What is Adstock? – Data Science for CMOs episode 5
What is Adstock and what differs Adstock from time lag? What impact does it have on future campaigns? Christopher Engman explains what adstock is in less than 3 minutes.
Adstock, advertising stock or advertising carry over, are three words describing the same thing. They try to catch the delayed effects of marketing. Over a longer time period you’re building up your brand by various campaigns which sometimes has a slow decay. Being able to use the previous work that you’ve done when calculating a new period, a new month for example, you want to take the old campaigns into account and then you use an Adstock model for that.
There are two versions of ad stock, a simple Adstock model where we take for example 90% of the previous periods value and move it into the new period. That becomes the base for the new period. On top of which you add all your variables with their multipliers. Or there is a more advanced Adstock model, which is also called distributed lag model. Each variable has a different decay in reality. So when we say that 90% of the previous period is moved into the next period we’re simplifying it a bit. In reality, especially if we’re changing the marketing mix, we come into a situation where we need to take into consideration the change of the mix. I give you an example, so an SCM ad might have delayed effect, which is very short. Maybe only 10% of the previous periods effect of the SCM ad can be carried over into the new period. Whereas the Van Damme epic split campaign has 95% of the previous period carried over into the new period.
So each different type of marketing has in reality different time lags. And because of this, if we want to be really advanced we need to use the distributed lag model. If you make very small changes in your mix from time period to time period, like month to month, you can use the simple version. But if you’re making big changes in your marketing mix you actually need to use the distributive lag model, which is the advanced adstock model where you take the different lags for each type of marketing into consideration.