Thought Leaders in Financial Technology: Bristol Gate Capital Partners CEO Richard Hamm (Part 2)

Sramana Mitra: When a bank does hire you to do this kind of work, double-click down to exactly how the technology works and what it is that it’s doing on behalf of the bank.

Richard Hamm: What the banks want us to do is to satisfy their investment-oriented customers by giving them a good return.

Sramana Mitra: I’m trying to understand what’s happening in the guts.

Richard Hamm: They don’t come in and have a very deep look at what we do in applying predictive analytics. We tell them what we do and we take them through the evolution of the system.

Sramana Mitra: What I’m trying to extract is what is the thought leadership that’s going on and how that applies.

Richard Hamm: If a bank was to look at what we did on the face of it and say, “What we do is what they can’t do.” We simply take them through the process of applying data and the records that we generate from the data into predictive models for dividend growth.

Sramana Mitra: I’m trying to go a bit further down in the process. We understand that at a 30,000-foot level. Give examples of what kind of data are you using to predict what.

Richard Hamm: We use data from Bloomberg and Fred. We prepare all these records. We run models based on about a thousand features of the data going back 20 years.

We train the model through machine learning to take what we have done in the past that has worked to find the same basic premise of prediction that we would have done by linear regression in the old days. Then that data is sorted into a training model.

The training model is then applied to the set of securities that we drive down to the number of features that we believe are key predictive features out of the thousand features we start with.

Generally speaking, we work it down to about 200 features. Then in that 200 features, there are 20 dominant features. As you can imagine, one of the most basic features is history. The history of what has happened is important in order to predict the future. Where we’re different is people don’t predict the future very well.

Our business is based on doing that. Some of the features are anything that you can imagine financially. That would go into our gradient boosting machine that would take what we have known in the past to be reasonably good and better than what we could do on our own.

We just keep that model running constantly. We always have a model running on top of an actual. If we find something that’s significant or that can change it, we’ll be that much better off.

Sramana Mitra: The end goal of all this is to maximize the dividend-based income for investors and public stocks.

Richard Hamm: Yes, these are SMP 500 companies only. That’s the most difficult market in the world to apply this to because it’s the most over-invested market in the world. We set out with the goal of doing it so that we can say, “Can we predict dividend growth better than analyst estimates?”

If we can lower our error rate in predicting dividends below that of analyst estimates, then we’re doing the right thing. We measure that kind of accuracy improvement. We are about 1.5 times better than analysts.

This segment is part 2 in the series : Thought Leaders in Financial Technology: Bristol Gate Capital Partners CEO Richard Hamm
1 2 3 4 5