Changing the rules of grocery retail with data and artificial intelligence at Kolonial.no

Kjetil Åmdal-Sævik
Oda Product & Tech
Published in
4 min readMar 28, 2019

Kolonial.no is a Norwegian technology and logistics company that happens to sell groceries online. Our ambition is to disrupt the entire fast moving consumer goods industry.

Since our humble beginnings we have been passionate about building phenomenal customer experiences and rethinking the use of technology and data to create the grocery store of tomorrow. For this we obviously need artificial intelligence.

Because our customers shop online, we have access to a wider and more diverse range of data sources than traditional grocery stores. This puts us in a unique position to build innovative solutions and challenge established truths in the industry.

With this data we can approach the challenges of building personalised shopping experiences for our customers and designing the value chain for fast moving consumer goods in completely novel ways.

Analytics and artificial intelligence is a natural part of this, and our value chain is tailor-made for applying machine learning and optimisation techniques in every nook and cranny.

To achieve our ambitions, artificial intelligence is not just important; it is absolutely essential. Let us tell you about the four main areas where we are currently building AI.

1. Reducing uncertainty about the future

Forecasting is one of the most important things we do in our company.

Reducing uncertainty about the future is critical for planning campaigns, procurement, warehouse staffing and delivery routes. Our value chain is a complex system where all individual parts need to work together like a well-oiled machinery to ensure that our customers get their goods where they want them, when they want them.

We aim for world class performance in this area, and therefore experiment with everything from Bayesian inference with Markov chain Monte Carlo sampling to recurrent and convolutional neural networks, depending on what types of problems we are trying to solve.

Forecasting the number of orders the Bayesian way!

2. Knowing which products our customers want to buy — and helping them decide when they don’t

Another important area for us is personalisation and product recommendations. We want to create a world where you can spend exactly as much or as little time you want to on grocery shopping, including no time at all. This means that we have to be exceptionally good at predicting what you need and when you need it.

For instance, you might want to have basic commodities magically appear in your shopping cart. And you want to be able to choose this week’s dinners among a selection of recipes; a list we have generated based on recipes you have previously purchased or combinations of foods you buy often. Maybe you also want our algorithms to replace some of the default ingredients in those recipes with your favourite brands?

Or maybe you prefer to be hands-on and do everything yourself? We’re fine with that too!

The possibilities are endless, and we want to offer our customers a shopping experience that morphs and evolves based on how their preferences change over time.

In order to do this, we have to be at the cutting edge of the recommender engine field. We are in a great position to do this, since our loyal customers shop frequently and abundantly. By doing so they are providing us with rich datasets that we use to build algorithms based on state of the art research and technology. In return they get a fantastic product that makes their lives easier.

3. Tuning our warehouse operation down to the finest detail

Optimisation of our warehouse operation is also critical for us; first and foremost to prove to the world that our business model is sound and that we are progressing towards being the world’s most efficient retail system, but also to be competitive in a cutthroat industry where margins and prices are under constant pressure.

Our product range contains a lot of perishables, such as fish and vegetables, which means that we have to get our goods in and out of the warehouse in an extremely efficient manner. Everything from the number of boxes we need to pack each order in to the placement of potato chips relative to the dip mix in the storage shelves matters for the efficiency of our operation — and by extension our profit margins.

In this area we have a veritable mountain of analytical tasks and optimisation problems which we have only just started climbing.

4. Understanding who our customers are

Last, but definitely not least, we are very dependent on detailed segmentation of our customer base to design and tailor different types of marketing, offers and other commercial initiatives. Our segmentation models, which have been developed by our data scientists, is an integral part of everything from budgeting to strategic planning at Kolonial.no.

We hope you have enjoyed this little teaser on how we apply artificial intelligence for building the world’s most efficient retail system.

Hungry for more? We look forward to telling you about how we do all of these things in more detail in the future!

Stay tuned for updates on how we do data and artificial intelligence at Kolonial.no.

--

--