What is adstock effect in marketing?

There are numerous factors that can impact the performance of an ad campaign, however, one of the most crucial ones is the Adstock effect which can be easily overlooked. The Adstock effect is a phenomenon that can help us calculate the impact of an advertisement over a period of time, rather than immediately and in one instance. This phenomenon describes how an ad's impact builds up gradually over time, rather than all at once. This is because people tend to see the same ad multiple times, which makes the message remain in their minds and have a cumulative effect. Adstock is especially valuable for campaigns that need long-term exposure, like those aiming to raise brand awareness or encourage people to make a purchase.

What is the adstock effect?

To put it simply, when you see an ad multiple times, it stays in your mind for a longer time, and you might remember it even after the ad stops playing. This is why companies show their ads many times, so you remember them even when you're not looking at the ad anymore! We are sure you can think of at least a few brands you know and most likely associate a certain advertisement style to them. Probably one of the most famous example of adstock effect were the Pepsi ads from the 2000s. Everyone remembers the celebrities and the cool vibe behind their ads. The adstock effect helps build a brand identity but also helps customers remember specific products or offers.

There are two components behind the adstock effect. Namely, the diminishing returns and the delay/lagged effects/carryover. The first one essentially means that the impact of advertising on sales tends to decrease over time. This means that the more an ad is shown to the same audience, the less effective it becomes. The second component is the delay/lagged effects/carryover effect, which refers to the fact that the impact of advertising on sales is not immediate, but tends to be delayed. This means that the effects of an ad may continue to be observed even after it has stopped being aired, due to the fact that some consumers may have been influenced by it and may still be making purchases as a result.

At its core, the adstock effect is the idea that the effects of advertising don't just happen in the moment when an ad is seen or heard. Instead, the impact of advertising persists over time, even after the ad is no longer being shown or heard. This is because advertising creates a kind of memory or impression in the minds of consumers that can influence their behavior and decisions for some time after the initial exposure.

However, these effects don't just disappear once the ad stops running. Instead, they tend to decay over time, gradually losing their power as the memory of the ad fades. This decay is what is known as the adstock effect.

Let’s look at a practical example: Your brand shows a TV ad to a group of customers on Sunday. Most likely, this ad will still have an effect on the customer behavior. Why? because viewers remember you, well…if you have their attention, then they remember you. However, statistically, people remember 74% of the ads images they see (source).

How can you measure the adstock effect?

There are multiple ways to measure the adstock effect. Let’s check out some of them:

• Conducting surveys: Marketers can use surveys to measure the awareness or perception of a brand or product over time. By conducting surveys at regular intervals, they can track how the impact of advertising is building over time.
• Analyzing web traffic: Marketers can use web analytics tools to measure how advertising is driving traffic to their website over time. By analyzing this data, they can determine the Adstock effect and adjust their advertising strategies accordingly.
• Using sales data: Marketers can measure the Adstock effect by analyzing sales data over time. They can determine the relationship between advertising and sales and use this information to adjust their advertising strategies.
• Analyzing social media engagement: Marketers can analyze social media engagement to determine the Adstock effect. They can track how engagement with social media content is building over time and use this information to optimize their social media advertising strategies.
• Running experiments: Marketers can conduct experiments to measure the Adstock effect. By varying the frequency or duration of advertising, they can determine the optimal advertising strategy for achieving their desired impact.
• Using Marketing Mix Modeling (MMM): As mentioned earlier, marketing mix models can be used to measure the Adstock effect. By including a lagged advertising variable in the model, marketers can capture the cumulative impact of advertising over time and optimize their advertising strategies accordingly.

How can you include the adstock effect in your marketing strategy with Marketing Mix Modeling?

In an MMM, the Adstock effect is typically captured by including a lagged advertising variable in the model. The lagged advertising variable represents the impact of advertising on future sales, which can help to account for the Adstock effect. By using a lagged advertising variable, the MMM can capture the cumulative impact of advertising over time, rather than just the immediate impact of an ad.

Aside from the benefits mentioned above, using MMM to measure the adstock effect comes with additional perks. By measuring the Adstock effect with an MMM, marketers can determine which advertising channels are most effective for their business. They can analyze the impact of different channels over time and adjust their advertising strategies accordingly. Moreover, measuring the Adstock effect can help marketers optimize their overall marketing mix. They can determine how advertising, pricing, product, and other factors interact to affect sales and adjust their strategies accordingly.

To introduce the adstock effect into MMM you can:

• Assume a certain industry standards for e.g. each media channel (let's say that TV should have a longer carryover, while SEA should have shorter carryover).
• Use hyperparameter optimization approach like Facebook Robyn does, so you don't include carryover as a parameter in the model but just run the model with different carryovers done as a pre-processing step and then pick the models based on e.g. goodness-of-fit / accuracy metrics. 
• Include the adstock parameters / variables in the model itself and estimate them properly from the data. The problem with this approach can be that you might end up having too many parameters in the model, making it e.g. prone to overfitting. This problem can be alleviated by introducing shared adstock parameters for media assumed to behave similarly.