Forecasting demand for promotional offers is notoriously difficult for consumer goods retailers and suppliers. Some promotions fail miserably whilst others wildly exceed expectations. Most, however, perform within broadly predictable bounds and ALL benefit from a dual approach to planning and execution:
- Plan realistic outcomes based on an accurate understanding of past performance
- Monitor execution attentively and react quickly to shopper response
The rest of this article will consider the first of these; how to plan realistic outcomes for your promotions.
Realistic planning
Let’s be honest – many “promotional demand forecasts” are hazy guesses based on limited understanding of past promotions. We can do better; much better.
Whilst past performance is never a guarantee of the future performance, in aggregate past promotional patterns tend to provide strong evidence for the likely performance of your next promotion. Let’s compare the worst-case approach to forecasting promotional demand – typically performed in a spreadsheet or operational TPM tool – to a best-in-class approach using fine-grained data and advanced statistics/ machine learning.
Typical approach
- Collect data on past sales performance for product(s)
- Calculate total achieved sales performance over the past year and divide by 50 for a weekly average volume
- Call your weekly average volume the ‘baseline’ demand
- Pick a desired ‘uplift’ – anything between 2.5x and 10x the weekly average
- Calculate the promotional forecast as uplift x baseline
Demand intelligence approach
- Continuously collect daily performance data for all products
- Deconstruct achieved daily sales, looking for explanations of variance, including:
- Price changes
- Seasonal patterns
- Special events (Christmas, Halloween, Olympics etc.)
- Your own promotions; lifting your promoted products and cannibalising others
- Availability and supply
- Derive and test a counterfactual baseline that models underlying demand with ALL variances removed
- Derive probable uplifts from daily variances (normalised for availability, seasonally etc.) achieved in previous promotions; tuning for depth of cut, distribution, POS support etc.
- Calculate the promotional forecast on a daily basis over the planned promotional period, using your daily baseline and promotional uplift model
Clearly, the second approach takes many more factors into account but requires a lot more data and a lot more effort, as well as a fair amount of complex mathematics. In short, it’s not an approach that a typical demand planner can use without the support of powerful software tools. The data and the tools do exist but historically these have been large, expensive and highly customised. SaaS demand intelligence solutions are emerging which use the modern data stack and machine learning to build better baseline models quickly from rich, detailed demand data.
To build a realistic plan that you have a high probability of being able to execute reliably, you should:
- Research and evaluate demand intelligence solutions
- Source and manage rich, detailed demand data
- Apply appropriate machine learning algorithms to produce accurate baseline models
- Forecast future uplifts based on the lessons you learn from past performance
Agile execution
However well you plan, all promotions will deviate somewhat… and some will vary significantly for a variety of reasons:
- Your retail customer fails to order sufficient stock before the promotion launches
- Your retail customer fails to ship sufficient stock from their depots to stores before launch
- Store staff fail to present sufficient stock, or supporting materials, before launch, or
- Competitors run cannibalising promotions at the same time as your promotion
- Shoppers fail to engage with your offer, or
- Shoppers love your offer, and you can’t replenish fast enough
The first three issues, left unresolved, will reduce the apparent demand for your promoted products. Your sales volume uplift will be lower than planned because shoppers can’t purchase your products even if they want to; the products simply aren’t on the shelf as the promotion launches.
The fourth and fifth occur when your promotion – in context with all of their purchase alternatives – isn’t as compelling as you expected. In the case of competing promotions, you will probably seek redress from the retailer for failing to disclose a parallel promotion, but sometimes your supporting price cut, advertising or in-store presence just doesn’t appeal to the shoppers you’re trying to attract.
In the final case, demand exceeds your plan and you fail to meet it through inability to supply the required volume of products.
In ALL cases, daily review of promotional performance – “in flight” – will provide you with the intelligence you need to select appropriate actions. Daily review is essential when promotions typically last for three weeks and the first three days usually set the tone for the entire promotion. Fail to meet potential demand in those first few days and you can be sure to miss your goals.
So how can daily demand intelligence help you act to resolve issues as they arise? Let’s take each issue in turn:
- Monitor depot stock levels and order patterns in the 5-15 days before the promotion launches and flag any concerns to your retail customer; remember to check stock at each depot and ensure that you have adequate regional coverage for planned demand;
- Monitor depot and store stock – and depot-to-store service levels – in the 2-3 days before launch to ensure that stock is moving from depot to store in anticipation of your launch; remember to monitor stock levels at all participating stores – lack of stock build is a clear predictor of Day 1 problems;
- Monitor store stock, sales volume and average selling price on Day 1 of your promotion in all stores; identify any stores where selling price isn’t in line with your promotional offer, and where volumes are below your expected uplift (within an appropriate tolerance);
- Monitor competitor pricing using price visibility tools like Acuity Pricing so that you can understand whether competitor activity could explain reduced demand for your promotion;
- If you’re sure that stock and point-of-sale material are in place, and all stores have activated your promotion as planned, but shoppers just aren’t engaging and it’s not down to competitor activity then talk to your retail customer and flag the issue early – maybe they can reduce the price further to avoid excess stock or waste at the end of the promotion… and consider shopper feedback from the likes of Kantar to avoid repeating the mistake next time; finally,
- Wild success is the “nice problem to have” but it still represents a problem! Monitoring sales, stock and availability every day through the first week of your promotion will enable you to detect sales above your expected uplift and pull future planned shipments forward and – where possible – adjust production runs to meet your increased forecast.