This site uses different types of cookies, including analytics and functional cookies (its own and from other sites). To change your cookie settings or find out more, click here. If you continue browsing our website, you accept these cookies.
As a new and successful business, adidas Western Europe eCommerce keeps on growing faster and faster; new services are being launched every week, an increasing number of marketing campaigns are being driven simultaneously, etc. This leads to more and more products having to be shipped out every day to our end consumers.This strong growth leads to an exponential increase of the complexities when it comes to forecasting our units and orders volumes, but also to bigger costs in case of forecasting mistakes or inaccuracies.
As these outbound volumes keep on increasing, we were being faced with the need to develop a new, more accurate, more detailed and more flexible operational forecasting tool.
Such a forecasting tool would have to cater to the complexities of having to forecast for 17 different markets rather than a single pan European entity. Indeed, warehouse operations and customer service depend on a country level forecast to plan carriers and linguistic staff. This is a very unique situation where on top of having a rapidly growing business we have to take into account local marketing events and markets specificities.
Finally, given the importance of ensuring consumer satisfaction through timely delivery of their orders, we also decided to provide a daily forecast for all 17 markets rather than the usual weekly format. Such a level of details improves the warehouse's shipping speed but also increase once again the difficulty of our task.
Describe the working solution:
Our first challenge was to find reliable sources of information. Both business analytics (financial and historical sales data) and web analytics (traffic information) data were already available to us through SAP HANA and Adobe Analytics. However, none of our databases were capturing in a single place all information related to marketing campaigns, project launches, events, adhoc issues, etc. That is why we started by building a centralized knowledge database, which contains all past and planned events that can impact our sales and outbound volumes. This tool is based on an Alteryx workflow, which cleans and blends together all the different calendars used by the other eCommerce teams. In the past, bringing those files together was a struggle since some of them are based on Excel while others are on Google Sheets, moreover, all are using a different format.
We made the best of this opportunity of now having a centralized event database by also developing a self-service visualization tool in Tableau, which displays all those past and future events. Such a dashboard is now used to:
1. Give some background to our stakeholders about what is driving the volumes seen in the forecast.
2. Have an overview of the business during our review the sales targets of the coming weeks, etc...
In a second time we created a workflow, which thanks to this new centralized event database, defines for each past and upcoming days as well as for each markets a set of "genes". These genes flag potential ad-hoc issues, commercial activations, level of discount, newsletter send outs, etc. This gene system can then be used to define the historical data to be used to forecast upcoming periods, by matching future and past days that share the same or at least similar genes. This is seen as the first pillar of our forecasting model.
The second pillar of our forecasting tool is a file containing our European weekly targets. These targets are constantly being reviewed based on new events shown in the centralized event database and current business trends. An Alteryx workflow derives from this target file our sales expectation for each upcoming day, market, category (full price, clearance) and article type (inline or customized). In order to do so, we use historical data defined by our genes in addition to a set of algorithms and calculate the sales impact ratio of each market and category. These ratios are then used to allocate a target to each one of the combination.
Finally, both pillars are brought together and we derive in a final Alteryx workflow, how many orders and units will have to be placed in each markets and from which article type.
However, since certain periods of time have genes combinations that cannot be matched, our working solution also gives us the flexibility to manually override the results. These forecast volumes are then shared with the team, warehouse, customer service call centers, etc. through a Tableau dashboard.
Describe the benefits you have achieved:
Thanks to the work that went into developing this new forecasting model in Alteryx, the adidas WE eCommerce business ended up getting:
A more accurate forecasting model, which allows for a better planning of our operations.
Reduced operational costs.
A more detailed forecast as we can now forecast on a daily level, when past methods required much more work and limited us to a weekly forecast.
A flexible forecasting model that can easily be modified to include new services and sales channels.
A forecast dashboard that lets us easily communicate our forecast to an ever growing number of stakeholders.
A centralized events calendar; that can be used by the entire department for much more than simply understanding the forecast (e.g. it is used to brief in Customer Service teams on upcoming events).
A massive amount of free time that can be used to drive other analyses, as it is not required from us anymore to manually join together different marketing calendars and other sources of information, create manual overviews of the upcoming weeks, manually split our weekly sales target, etc.