This article is an excerpt from our ebook Sales Promotion: How to improve the performance of your investments which you can download for free.
Manufacturers and distributors around the world spend nearly 500 billion dollars on sales promotions every year (source: Nielsen). However, the performance of promotional campaigns remains stubbornly difficult to measure.
To regain effectiveness, campaigns must enter a new era. New technologies, personalised marketing applied to promotions and the exploitation of the huge amount of data collected open up new perspectives.
Lower yields from promotional activities are not inevitable: here are two solutions on how to better measure ROI and optimise promotional budgets.
ROI measurement based on granular data, A/B testing and predictive analysis
Two methods of promotional ROI analysis, A/B testing and predictive analysis, both rely on the use of granular data, meaning data relative to:
- A person and a period of time (for example, consumption of a product by a person on a given day)
- A small segment and a period of time (for example, consumption of a brand’s products by a segment on a given day
Conversely, overly aggregated data (consumption by all consumers on a given day or total consumption ever by a customer) does not allow us to calculate the ROI.
We need to have the necessary variables: to run A/B testing, calculate the investment and calculate the profit.
Run A/B testing
Variables belonging to the control group or test group. If necessary, we can also supplement these variables with everything that has made a consumer eligible for a promotion.
Variables indicating who has been targeted with a promotion, who has benefited from a promotion and what a promotion has cost.
The incremental revenue, the number of target products purchased, the frequency of purchase or any other variable showing a positive impact from the promotion.
For predictive analysis, any data that helps to improve precision should also be included.
Predictive analysis to optimise promotional campaign scenarios
A/B testing and econometric methods help to effectively measure the impact of a campaign retrospectively. By expanding the range of available algorithms, machine learning helps to predict a consumer’s behaviour on a given date and for a particular product, with far superior precision than the usual calculations. When used to measure promotional performance, it can:
- Personalise for an individual consumer, rather than segment-wide, on a specific date for purposes of one-to-one marketing,
- Reproduce A/B testing results and make them more precise,
- Do an initial analysis of performance prior to the campaign and therefore completely review the promotional strategy in the mid-term.
This increased precision, obtained by creating behaviour scenarios, helps to personalise smart promotional messages.
These predictions, at an individual consumer and/or product level, can of course be combined with traditional statistical analysis (business intelligence, scoring) and re-used in strategic planning models.
However, we must bear in mind that these methods are made to be precise according to a given criteria (average relative error, for example) and aren’t necessarily ‘explanatory’. In some cases this can lead to ‘black box’ solutions. Nevertheless, there is generally a happy medium.