Analytic Forensics Logo 






MMM (Marketing Mix Modeling) depends on the concept of “Ad Stock”.  (To access a discussion of how Marketing Mix Modeling works, click here MMM.)   Advertising is a catalyst; it changes a customer’s awareness, attitudes, and opinions, that over time change their behaviors and hopefully increases sales.   But the process is complex with many steps within the minds of the customer, so the effects are seen over time.  Ad stock is the distribution of these effects over time.

Ad Stock is just one of many non-linear effects in Marketing Mix Models.  (To access a discussion of some of the other effects, click here Non-Linear Effects.)

It is calculated (to over-simplify) but averaging the volume of sales after each ad placement (if possible with other factors and effects removed first).  You look for the patterns and usually “smooth the curve” to reduce noise.  Usually after you do this you get a graph like the one below.

Often a Marketing Mix Modeling project involves convincing the client that ad stock exists, and that it often builds rather than always starts at its highest point.  People naturally think that if they place an ad today they should see results tomorrow.  But that is just not how advertising works.  To prove this, I usually show them a project I did for a client (the graph below).  The client offered a 1-800 number in their ad so people could request more information.  Note, that anyone calling a 1-800 number is likely already 90% sold on buying the product.  It probably took them months of viewing ads just to get to the point of calling for more info.  But I took the names and addresses of the people calling the 1-800 number and matched them to names and addresses of people who eventually bought the product.   I was looking for the time between calling the 1-800 and the actual purchase.  As you can see in the graph below, sales built to a peak over the 11 months from the 1-800 call and then slowly declined.


The above graph is for a “long term considered purchase” product.  You would expect the consumers to take time to make the decision to buy.  But the same build of effect occurs even for impulse purchases like a candy bar.  If you see a candy bar ad, you don’t run out and immediately buy the candy.  More likely you won’t be in a store that sales candy bars (and be hungry, and…) for days.  But even if you are in a store within hours of seeing the ad, in that case more likely you had already decided what candy bar you were craving before you saw the current ad (but maybe the current ad will inspire the next candy bar you buy after that one).

So Ad Stock often involves some build to a peak and then decline of sales over time.  But not always.  Some media do have a more immediate effect.  The below graph is the ad stock for an ad that pushed an immediate discount on a product (with a time limit on the discount).  But it also had a long term delayed effect because it got the customers thinking about buying the product, which effected sales even after the discount was over.

But even “immediate effect” ad stock is more complex than most people think.  The below ad stock curve was for an in-store promotion.  It was very effective at selling the item immediately (especially since it caught customers who were in the store already), but also lead to “pantry loading”.  Customers bought so much of the product they didn’t have to buy it again for a while and that hurt future sales.  However, the total effect was positive, so the promotion did what it should have done.  But a statistically significant drop in sales afterwards also had to be put in the equation.

Most advertising (TV, Print, Digital banner ads, etc.) do follow a build to a peak and slow decay pattern (usually following a “gamma curve” distribution).  But it takes a lot of statistical effort to calculate the shape of these curves, their length over time and where they peak.  And they are different for every media and marketing factor and every product.  There are no “one-size-fits-all” ad stock curves.

And often ad stock curves can get very complex.  Competitive ads ad stock is particularly complex.  Competitive ads are a major factor in models for new products trying to make a mark against an established market leader.  In the case below, my client’s sales dropped immediately when their large competitors run a big ad campaign, but then would grow after that (as more customers entered the market for their product category based on the competitor’s ads), and in the long run would drop again (as the large competitor won new loyal repeat customers). 


Home Page


Analytic Forensics Logo




Text Box:                                                                                                                                                To Contact Us:







                                                                                                                      Phone:                                                     212-529-5337

                                                                                                                      Voice Mail:                                               917-838-7966


                                                                                                                      Address:                              23 East Tenth Street #304

                                                                                                                                                                New York City, NY     10003