For the final part of their discussion on the future of energy retail, Gorilla CEO Ruben Van den Bossche and VP of Product Joris Van Genechten provide the link between parts 1 and 2. In part 1, they discussed the challenges of energy retail and their experiences working with retailers. Part 2 zoomed out to look at the electricity grid and the ongoing changes it faces. What is the way forward for energy retailers? How can they move forward with net-zero and maintain competitiveness in a tumultuous market that is constantly changing?

Getting to grips with truth in data

Ruben: I think it's part of any enterprise's data strategy to come up with a single source of truth in an organisation rather than having seven forecasts in seven departments each with their version of the truth. What in your opinion is the true underlying problem? Why is it so hard for energy retailers to get there?

Joris: I think there are probably multiple problems. You can’t underestimate how complex each of these parts are. A lot of these companies come from very traditional, maybe even government-owned organisations, and they've really focused on stability. That's been the focus for a long time: reliability, stability, just making sure everyone has energy.

This has transitioned in the last couple of years to them becoming the innovators, companies that like to make a difference, but that means a big organisational change. They have to break open a lot of the silos that have been formed in the past and put in place a true data strategy with strong data capabilities that can tie all of it together.

I think for a very small retailer, you might be looking at the same portfolio across all of your departments because it might still all be in one system. But you look at the large ones who have a C&I business, have an SME business, who have a B2C business, and who might've purchased some smaller energy retailer during the crisis to help them out. They're looking at a lot of legacy systems all over the place. So it's going from that sort of legacy patchwork of systems to a dedicated data strategy that can tie it all together. This is the underlying challenge for those companies to overcome.

Ruben: It feels like a lot of energy companies now understand how crucial it is to do forecasting and pricing better.

I think they got a few wake up calls in the last two to three years. Some of them really believe they need to own it themselves. But that does not mean they do it in a data driven way. An energy company is not a software company. They seem to be struggling with everything from data architecture to pipelines to segmentation. How are we helping them overcome those challenges?

Joris: I think what we've heard and what we've seen in data is exactly as you mentioned, some energy retailers are heavily investing in that themselves, really trying to be that software company to do that. Sometimes it’s leading to ridiculous costs to the company. I think that's always something to be very cautious of. You want to invest, you can invest, but knowing if you build a software company like we are, that also costs a lot of money and you’re doing it only for yourself. It means you need to maintain it, you need to host it, and you need to continually support it, which is hard to do as an energy retailer. It takes you away from your core.

What we try to do is give you the platforms to be able to do that, give you a set of base models that you can already start with so you can - out of the box - have automation to run it through the data management, the data storage, and the infrastructure. But, it still gives you a lot of configurability. 

We are energy experts, but our customers are the ones that are going to make the difference. Being able to use their data themselves and configure what the optimal solution is for them is still something that we put in their hands.

We are not going to tell them how to run their business. We're just giving them a suite of tools that are already energy specific, but that they can bring to the next level to run their own business. 

We've laid those foundations. We've even added on an energy specific layer for our customers, but we still expect them, the experts, to configure it and make it their own. We're trying to make it as fast and as automated as possible, but they need to put in their extra layer of expertise.

A model decision?

Ruben: One mistake that I see energy companies make is with their data scientists, the people that truly understand energy and how to model the behaviour they want, whether it's forecasting or pricing.

What they forget is to get that into a production environment and run it on a portfolio of millions of meters, you don't just need data scientists, you need a data platform. 

There are enough data platforms available out there, but they're horizontal. 

So on top of that data platform, there's still a huge amount of things to be built before you can bring that one trained machine learning model into production and run it with integrations into your CRM, your billing, and your trading system. 

They keep changing their minds when they find a new model because the pace at which those models are evolving right now is ridiculous. 

Once you have the new model trained, which might be in two months already, you can't go through a change process of six months to just get it live and test it again in production.

And that's what we're good at, right? We're good at that layer underneath and giving them the flexibility of going in and deploying those new models, whether it's pricing or forecasting, and making sure that they're integrated once and they continue to work.

Joris: I don't remember who said it, but you are 100 % wrong on all the things you don't forecast. 

And that is what we see sometimes, they're investing so much in getting new models and trying to focus hard on that accuracy that it basically gives them a gap where they're not forecasting or where they might be missing parts of the portfolio in their forecasts. 

If you don't have an operational production system in place you're 100 % wrong because you've not done it properly. 

In some cases, it's actually smarter to go a bit quicker and start a bit simpler, but just get it live. Make sure you have all the data, make sure it all runs and is stable, and you don't have to look at it and then start innovating because spending so much on innovation might never get you there. 

And we've actually seen that at one of our prospects that invested so much and eventually cancelled the project, which is very sad of course, but it happened.