Black Friday, Cyber Monday and Christmas sales – can we forecast their effects?

30 November 2015

Can we predict and manage the effect of promotions in supply chains? Are promotions beneficial to retailers? Nikolaos Kourentzes and Oliver Schaer discuss.

This year ASDA decided not to take part in Black Friday sales and recently Sainsbury’s changed its promotional strategy, resulting in a simplified and less intensive set of promotions, mirroring the position that other retailers like Aldi have maintained. There are many arguments to explain these moves. Nikolaos Kourentzes and Oliver Schaer focus here on two aspects: can we predict and manage the effect of promotions in supply chains; and are promotions beneficial to retailers?

With the end of year upon us, shoppers are searching for the best deals. Black Friday, Cyber Monday sales and the build-up of promotions towards Christmas highlight how retailers put a lot of effort to boost their sales. Crowded stores and busy online stores seem at the first glance to be appealing for retailers. Without a doubt promotions are highly effective in attracting consumers’ attention and generating sales. However, promotions of the various types that different retailers use are disrupting the demand patterns and impacting the supply chain. This results in a substantial forecasting challenge: getting the impact of promotions right.

Can we forecast promotions?

It has been argued that a reduced promotional portfolio can result in increased forecasting accuracy. One might get the impression that the effect of promotions cannot be forecasted effectively and therefore companies might prefer to use them more sparingly. In practice companies often use a baseline statistical forecast adjusted by experts to reflect promotional uplift, as well as pre- and post-promotion demand.

Research at the Lancaster Centre of Forecasting has shown that human experts add value to the forecasts, but their performance is inconsistent (Trapero et al. 2013) and should be aided by appropriate forecasting support systems to maximise gains (Fildes et al. 2015). Going one step further our research has shown that we can construct relatively simple forecasting methods to predict promotions accurately and automatically in a variety of settings (Trapero et al. 2014 and Kourentzes & Petropoulos 2015). Recent work has shown how to incorporate competitive and cross-effects in promotional forecasting models (Ma et al. 2015). To sum up, promotions are not unforecastable, on the contrary it is possible not only to predict their effect, but also understand the impact they have on demand uncertainty and plan accordingly.

Are promotions useful?

Promotions are used to drive sales. But are they useful? The standard argument is that promotions increase revenues, help to clear stock and generate spill over effects. Furthermore, retailers use promotions for competition reasons. A study by Nielsen showed that in the UK more than 30% of all products listed are constantly on promotion. Therefore promotions are no longer special or even unexpected and consumers have adapted their buying behaviour to account for them.

In many cases customers only buy when their desired product is under promotion. In particular for long lasting products such as clothing and household products people tend to hold back their investment until such products are under promotion. Research has shown that although promotions have a substantial short term effect, which may not always be profitable, there are limited long-run gains from promotions. In other words, products that have been promoted do not observe increased sales after the promotion has ended. The argument here is that although promotions are very effective at manipulating revenues and stock levels, the result on the retailers’ profitability is far from straightforward and often in the long-term there are no gains. Nonetheless, we should not underestimate the importance of promotional activities to manage market share and competitive actions.

So what should one do?

There is no generic solution and retailers must understand what are the effects of their promotional mix on their demand and the respective uncertainty. Innovations in forecasting and marketing analytics models permit us to capture complex promotional structures and understand their effect. These should be considered in conjunction with the various marketing instruments that organisations use. Only then it becomes possible to gain transparency on how promotions help meet the different short- and long-term objectives of companies.

Researchers at the Centre for Marketing Analytics and Forecasting are actively looking at understanding and modelling promotions, from the impact on demand and its uncertainty, their profitability implications up to how they propagate across the supply chain and impact the various organisations. A recent workshop on promotions and marketing analytics contained presentations from SAS, Nielsen and MarketingQED and a summary can be found here. Follow our newsletter to keep updated on the latest innovations in the field. If you are interested to know more or collaborate with LCF contact us