Post Promotion Dip

The Post Promotion Dip: Does it Exist?

Professor John M. McCann
Fuqua School of Business
Duke University

March 21, 1995

The subject involves the pattern over time of sales of a brand during and after a retail promotion. The question involves the post-promotion dip ... the dip in sales following the promotion; does it exist? Consider a brand in a competitive consumer package good market such as toothpaste, sodas, etc. When you look at factory shipment or warehouse withdrawal data, you see patterns like the following:

This chart shows an increase in shipments during a two-week trade promotion (weeks 6 and 7), following by a dip in shipment for the next three weeks. This is a typical pattern that is caused by retailers buying for inventory, and then working out of the inventory for a period of time. This pattern has existed for a long time, and has become more pronounced as trade promotions have become more prevalent and retailers have learned how to optimize their buying.

Brand managers have long believed that a similar pattern holds for retail sales, for sales to the consumer. They formed this belief during the days of Nielsen store-audit data. Although the bi-monthly nature of the data masked the promotion effect, they believed it happened based upon common sense, shipment data, and a few marketing models. Researchers such as Shugan & Jeuland have built consumer purchase models based upon a pantry inventory assumptions: they assume that consumers inherently use an EOQ model to determine how much of an item to buy when it is on deal. The resulting models would lead to a post-promotion dip.

One of the surprises from the weekly scanner data is that the post-promotion dip is hard to find when you examine weekly store sales. You see data such as the graph on the left. The general pattern involves a bump in sales of 2 to 8 times normal sales, and then a return to the pre-promotion level. There is no dip in consumer sales.

There are several ways to explain this pattern, and I prefer the following segmentation argument. Some consumers are brand loyal; they are going to buy the brand on a regular schedule and are not influenced by the promotion to any noticeable extent. These consumers did not cause the bump in the brand's sales, and they will not cause the bump in the sales of competing brands. Each brand has its set of such core buyers. In the above graph, these consumers account for the 10 or so units sold each week.

There is a segment that is brand loyal, but who never pay regular price for the brand. These consumers have been trained by manufacturers and retailers to expect a promotion on their brand ever so often. When one does occur, they purchase enough to hold them until the next promotion. We could call this the "rational expectations" segment, or the "brand loyal/inventory buying" segment. These consumer cause part of the bump in sales during the promotion, but they are not part of the sales between promotions.

There might be a related set of consumers who move between types of retail trade. They may normally buy the brand in mass merchandise or club stores. Their purchases will only appear in supermarket data when a chain offers a deep discount on the item.

There is another segment of consumers who can be called the "evoked set" buyers. Each consumer has a set of brands that s/he will select among, depending upon who is on promotion. These consumers have also been trained to expect a promotion on a regular basis, and they make their inventory decisions based upon their expectations. They are part of the bump, but not part of the normal sales.

We can identify yet another group as the "deal consumption" group. These people do not consume the category on a regular basis, but will buy it for immediate consumption when sufficiently motivated by a deal. Again, they are part of the bump, but not the base.

Then we have the mythical segment; the one that does not exist; the one that would account for a post-promotion dip. These consumers are brand loyal, and they are motivated by a deal to buy more than the normal amount of the brand. They do not increase their consumption based upon having more in their pantry. They just delay purchasing the brand until they use their stock. If this was a significant segment, we would see a dip. When we do not see the dip, we can infer that this segment does not exist. Since it is rare to find the dip, I have come to believe this segment is quite small.

Does this mean that we never see the dip? No, that is not the case. But we can identify conditions where we might see it.

  1. Infrequent promotions: consumers would have to purchase too much to buy deal-to-deal.
  2. Unpredictable promotions: consumers could not anticipate the timing of the deals and would thus not buy deal-to-deal.
  3. Very large market share: the brand has such a large share that there cannot be an evoked set ... there is no viable alternative.
  4. Very deep discounts: loyal consumers are motivated to buy for inventory and work out of that stock.
  5. Pantry stock is not related to consumption.

Some people report the existance of post-promotion dips in scanner data. This may be the case, but I think it is a relatively rare phenomenon because the conditions listed above do not occur often.

The noisy nature of real data make it difficult to determine the existance of a dip. This chart plots weekly retail sales data for a brand of coffee creamer in a single market. Detecting post-promotion dips is difficult because it is hard to find a promotion that has steady state conditions before and after the event. In these instances, one must resort to econometric models to take out trends and the effects of known events, as well as model the dip.