Finding the ‘next Max Verstappen’ with data and smart algorithms.
Van Amersfoort Racing
Prediqt
Predict sales and support store ordering with data.
Align orders with expected demand for each product and location.
Trusted by leading organisations
Why we built this,
Store orders are often put together based on experience, historical knowledge and standard order lists. Store managers have to estimate how much stock is needed for each product, while demand can vary by location and period.
Ordering too little leads to empty shelves and lost sales. Ordering too much creates excess stock. For fresh products, it can also lead to more food waste.
Prediqt uses historical sales data to forecast future demand. Based on this, the application generates an order recommendation that can be used within the existing process.
What you can expect,
A demand forecast by product and location
The tool analyses available sales data to forecast demand for an upcoming period. It can account for differences between stores, products and sales periods.
Which factors are included in the forecast depends on the available data and the question at hand.
From forecast to order recommendation
A forecast is turned into a specific recommendation for the next order. This needs to account not only for expected sales, but also for current stock and the applicable ordering conditions.
The order recommendation can, for example, take into account:
- Historical sales
- Available stock in store
- Expected demand
- Lead times and ordering schedules
- Agreed stock limits
Fitting into the existing ordering process
The recommendation can be made available through the workflows and systems employees already use. Store managers can review the proposal and adjust it where needed before an order is processed.
Our approach,
-
Start with a defined product category
We first examine how orders are currently put together and what data is available for that process. We then select a familiar product category and a small number of stores for the initial rollout.
This lets us compare the predictions with the existing approach. -
Develop and test a predictive model
We develop a model using the available sales data and test it against historical data. We look at which patterns are predictable enough and where additional information is needed.
Together with staff, we also assess whether the resulting ordering recommendations are useful in day-to-day operations. -
Integrate and expand in a controlled way
After validation, we integrate the solution into the existing ordering process. We monitor the difference between predicted and actual demand and adjust the model when needed.
The solution can then be extended to other product categories, stores or relevant data sources.
Not an order placed on autopilot, but an evidence-based recommendation that helps store managers better align demand and stock.
The tool supports store managers in their day-to-day decisions without overlooking their knowledge of the local situation.
What we’ve built before,
Three projects from our work with data and AI.
Further reading
Blogs and in-depth articles from our team.
24 Aug 2026
Three lessons from our AI projects
What we learned from dozens of AI projects: start with the decision, treat data quality as an organisational issue and arrange management in advance.
18 Aug 2026
From pilot to production: why most AI projects stall
Most AI projects stall not because of the technology, but because of data, teams and processes. Read what it takes to move from pilot to production.
25 Jun 2026
Responsible AI: from principles to practice
AI is developing rapidly. New models emerge almost every week, and more and more organisations are experimenting with AI.
Get in touch
Stef Stoelwinder
Business DevelopmentE-mail info@datasciencelab.nl
Phone +31 (0)20 3636 724
Frequently asked questions,
Historical sales data by product and location provides the foundation. Depending on the question, stock levels, lead times, promotions and other relevant factors can also be used. We first assess what data is available and whether it is reliable enough to process.
The proposed solution allows ordering recommendations to be integrated into existing systems. How this is set up technically, and whether an employee approves orders first, depends on the current ordering process. The available integrations and degree of automation still need to be confirmed.
A store manager may have local information that is not yet reflected in the model. It is therefore important to establish how employees can review and adjust a recommendation. The exact options for manual changes still need to be determined.
Future demand can change due to promotions, seasons, local circumstances or changes to the product range. Historical data remains important, but must be used in the right context. The additional factors needed vary by product group and organisation.
We regularly compare forecasts with actual sales. This makes it possible to see when demand patterns change or the model performs less well. Monitoring and periodic adjustments are needed to keep the application aligned with new data and changing circumstances.