictions we make will be. If our standard deviation on case duration were only one month instead of roughly eight months, we could be 90% confident that cases would typically be completed in a span of 14 to 18 months-a considerably tighter window for predictions.

Lucky for you, Excel makes this easy with its =STDEV() function. 

#3 Assess Statistical and Economic Significance

This type of prediction model is also great for assessing what type of cases are likely to be more or less profitable for investment. 

For example, we might estimate that premises liability cases area have an IRR of 29% based on past data collected, while mass tort cases have an IRR of 24%.

If the standard deviation on intellectual property is 11% versus 5% on mass tort, that gives us a 90% range of IRRs of 2% to 46% for intellectual property, versus 14% to 34% for mass tort.

These numbers are hypothetical, of course, but you get the point. Data can help us to decide if investing in particular areas makes sense. Do we want a safer, lower return from mass tort, or a more uncertain, but potentially higher return from premises liability? The answer depends on the objectives of each firm.

In addition to analyzing the statistical significance of the range of expected returns, we might analyze whether those return differences are economically significant. If mass tort cases are profitable, but there are not enough of them to use all of our available capital, then the difference in case profitability is a moot point.

Overall, when considering data in the context of litigation finance, it is not only important to consider numbers, but also what these numbers mean. Studying variation in data over time helps firms assess whether or not the future will echo the past.  Again, the key in all data-driven endeavors is to treat the analytics as a process rather than a one-off effort. Taking this approach will eventually pay big dividends from an investment standpoint.