data scientist
Data science is changing the way retailers do business.

Not every company is ready or able to hire a full-time data scientist. For many, the wisest decision is to outsource this skill; however, adding this degree of separation may create even more confusion, causing some companies to put on the brakes.

Today, we speak with Ed Cockcroft, head of data science at PDI, about the value of data scientists and the challenges they sometimes face when interpreting data for organizations like yours.

Ed Cockcroft, Head of Data Science at PDI

First, a little background: Ed has a first class degree in mathematics, plus a Master of Mathematics with distinction, from Cambridge University. Ed has been working with IIS FuelsPricing, a company PDI acquired in 2018, for 20 years, so he’s no stranger to the market. He has been a senior architect for the PDI Fuel Pricing retail solution since 2011 where he was head of research and development and is now head of data science for PDI.

Data scientists are hot commodities in the job market right now. Would you tell me a little bit about the rising demand for data scientists, and what makes a “good” data scientist?

It’s kind of come out of the big data revolution. You could argue that a little bit of it is fashion. Companies feel they need to do this, otherwise they’re going to lose their competitive advantage. I guess what drove that in the first place was the advent of new technology, techniques, and mathematics. Also, there’s a lot more open source stuff around, so it’s easier to get into it. People need a strong numeric background in science or mathematics, attention to detail, and a lot of logical reasoning.

I’m sure you face new and unique experiences every day in your field, but can you generalize about a day in the life of a data scientist?

We have two streams. First, we have a services stream where we are handling requests from our customers for data science work for things we’ve already developed—things like running data techniques, results and analysis. Then, we have an R&D stream, where we try to improve PDI products by building some prototype models and seeing how they perform. That’s more enhancing the product.

What challenges are fuel wholesalers and retailers facing today that a fuel pricing data scientist could solve?

A lot of them are the classic problems that have been around for years. There’s a perception that margins are thinner and times are harder, and we need to take every opportunity. The types of problems we’re looking at are how do we price in many ways? Whom do I compete against? Whom do I need to watch? Where can I afford to position myself with respect to them? How do I need to time my price moves?

In c-stores, people are more interested in fresh food and coffee because the greater percentages are driven through c-store sales rather than fuel sales. So, people are looking at how to be more intelligent about how they use the fuel to drive the shop and how to use the shop to drive the fuel and how to tie all that together. People want to get more sophisticated with that.

Loyalty programs are the next stage. If you can have a loyalty program where you actually know who your consumers are, you can look at their buying habits. Data like that is becoming more available with improved hardware and software. Customers are beginning to have good datasets on these things.

The next big change will be to start looking at targeted pricing. Supermarkets are already doing this with their loyalty programs, where they are giving people money-off vouchers based on their individual buying habits. That’s a game-changer. We move from picking a single price to stick on a big sign to give to everyone, to determining price sensitivity based on an individual purchase habits. That’s a whole new game, and it will be a step up for the involvement of data science in fuels pricing.

What challenges are fuel pricing experts and data scientists like you experiencing that impact the ability to make sense of available data, and how do you overcome them?

Data quality is often something you have to tackle. Customer expectation is as well. One of the challenges of fuel pricing is that datasets are quite small, so we need to adjust our modelling approach to handle that. We need to make sure that our techniques are applicable given the relatively small datasets we have. And not just data quality, but data visibility. Having really good, timely observations of what competitors are doing. Retail fuels environments are generally fairly tricky. You’re measuring effects that can get diluted quite quickly if you’re not sure. That’s definitely a challenge. On the flip side, you have to manage customer expectations. Given the size of the data available and the noise in the data…there’s obviously some play there.

Stay Tuned for Part 2

Still unclear on the need for a data scientist? Be sure to read Part 2 of our Q&A with Ed Cockcroft, where we look at how data scientists manage the challenges of a volatile fuel market.

Did You Know: Your Source for PDI News provided by PDI the leader in enterprise management software for the convenience retail and petroleum wholesale markets.