Long-term effects of advertising

When marketers aim to propose or justify their advertising budget, the question arises:

What are the long-term effects of our advertising?

Instead of a deer-in-the-headlights reaction, marketers should welcome this question as it shows a willingness to treat advertising as investment, and allows to make the case where it should get prioritized over actions with larger short-term but smaller long-term effects, such as price promotions or temporarily boosting the sales force. Indeed, as Michael Wolfe notes for the 26 brands he analyzed for the leading picture , the short-term ROI of advertising is often negative, while the long-term ROI is often positive. Thus,

For those who have been relying only on short-term ad metrics, we think that this has often resulted in less investment in advertising, overall.

Fortunately, the long-term effects of advertising have been studied for decades. John Little specified back in 1979 that any model of advertising should incorporate industry wisdom that it is not just having an immediate effect on sales, it also increases sales in later time periods. This can happen in several ways:

1) delayed ad effect: consumers take a while to act on the ad

2) customer holdover: consumers act on the ad, and repeat purchase

3) distribution: retailers may react to ads by giving the brand better shelf space

I will focus on the delayed ad effect here, as modeling it involves interesting choices.

The delayed ad effect is typically tied to the purchase cycle, e.g, customers may not be currently in the market for the product, or consumers have to wait for a doctor's visit to ask for the advertised prescription medicine. In such situations, ads can build memory structures and/or keep the brand top of mind when the purchase occasion arises.

A key question has been how to model these 'lagged effects'.

In principle, lagged ad effects can show wear in and wear out, just as the product introduction and price promotion effects quantified in my sole-authored Marketing Science paper on how consumers, competitors and company decision making drive long-term marketing effectiveness:

?In this first (dark line) scenario, the peak impact of advertising is only reached after a few periods (wearin), after which the impact declines (wearout) until indistinguishable fom zero (return to baseline sales).

However, capturing this wearin and wearout requires modeling each lagged effect (t-1, t-2, etc), which requires many parameters to be estimated. This seems overkill for the second scenario of a promotion in above figure (grey line): the peak impact is immediate (no wearin), and the wearout is gradual, similar to exponential decay in chemistry. In such scenario, we can capture the dynamic effects with just 2 parameters: the immediate ad effect, and its decay.

Enter adstock, coined by Simon Broadbent. The adstock theory hinges on the assumption that exposure to television advertising builds awareness or 'goodwill' in the minds of the consumers, influencing their purchase decision. Each new exposure to advertising builds awareness and this awareness will be higher if there have been recent exposures and lower if there have not been. In the absence of further exposures adstock eventually decays to negligible levels.

?This mathematical model of exponential decay is more broadly known as the Koyck model, created by Dutch economist Jan Tinbergen and developed by the economist Jan Koyck.

?

What is the likely decay rate of adstock and its impact on future sales?

The rule of thumb is that the long-term effect of an ad is twice the short term effect, which implies an exponential decay parameter of 0.5 (1/1-0.5 = 2). This doubling was first observed in single-source field experiments were households were exposed to a heavier ad schedule for 1 year, and continued to show higher sales after the schedules were equalized. It was confirmed by Wildner and Rodenbach (2015) in their analysis of 204 German TV ads:

Importantly, the decay rate of adstock is estimated and may differ by campaign and channels (Zaldivar 2023) (Kantar).?? The Kantar data for weekly decay rates of ad recall are show in the figure here.

1) TV ads tend to have a slower decay rate, with recall retention at 0.7-0.9.

2) ?The other decay rates are as shown, with digital (display) having the greatest decay rate, retaining weekly recall at a 0.1-.06 rate.

3) All rates presumably depend on more granular factors such as the length of the video, the context of the placement, etc.

?

All of these analyses show that the return of advertising is probably significantly underestimated by typical marketing incrementality measurement methods.