Mastering Promotions: Essential Strategies and Insights
Promotions are a critical lever for driving revenue growth in the fast-moving consumer goods (FMCG) industry. In a competitive market where shopper loyalty can be fleeting, well-executed promotional strategies not only boost short-term sales but also enhance brand visibility, stimulate the trial of new products, and encourage repeat purchases.
Promotions account for a substantial portion of FMCG sales, with some categories seeing as much as 50% to 60% of sales occurring on promotion. For instance, over 30% of all food sales in the European grocery market are driven by promotions. At the same time, in categories like personal care and household products, the figure can rise to nearly 60%.1
This significant reliance on promotions underscores their importance as a revenue growth management (RGM) tool. Effective promotions can help FMCG companies optimize pricing, manage inventory, and ultimately improve profitability.
However, the challenge lies in balancing promotional activity with maintaining brand value and margins. As consumer expectations continue to evolve, FMCG companies must fine-tune their promotional strategies to ensure they remain a powerful driver of growth while safeguarding long-term brand equity.
In this blog:
- Outline the main reasons why FMCG companies run promotions.
- Discuss the core promotion dynamics and traditional tools and methods for analyzing and managing promotions.
- Demonstrate how to best steer and optimize promotions using current advancements in AI technology that significantly improve RGM teams’ understanding of how different promotions affect sales, revenues, and profits.
Why promotions?
There are three principal reasons FMCG companies run promotions that are of key interest for RGM professionals (see Figure 1). First, they are used to improve relevant KPIs such as sales, revenues, profits, market shares, and promotion return on investment (ROI).
Further, retailers often require them, as they need promotions to attract shoppers into the store. Promotions and the manufacturers’ share in their investment are important elements of the overall business relationship and economics of the manufacturer-retailer relationship.
Finally, the ultimate goal of any promotion is to attract new shoppers who would normally not buy a product, test it during the promotion and then continue buying it at the regular price.
Figure 1: Why promotions?
Of those three, particularly the third objective is notoriously difficult to measure. Here, we will show how it can be done using AI technology.
Basic promotion dynamics
There are two basic effects at work when a product is promoted.
First, it creates an overall increase in demand that comes from the heightened attention on the product, particularly at the point-of-sale (POS). This results in a shift in the demand curve from (I) to (II), as shown in Figure 2. With this shift, sales at every price will be higher than without the promotion.
In addition, the typical promotion price discount further increases sales during the promotion. The graph shows a movement on the shifted demand curve towards lower price and higher sales.
Figure 2: Promotion dynamics
- Besides the direct effects on the promoted product during the promotion, there can be further effects on other own and competitor products at the same time, if shoppers buy the promoted product who would otherwise have bought something else.
- Further, there can be long-term effects if shoppers store non-perishable goods and then buy less in the coming weeks.
- There can be a sustained increase in demand because a promotion has attracted additional shoppers.
- Finally, promotions at one retailer may attract shoppers who would usually go to other retailers and reduce demand there.
Figure 3: Promotion Effectiveness
Figure 3 shows these different effects sorted by their proximity to the promoted product.
First, there are the immediate effects directly on the promoted product that typically lead to an increase in sales, and the indirect effects on alternative products not being sold because shoppers switch to the promoted products. This primarily affects sales at the same store but to a much weaker degree at other stores.
Then, there are two types of potential long-term effects: a loss in future sales because shoppers stock up on products, and a increase in future sales because the promotion has attracted new shoppers.
One key metric for determining the success is the promotion’s return on investment (ROI). It is computed by dividing the net profit impact by the promotion investment.
The promotion investment includes all direct additional expenses for the promotion. For example, additional discounts for the retailer or payments for additional marketing at the POS.
The net profit impact is the additional profit (gross margin) gained from a promotion compared to the profit without promotion. In most cases, the net profit is computed only for the promoted product during the promoted period. This ignores many of the effects highlighted in Figure 3. We will show how these can be included in the final section of this blog.
Traditional analyses
A key challenge for steering promotions is understanding how a future promotion will affect sales. Most of the RGM and sales teams we have seen work with two methods to determine the effects of a promotion:
- Promotion uplift: A simple way to capture the effect of a promotion is by determining the promotion uplift. For example, a “2 for 1” promotion increases sales by 400% compared to the baseline. The uplifts can be determined using a simple linear regression analysis or more advanced analytics considering seasonal trends, interactions, and a range of potential other effects.
- Promotion price elasticity: The more common approach to compute the effects of running promotions is using promotion price elasticities. Agencies provide these together with base price elasticities for key products. Table 1 shows an example of base price elasticities and promotion price elasticities for different promotion types. Here, the promotion price elasticities are typically larger (i.e., more negative). This means that a price decrease tends to have a larger effect on sales during a promotion than a regular base price change.
Table 1: Base and promotion price elasticities
While promotion price elasticities are widely used, they can also be misleading. To understand this, consider the two promotion effects shown in Figure 2: the shift in the demand curve and the movement on the new (promotion) demand curve towards the upper left corner because of the price discount.
With real data, the demand curves is not observable in full, but only at selected price points. Figure 4 shows this as an example of a product sold at three different price points: first at the unpromoted price of €4.99 and then after a price increase at the price of €5.99. For both price points, the promotion price was €2.99. This would result in a base price elasticity of -0.7 at a price of €2.99 (bold line) and two promotion price elasticities depending on the start point of -2.2 and -4.8.
This simple example highlights that the promotion price elasticity does not capture the promotion effect well. This is because the actual promotion effect comprises the two effects described in Figure 2, with different price discounts, and the promotion price elasticity aggregates these two effects. This can lead to great discrepancies if the base or promotion prices are changed, as the difference between -2.2 and -4.8 highlights.
Figure 4: Estimation of base and promotion price elasticities via linear regression
Another key challenge with both traditional methods of capturing the promotion effect is that they do not lend themselves to an easy extrapolation beyond the specific cases available in historical data.
For example, the above case already highlights that a specific measured price elasticity is only valid for the specific price change and cannot easily be applied to other price changes. The same is true for promotion uplifts. That means, if in the past only a 30% discount had been applied, it is difficult to assess what would happen after a 20% or 40% discount. Furthermore, assessing what would happen if another product is promoted instead of or in addition to, or if another promotion type is used, e.g., a “2 for 1” instead of the 30% discount, becomes so imprecise with these traditional analyses that most teams shy away from considering meaningful changes in how they run promotions.
To summarize, the two traditional methods can provide a decent approximation in a stable environment with little changes in how promotions are run. However, they are not made to explore the full spectrum of promotion opportunities. That is why most companies do not realize the full potential of their promotional activities.
Figure 5 shows an overview of the levers available to steer promotions. These include the promotion type, the selection of products that are promoted (together), and the discount level. Their combination provides a rich set of options to choose from to optimize promotions and how they interact with the other RGM levers.
Figure 5: Promotion levers
What keeps RGM and sales teams from exploring these options is the inability to assess how new promotions will affect sales, (net) revenue, and profit – and the risks associated with this. Here, generative AI-based RGM tools come into play. These allow teams to model the effects of new promotions. The next section outlines how this works.
AI for promotion analysis and optimization
Recently, RGM professionals have put high expectations on Artificial Intelligence (AI), specifically Generative AI (GenAI), to help improve analytics to better steer RGM levers and specifically promotions. RGM professionals have tried different techniques, ranging from various regression techniques to more advanced methods like gradient boosting, with mixed results so far.
At Buynomics, we have developed a different technology based on first principles called Virtual Shoppers AI. The technology is built to generate large groups of virtual shoppers that simulate purchasing decisions, similar to actual shoppers.
As illustrated in Figure 6, the system creates hundreds of thousands, or even millions, of virtual shoppers, each with unique preferences for brands, product sizes, and other attributes. These virtual shoppers also exhibit behavioral traits, for example, how they respond to price thresholds or a different placement of products in stores during a promotion. Virtual shoppers make individual purchasing decisions based on their specific characteristics and the options presented to them.
For instance, if the price of a product increases, it may result in a loss of sales as some virtual shoppers decide not to purchase it anymore. Some may switch to another product, while others might choose not to buy anything.
Figure 6 shows how changing prices for a product through a range produces a demand curve and, with this, insights into revenue, profit, and all other relevant KPIs. Similarly, altering a product attribute, like its size, could cause certain virtual shoppers to switch their preferences to or from other products. This comprehensive model allows for a holistic prediction of the effects of all RGM lever changes—whether in isolation or in combination—rather than focusing solely on price or promotion adjustments.
Additionally, this model helps users understand product interactions, such as the impact of changing the price of a single product versus the entire portfolio, or how a price change combined with a size modification or product delisting affects overall sales.
The virtual shoppers are trained using all available historical RGM data, including sales, surveys, and panel data. For sales data, the AI identifies the virtual shopper preferences that align with observed sales patterns across products and channels over time, given the specific prices and offers during the training periods. Moreover, survey results for a potential new brand can be incorporated into a model built from existing sales data and value drivers, enabling teams to evaluate the potential of a new brand with a robust model that integrates historical sales data and fresh survey insights.
Figure 6: Sales dynamics using Virtual Shoppers
This technology is particularly relevant for understanding and steering promotions. Figure 7 demonstrates this with a simple promotion example. It shows the offer of products A, B, and C in two periods.
In period 1, no product is promoted. In period 2, product B is promoted with a price reduction from € 7.99 to €6.49. The colors of the figures indicate what product each one would buy based on its preferences for the three products and their price sensitivities.
Figure 7: Example of Virtual Shoppers reacting to a promotion
The example illustrates how the Virtual Shoppers technology predicts the effects of promotions, including the interactions with other products, as 30 of the shoppers who bought product A in period 1, when no product was promoted, bought the promoted product B in period 2, and 20 shoppers switched from product C to product B. Similarly, virtual shoppers can stock up and buy less in period 3, and so on.
This provides insights into the promotion effects over time and allows RGM professionals to holistically assess promotion effects, not only for the promoted product, but also for other products and over time.
Further, the Virtual Shoppers AI allows users to model variations in the promoted products. For example, if only product B had previously been in promotion, then the promotion may be moved to product A or C, or the joint effect of promoting products A and B together could be evaluated. Together, this opens up a new universe of promotion options available to RGM and sales professionals.
Finally, the Virtual Shoppers AI also allows users to understand how promotions affect preferences and willingness-to-pay of shoppers who tested the product during a promotion, then learned that they liked it, and will continue buying it at the unpromoted price in the future.
In the video below, we show a specific scenario and how Buynomics platform can forecast precisely the promotion impact to the portfolio over the weeks.
Assessing promotions
Finally, we need to assess promotions. There are two perspectives that are important when assessing promotion effectiveness:
- The effectiveness of a single promotion
- The optimization of a full promotion plan
This assessment should always be performed to plan promotions beforehand, using predicted effects (e.g., using the traditional and AI-based techniques described above), and then after they have been conducted using actual sales data.
Effectiveness of a single promotion
Evaluating a single promotion (e.g., a promoted price reduction in one week for one or many products) is very straightforward conceptually, but it can be tricky in practice.
Table 2 shows the analysis for the promotion example from Figure 7. Product B is promoted in period 2, which increases sales from 200 to 400 units. In addition to the 30 units lost from product A and 20 units from product C, product B also loses 20 units in periods 3 and 4 each, as some shoppers move forward purchases from these periods into the promotion. With the prices, sales, and cost incl. the promo investment of € 0.40 per unit sold during the promotion, the promotion produces the following results.
Table 2: Example of a promotion analysis
The promotion has a positive profit of € 38 vs. the base case with a promotion ROI of 24%, if only the effect on the promoted product in the promoted period is considered. If the total effects in period 2 are considered, incl. all products, then the profit effect is negative with a promotion ROI of -104%. If all three periods (periods 2, 3, and 4) are considered, including all three products, then the promotion ROI is -204%.
The problem for many companies is that this outcome is not too uncommon. While the visibility of the latter two numbers is often tricky to achieve, as the cross effects are difficult to assess using traditional methods such as promotion price elasticities, they cannot capture cross effects between products and periods well.
Optimization of a full promotion plan
By evaluating individual promotions, teams can choose different types, discounts, and promoted products that best fit their objectives. As effects may differ throughout the year, KPIs like the promotion ROI can differ between weeks.
A simple way to define a promotion plan is by computing the expected promotion ROI for all potential promotion weeks and then selecting the weeks with the highest ROI, considering constraints. These plans typically include a maximum promotion investment per retailer or channel and heuristics such as always leaving two unpromoted weeks between two promotions to mitigate the negative effects on promotion sales from shoppers stocking up on products.
Advanced technologies, such as Virtual Shoppers AI described above, support for a more precise and automated assessment of promotion plans by allowing teams to model not only a broader range of individual promotions but also the interactions between promotions and base sales over time. With this, a true promotion optimization that identifies the best promotion for each week, given the planned promotion spend, and precisely predicts the resulting sales and other KPIs is possible.
Get Started Today to Reach Your Goals Tomorrow
Request a demo today to see how Buynomics can support your RGM team and helps you make data-driven decisions.
Additional resources
Webinar: Promotion Optimization for Balanced Revenue and Profitability
1Sources: NielsenIQ, "Beyond Cost Cautiousness: What’s Next in FMCG Category Trends," October 2022. Circana, "Inflation Impact on FMCG Market: Prosumers Adapt Amid Declining Sales & Changing Behaviors in Europe's Top Markets," 2023.
September 17, 2024