At least once in one or two months I read a blog, discussion or an analytical note which raise objections against MMM recommendation on the budget split between media advertising and price promotions.

In some companies, advertising or media budgets (typically managed by the Marketing department) and a trade budget (managed by Sales) are determined separately. On the other hand, frustrations of media planners over shrinking budgets indicate that many brands maintain more holistic view of marketing activities. Their marketers are trying to find a right balance between the two areas.

The objections are typically raised by the media side stating that advertising budgets are cut while budget for promotions are gaining.

If short-term targets set large sales volume as a priority, an immediate effect of advertising campaign is indeed often below incremental sales generated during price promotions.

Decisions based on the lift - incremental unit sales above the base sales - against the cost of respective activity may lead to cost-based adjustments preferring promotions at the expense of media advertising. The effects become self-fulfilling with advertising driving less and less incremental sales.

But are mix models to blame? Unlikely. Model-based optimisation is driven by brand targets and KPIs. The focus on short-term gains typically has repercussions on longer-term results.

The question arises how to get the split between media and promotion budgets right? How to achieve sustainable growth of brand? Should marketers completely dismiss price promotions and focus instead on building the brand equity and long-term customer loyalty?

Before making any decision, it is critical to properly measure overall gains and costs of different strategies. In addition, scenarios need to be evaluated for several indicators of brand performance such as sales volume, revenues, profit, market share and brand tracking metrics.

The role of models is to facilitate such analysis. For this purpose marketing mix models need to cover a sufficient time period and provide an adequately detailed outcome. Let’s discuss these points more in detail.

Model Time-Window

A reasonable modelled period lies between two to three years of data.

  • A longer-term effect of advertising needs to be measured over extended period of time to have sufficient variation in the budget and other factors, e.g. seasonality, level of competition, consumer trends etc.
  • Some events, for example price change, may happen only occasionally and a modeller should make sure that all important events are present in the modelled period. Otherwise, one cannot report their outcome and use it for scenario projections.
  • If a longer model window enables covering several occurrence of the event, such as TV campaign, specific price promotions, the estimate of this impact will be more robust.

The Impact of Advertising

Advertising effects consist of a number of components. Modellers should go an extra mile to explore which elements of brand performance can be legitimately attributed to the communication activity.

Short- to Medium-Term Impact of Advertising

The main direct advertising effect is measured using proved metrics of media exposure. The estimate should include memory effects of advertising, diminishing returns, wear-out effects etc.

Longer Term Effect

There are several approaches to estimating longer term effects of communication activities. Generally, the focus here is not on the incremental impact of advertising discussed in the previous point. This is about the effect on base level dynamics not explained by other factors.

Base level of sales is the result of past bran-building efforts. In addition to WOM, referrals, improved retention of customers, part of its variation can be attributed to past media communication that attracted and retained brand loyalists, triggered buzz and discussions, shaped brand perceptions. The impact can be measured through past advertising spend or by an intermediate indicator of brand performance such as tracked awareness or consideration.

Indirect Effect of Advertising

In addition to the direct effects of communication activity, an analysis should be undertaken to determine which other contributions to sales can be attributed to effective advertising.

These include but are by no means limited to:

  • Increased distribution / improved visibility in retail store due to advertising
  • Lower price elasticity, which implies lower negative effects if a price is raised
  • Lower elasticity to competitor advertising and promotions

Communication with stakeholders and a systematic study of all factors are critical to accurately quantify how much of indirect effects can be actually attributed to advertising. Nevertheless, if causality is there, it should be included in the overall impact.

Effects of Price Promotions

Immediate Effect on Volume, Revenues and Profit

The estimation of immediate impact of price promotions is straightforward. It is generally captured by a price discount variable. More detailed modelling can take into account differences in promotion mechanic, diminishing return over weeks and non-linear impact of discount depth.

Revenue Analysis of Immediate Effect of Price Promotions

Even a simple analysis of revenues sheds more light on the immediate impact of price promotions and their viability. Let’s assume a short term target is to maximize revenues instead of sales volume. What does this imply for price promotion planning?

The chart on the left depicts the frontier line which shows for each discount level, the required volume uplift above the base line for revenues to break even. If the actual volume uplift during price promotions is below the line, the marketer is better off without promotion. The main takeout is that with rising discounts one needs exponential growth of volume uplift to achieve the same revenues as without price promotions.

The S-shaped curve line represents the response of sales to promotions at different levels of price discount. It is based on MMM estimation. In the given example, the expected marginal response actually diminishes and becomes flat which makes deeper discounts unviable.

The response curve indicates that from a revenue perspective, the discount should be between 10-45%, with the optimal level around 30%.

Post-Promotion Dips

Depending, what drives the volume uplift during price promotions, brand sales can suffer dips in the weeks following the promotion. This is mainly the case if:

  • Volume uplifts are driven by cupboard stocking of base customers and
  • Category demand is inelastic, i.e. cheaper price does not increase the overall consumption in the product category

This means, if most of extra purchases during price promotions are driven by cupboard stocking of customers who would purchase the product anyway and if at the same time, consumers can use only certain amount of product over time (e.g. toothpaste), the post-promotional dips in sales are likely. The extent of this effect should be quantified.

Cannibalisation Effects

In addition, if the volume uplift is driven also by those customers who would otherwise buy another product from the brand portfolio, (or a different size of the same product), the actual net volume uplift is more moderate compared to an initial estimate.

Longer-Term Effects of Frequent Price Promotions

Even though the immediate positive impact of price promotions is easily picked up in mix modelling, the effect does not stop there. If a brand relies largely on price promotions to achieve sales targets, consumers tend to factor into their purchase decisions an expectation of the next promotion. On top of that, a recall of past prices the consumer has paid leads to a new internal or temporal reference price which is below the actual base price.

What does this mean for the brand? Decreasing base sales may arise in response to regular price promotions. The main factors behind - lower reference price and changed perception of brand by customers - are reflected in higher price sensitivity

All the discussed points should definitely be taken into account for the optimal split of the marketing budget between individual activities. An optimisation based on the full range of relevant factors quantified over the sufficient period of time is a prerequisite for sensible split recommendations.

Optimal splits can be generated for individual KPIs, such as sales volume, revenues and profit. This enables considering different scenarios and their trade-offs.

How Customer-Centric Analytics Make the Split Less of a Problem

Price promotions are mainly associated with FMCG using retail stores as the main or the only distribution channel. Products placed on shelves just next to their competitors and private labels are under pressure to attract sales and keep their shelf space. However, the stronger are the base sales and the smaller is the scope for volume uplift, the more costly is to do market-wide price promotions. For new brands, a positive effect of increased sales is in the longer run limited.

Now, let’s think one step further. Instead of optimizing the budget split between media advertising and price promotions, it might be better to ask:

How can marketers go about building the brand in the long run and at the same time maximize sales in the short run?

Marketers should definitely focus on building brand equity. The other objective of maximising sales should be driven by proper understanding of customers. Who are my customers? How can they be segmented based on their characteristics, behaviour and attitudes towards the brand? What does segmentation imply for media planning to acquire new customers? What retention strategies would work for individual segments?

Advances in customer-level data analytics offer more possibilities to marketers of brands sold online or where most of customer communication is tracked given the nature of the service or through brand loyalty cards. Customer segmentation allows offering price discount only to the most price-sensitive customers. The product offers can be personalized and tailored to customer needs. Discount coupons can be sent via email or mobile phone. This option is increasingly used in a number of product categories – mainly in companies that invest into insights and analytics.

But how about FMCG that rely mainly on networks of retail stores to get their product to customers? In store distribution, the product is located just next to its competitors, some of which might promote aggressively. A moderate approach to price promotions will very likely cost business some sales from switching customers.

However, there are more opportunities to apply insights also to products sold through retail stores. Segmentation allows identifying the most price-sensitive customers. These can be offered discounts through their loyalty cards, communicating the offers via email or mobile application. As for potential new customers, or even the switching ones, these can be still kept using coupons in store catalogues. Price-sensitive bargain hunters are likely to go extra step to get a discount. If these efforts are successful, some marketers can reconsider their approach to market-wide price promotions, and to gradually phase them out.

This strategy will free resources for brand-building activities. Stronger focus on brand will in turn lead to more loyal customers and we will hopefully see fewer complaints about the media budget decline at the expense of brand harming price promotions.

Sydney, 30 May 2013,

Elena Yusupova, PhD