Trade promotions, where a food manufacturer or consumer products company negotiates some form of promotional activity with a retailer, are big business. But Nielsen research shows that 59% of Promotions fail to deliver value in Europe’s top 12 markets, with €27 billion wasted each year. So why is it so hard to plan and execute effective promotions?
The first step for any company looking to understand their trade promotions is recording all of the promotions that you are doing. In far too many cases this is a spreadsheet detailing the product, customer, dates and a predicted outcome that has not been validated after the fact. In more advanced organisations, actual performance data is incorporated to create a dataset giving actual promotion performance. But even where this valuable data exists, it frequently has no data validation and requires a lot of cleansing to make any sense of the data. While we don’t advocate for spreadsheet databases, if they are used as a tactical solution, please ensure your organisation has taken the 10 minutes required to add data validation!
TPM is primarily concerned with the process of recording and executing promotions, but not typically making sure the best promotions are planned in the first place. However, with quality data on past promotions and retrospective analysis of promotion performance, the data out of TPM is the best place to start. TPA is descriptive analytics of these historical promotions. The aim of TPA is to understand which promotions performed well so that you can run them, or similar promotions, again. Equally, by identifying poor performing promotions, you can avoid doing them again.
The simplest form of TPA takes all the promotions run each month, calculates the incremental volume, revenue and margin delivered and them calculates the return on the trade spend for each promotion. This data can then be visualised and plotted to show how promotions for different brands, categories or Account Managers perform over time. This information can be used by Category Management to make strategic decisions around the types of promotions to focus on, and where to focus trade spend to maximise impact on category KPIs. But to really move the needle on trade promotion performance, you need to optimise the specific parameters of each promotion.
Using the TPM historical promotion data, you can train a machine learning model to learn what factors are important in driving promotion performance. A primary driver is of course the type of promotion, but it will also learn the impact of the size and price point of the product, the time of the month and duration of the promotion, and which customer the promotion is run with. It is very hard to capture all factors which will drive promotion performance, and there will be many scenarios where promotions are not executed equally in store which we will pick up on later. But this is a much more robust and predictive approach to what your organisation may be doing already – does plotting price elasticity charts sound familiar? Well that is the same principle that machine learning uses, but with much more statistical rigor.
What a machine learning model allows you to do is predict the volume, revenue and margin impacts a promotion is expected to have. Add in the ability to scenario plan different trade spend values and formats and you have yourself a very useful planning tool for negotiating trade promotions. Add in the consumer retail price and you can understand your customers margin incentives and how much trade spend they are likely to ask for ahead of time.
Incorporating trade promotions into demand forecasts is crucial to hit customer availability targets and optimise supply chain operations. Yet even for many large organisations, this is not managed well. The best practice is to use a machine learning forecasting solution for the baseline demand forecast, which excludes the impact of promotions on demand. An additional layer is then added to this, specifically accounting for promotion impacts. These promotion impacts can be based on the previously mentioned machine learning promotion model, which will give the expected incremental volume for each promotion. This not only increases the accuracy of demand forecasts, but removes the need for manually estimating promotion volumes and trying to add these into the automatic forecasting process.
As a brief tangent, all of what has been discussed so far can be further developed by incorporating analysis of the halo or cannibalisation effects promotions have on other SKUs. TPA will incorporate the changes in demand of other SKUs during the promotion period, using a rolling baseline of sales prior to the promotion. TPO will use machine learning to understand what drives these changes in demand, and predict halo or cannibalisation for future promotions. Finally, these effects can be incorporated into the demand forecast to improve accuracy.
In addition to the traditional solutions discussed above, we have developed specific solutions for our clients. The first is a process which analyses planned promotions for the subsequent 2 months and predicts the expected incremental volume. It then calculates the expected financial performance of the promotion based on the trade spend and flags when planned promotions are expected to destroy value. There is also the functionality to run promotions with engagement based KPIs. The solution uses a threshold the company is willing to invest for engagement to flag when promotions are not expected to meet this threshold.
The second solution we created analysed the performance of live promotions based on ePOS data. Where promotion performance was >50% below the predicted performance, the promotion was flagged to the Account Manager for further investigation. In almost 40% of scenarios, these promotions were found to have been incorrectly executed at some or all the retail locations. This enabled timely correction of the issue and improved the return on trade spend by 7%.
Trade spend is one of the biggest expenses on the P&L and there are many ways that data-driven decision making can avoid poor promotions and improve the performance of others. Even the most mature organisations benefit from solutions like the two described above, both to improve return on trade spend and unlock time for more strategic activities.
Purpose AI is an AI and Analytics Consultancy delivering data-driven solutions for complex Supply Chain and Commercial decision making. We work across the Consumer Products and Retail Industries and have worked with many Top 20 CPGs and Retailers in Europe and North America. Our USP is that we have developed industrialised solutions to common industry problems, reducing the time and costs to deliver value. But unlike SaaS providers, there is no licensing costs, we customise our solutions to your requirements and embed a strong business process to ensure they deliver value.