From myth to maths: Harnessing the halo effect of promotions

At any given time, up to one third of a retailer’s assortment is on promotion. These promotions are set up in such a way that the expected uplift in volume and the subsidies provided by suppliers will offset the negative margin impact of the discount. So much for the theory. In reality, many promotions don’t turn a profit at all, or at least they don’t add nearly as much profitable revenue as retailers expect. To manage promotions according to their real ROI, retailers would have to be able to quantify the sales of non-promoted items triggered by a promotion, the so-called halo effect. But so far, there has been no reliable method to quantify the halo effect of promotions. Even more sophisticated approaches to measure the impact of promotions have usually been limited to the items on promotion.

A holistic approach to full store promotion impact modeling

However, thanks to advances in statistical modeling, computing power, and a growing base of ever richer transaction data, a new method to measure the halo effect of promotions is emerging.

For the first time, it is now possible to measure the impact of promotions on non-promoted products across a store’s entire assortment with sufficient reliability and precision to take action. To what extent does the promotion trigger substitution and what is the influence of the promotion on average basket size? Do buyers of promoted items also buy other, regular priced items that they would normally not be buying?

In the past, the answers to these questions were a matter of experience, mixed with a fair amount of anecdote and speculation. Now, these questions can be answered with scientific rigor. State-of-the-art models differentiate between baseline sales (that would happen at cull price without promotion), promotional sales (sales of promoted items), and full-store lift (incremental sales of non-promoted items as a result of promotions). The foundational promoted item category models have to have very low forecast error to accurately assign full-store lift to category and item level promotions. Other features of our breakthrough include:

  • Control for external factors, such as store-wide coupons, seasonality, and floating holidays to isolate the impact of the promotion itself.
  • Bottom-up: Treating each basket as an observation that allows reconciliation at lower grains, i.e. review of relevant basket archetypes, to avoid double counting and separate cause from correlation.
  • Top-down attribution to category promotional linkage and, then driving down to the KVI or Ad Block level.
  • Application of advanced statistical techniques to decouple the impact of a given promotion from other simultaneous promotions in terms of drivers of Full Store Lift.

In effect, the latest generation of models provides retailers with the means to review the full-store impact of past promotions and to design future promotions to maximise true ROI, recognising both the direct impact and the halo effect. Typically, it requires three to five years of past transaction data. The bigger the data base, the better. In most cases, the implementation of systematic measurement of the halo effect of promotions is part of a bigger transformation toward more data-driven decision making in retail.

From myth to maths

The objective of halo effect modeling is to supply category managers with the facts they need to make better, more holistic ROI decisions. Many of them, however, are sceptical at first. But once they see the value, they don’t want to go back to rules of thumb.

Retailers will benefit from putting rules of thumb to the test. Often, it turns out that only a small group of SKUs is fit to balance the objective to drive revenue with the objective to protect profitability, and this group of suitable items varies greatly across categories, regions, and even individual neighbourhoods. Also, consumption patterns and shopping habits change over time. This is why it pays to move from myth to maths.

Retailers who adopt state-of-the-art full store ROI, including halo effect modeling, can often reduce their promotional spending by up to 10% or re-allocate these funds to more promising commercial levers to achieve a full percentage point of margin expansion. For retailers with single digit total profitability this is a very big opportunity. Building that into your everyday way of working via process, tools, and central support can happen in as little as six months.

Working with data-driven decision support systems trains category managers to think more critically about the real effect of promotions and prevents them from making excuses. Consumers want creative promotions that catch their eye and are aligned with their own shopping behaviour. The retailers that succeed will draw more customers, build brand loyalty and drive up profits. It’s time to take a new data-driven approach to promotions.

What’s Hot on Essential Retail?