Thirsty for more expert insights?

Subscribe to our Tea O'Clock newsletter!


How to accurately model Search Engine in MMM with Agent-Based Models

Romain Warlop
Published on
Search Engine is a peculiar media channel in that it behaves differently than other media, like social or even TV. A Search Engine is a pull media (in opposition to push media), making it only relevant if users are taking action toward your market. By design, search engines are also frequently last-click media, whose standalone impact is often overestimated when using attribution methods. Unfortunately, classic MMM usually categorizes Search Engine as any other media, which makes little sense. Even if other solutions exist, at fifty-five, we prefer to rely on a hierarchical method combined with agent-based models, or ABMs, which in our experience is often the most effective solution.

The hierarchical method

The first step is to model the impact of other media on Search Engine. Indeed, a TV campaign will lead to more search and, thus, more impressions and clicks. But those clicks should be “credited” to the TV campaign. With this first model, we are able to estimate how many clicks are “natural” and how many should instead be credited to other individual media. With that data acquired, we can move to agent-based models. 

Agent-based models

ABMs’ main appeal is their ability to model the behavior of consumers (agents) in response to stimuli (such as how many times a consumer sees or clicks on an ad). The marketing strategy will increase consumers' perception, making them more likely to buy from your brand. This is true for TV or social campaigns, but to a lesser extent, this is also true for SEO and SEA. However, assuming that Search Engine has little impact on consumer perception will result in a very low contribution rate, which has been proven wrong when experimenting. The reason being that if you don’t exist on search engines, consumers will click on and buy from your competition. To take this into account in ABM, if a consumer clicks on your brand’s search links, their purchase probability will be higher according to the model. Thus, among other terms that characterize a user’s purchase probability for a brand, search engine clicks play an important role along with classic perception. This allows us to mimic reality and obtain a significant contribution from search engines. 

But that’s not all. Search engines can be divided into 4 categories: brand / non-brand / SEO / SEA. Those 4 levels are taken into account in the hierarchical approach as well as in the ABM. For instance, in the ABM, thanks to CRM data and market surveys, we are able to create a model that estimates a consumer’s affinity with the market (likelihood to purchase, for instance) as a function of user attributes (gender, age, geolocation, behavior on the market …). We thus have two variables that describe agent affinity: perception, which is their affinity for your brand, and affinity with the market. Brand search engine clicks will be more correlated with market affinity AND perception while non-brand search engine clicks will only be correlated with market affinity. 

As a result, fifty-five’s hierarchical approach combined with our advanced ABM allows us to accurately mimic consumer behavior on brand, non-brand, SEO, and SEA, as well as their consequences on revenue.

All articles

Related articles

Why use reach and frequency instead of impressions in Marketing Mix Models?

4 mins
Romain Warlop

Data Science @fifty-five

4 mins

The decision-making process: an overview - Part 4: Nudging: the mechanisms of influence

5 mins
Sabrine Hamroun

Thirsty for more expert insights? Subscribe to our monthly newsletter.

Discover all the latest news, articles, webinar replays and fifty-five events in our monthly newsletter, Tea O'Clock.

First name*
Last name*
Preferred language*
Merci !

Votre demande d'abonnement a bien été prise en compte.
Oops! Something went wrong while submitting the form.