Understanding weather impact on beer sales in denmark


Keywords: Big Data, Beer & Beverages, Weather forecasts, Data mining, Sale intelligence, Sale forecasts.

Branche: Beverage

Data set: Yes

Focus Area

The weather has a significant impact on the sales of our products.

  • On the negative side, this has historically created problems for us in terms of e.g. bottlenecks and under-/over delivery to our customers.
  • On the positive side, we believe that a better and more detailed understanding of the dynamics of weather impact can give us a competitive edge since we need to be available with the right products when the consumers need them and want to buy them. Failing e.g. on ‘those 5 great summer days’ that can unfortunately end up being the only ones in a Danish summer can potentially have a big negative impact on the full year result!

For the past years, Royal Unibrew has grown to become a strong competitor and challenger to Carlsberg’s dominance in the Nordic region. We have done so through our focus on customer relations, agility, and innovation – and needless to say, through optimal use of the wide range of strong, well-known brands in our portfolio – in Denmark: Faxe Kondi, Pepsi, Nikoline, Booster, Egekilde, Royal Beer, Heineken, Lays, Doritos, Buggles. This has earned us the vote as the best supplier to the Danish retail trade for 3 consecutive years now.

We are also sponsors of some of the biggest music events in Denmark such as Smukfest, Distortion, Tivoli Fredagsrock, Copenhell, Heartland, and many more – and proud sponsors of a number of Danish superliga football clubs and sports events, including Champions League (Heineken and Pepsi) and Formula 1.

Can you help us become an even stronger challenger? Can you help us become even more agile and intelligent in the way we operate and serve our customers and consumers?


Imagine an intelligent system that will continuously give us e.g.:

  • A ‘weather modifier’ on normal/average baseline volume sold nationally, per region and possibly per store – e.g.:
    • ‘Weather beer volume modifier for Monday: -2%’, weekly +1%,
      • Århus area +3%
        • Stationsgade Superbrugsen +2%
        • Heineken 24 crate 33cl bottle +7%
        • Royal Classic 24p cans -2%
        • Etc.
  • On top of that, possibly a system that will then ‘learn’ from actual sales and hence continuously adjust the ‘algorithms’ to give more and more precise predictions on weather impact.

Hence: We need you to help us build a system that will use detailed weather data as well as weather forecast data on a regional level (sunshine hours, clouds, temperature, precipitation, etc.) throughout the year to better predict the sales of beverages across our big four categories (beer, carbonated soft drinks, bottled water, and energy drinks) and across store types and channels – e.g. convenience, small supermarkets, hypermarkets, discount stores. Finally, comparison of weather and weather forecast data to understand dynamics of forecast vs. actual weather.

That system will help us both produce and deliver the right amounts of the right types of products – and furthermore help us understand e.g. what products, pack types, etc. run more quickly out of stock with empty shelves, what to put on promotions and special displays (the ones with stronger weather impact), etc.

Finally, we would like the system to include the possibility to not only insert volumes but also margin figures on both our side and customer side so that we can optimize sales not only from a sales perspective but also in a profit perspective.

Depending on the possibilities, we would like to get as detailed an understanding as possible – i.e. perhaps down to SKU (product) level, store level and week, day and even day part level.


Lots of factors affect sales. We probably need to be able to filter these factors to isolate weather impact on ‘baseline volume’, e.g.:

  • Own vs. competitor…
    • Shelf pricing
    • Promotions: Secondary placements/displays, leaflets, price reductions
    • Assortment size, facings etc.
    • Customer stock
    • Cooler availability
    • Advertising, events and sponsorships
    • Store visits from Royal Unibrew personel
    • Delivery days


We imagine for a pilot study that we will be allowed access to very detailed data from one of our retail customers – e.g. COOP who are also taking part in Oi-X.

That means store level data down to SKU/product level for all products in the category with volume in both units and liters and literprices, at store level and at least at day level, possibly even at daypart level. Possibly with a split on promo/shelf sales. We have yet to clear up both if and what data we can get access to.

As a pilot project, we may also select just one category – e.g. beer – to start out with.

In addition to this – or as a backup if we are not granted access to the most detailed data, we can get access to:

  • Nielsen scantrack data (census data on SKU level but only down to weekly level and banner level – e.g. weekly data for Superbrugsen) with volumes, units and average prices
  • Internal data on own sales from brewery to customers (e.g. RU sales to Superbrugsen)
  • Promotion calendar per banner
  • Weather data on regional level (from DMI)
  • Brand activity/Share of voice data RU vs competitors