The Challenge

Our client produces more than 500 different products (medicines) that are sold all over the world. To optimise their production process, they need to better forecast their sales volumes for the coming months up to a year. This is not easy, as the sales volumes are different for each product and different parameters influence the growth for each product, such as seasonality trends, the weather, historical growth, etc. Our client asked us to develop a prediction model for each product and build a tool with which they could analyse and optimize the forecasted volumes in an efficient way.

The Approach

We started of developing the forecast model in our own environment. By gathering, understanding, and cleaning the data and by making the first predictions. Then we optimized the model stepwise by investigating multiple models and new data that could improve our output.

After we developed the forecast model and showed our client that the model performed better than their current, more handmade, model, we started building the tool around the model. There we focused on the end goal, making sure the users of the model could identify the flaws in the forecast very easily and optimize the forecast sales by hand. We worked in sprints throughout the whole project, making sure we stayed close to the client needs and get feedback as quick as possible. We managed to deliver a working tool including the forecast model within 3 months, on time and within budget. The implementation started straight after and did not finish until the client uses our tool with great pleasure.

The Solution

We created a forecast model for sales volumes of more than 500 different products. Next to that, we created a tool to analyse the forecast and optimize the predicted sales volumes in an efficient manner.

Impact and benefits for the client

Now, the client has much more grip on the sales volumes, so that they can:

  • Optimize their working capital (with less production excess and therefore less risk of depreciation of stock)
  • Have less out of stock products (with higher sales volumes and better service for clients)
  • Have more stability and regularity within the organisation (with less disturbances and more uniform processes)

Client – The whole process was interesting to witness: from our question, to the proposal, sharing the data, stepwise development of the forecast model and dashboard, to implementation and optimizing the predictions by pinpointing the deviations with intuitive visualizations. All done in a very timely manner!”

Our Learnings

We showed:

  • How to quickly develop a forecast tool with all sorts of information in close collaboration with the client
  • How to build a tool that picks out “the needles in the haystack” of predicted sales volumes that need optimizing for specific products

Curious what we can do for your organisation?

Would you like to know more this case or what the Amsterdam Data Collective can do for you? Get in touch with Dennis Diederix at, or check our contactpage.