Forecasting for Retail in an Amazon World

The “Amazon effect” has already dramatically changed the landscape of retail but the disruption goes much deeper than just the convenience of 24/7 shopping and next day delivery. It has changed retail cycles as retailers are forced to change the timing of their typical end of season sales to align with the Amazon driven Black Friday and assorted Amazon Sale Days throughout the year.

As a result, forecasting for a retail business or one of the myriads of distributers in their supply chain is very different to what it was just a few years ago.

With e-commerce comes e-planning and if you are not already forecasting differently you will or should be.

Online shoppers don’t want the pre prescribed experience of the traditional department store, they expect an experience tailored to their likes and preferences. To exploit these differences, we need to plan and forecast more flexibly and take into account macro and micro changes in consumer behaviour.

Traditionally, forecasts were based on historical demand and performance and the assumption that largely history will repeat itself.

In principle these methods hold true but without the ability to build and test new scenarios into a forecast, these forecasts will struggle to keep up with a rapidly changing dynamic landscape.

In today’s landscape, changes in the marketplace are swift, sudden and don’t follow a historical trading trend or pattern. Just looking at historical trading will not give you the whole picture.

To have a credible forecast you need to make judgements based on multiple scenarios, such as changes in buying habits and peak sales times, introducing new products more frequently than has been done previously, flexing your work force around these new dynamics.

To predict demand people are turning to machine learning to try to understand the drivers for demand of what and when, Big Data gives us access to an almost infinite array of parameters to consider, however without analysis Big Data is just that, a lot of data without any credible output.

How to harness the power of the available data to produce an accurate and credible forecast for your business is one of the many challenges we’ve set ourselves at ProForecast. We work together with local Universities to harness the power of Machine Learning to improve the reliability of our output. In the meantime, we offer you the ability to customised and run multiple scenarios both micro and macro to come up with the best way of planning for demand and the associated costs moving forward.

Visit us at and book a demo to see the most powerful forecasting application on the market.

Mark Harrison

Chief Commercial Officer