Recently, we received a question from a reader who was puzzled because he was seeing negative numbers for subsidized volume. If you’ve also wondered about this, take a look at his question and our answer.
If you don’t know what we’re talking about, you might be interested in this post all about Subsidized Volume. In a nutshell, subsidized volume happens because, when a product is promoted, there are always some shoppers who take advantage of the promotion but would have bought the product anyway. You are most likely to see negative subsidized volume (if you ever do) at a very low level of detail – one week in one retailer in one market on one brand or item. When periods, geographies, and/or products are aggregated you’ll rarely, if ever, see negative subsidized volume.
Hello again Sally,
I’d like to share some thoughts about negative subsidized sales. Based on the title of the article, I thought this would be the most appropriate place to post.
Thanks to other articles on this site, I completely understand how incremental sales can be negative since it is ultimately derived using an estimated value (i.e., base sales). However, after much thought and deliberation, I do not believe it makes sense to speak of or use negative subsidized sales in one’s analysis. Allow me to explain…
Based strictly on arithmetic, using the formula:
Subsidized Sales = Promoted Sales – Incremental Sales
… it is trivial to see how subsidized sales could result in a negative value whenever incremental sales is greater than promoted sales. But what does that actually mean, if anything? How do you extract actionable insights from this?
By definition, subsidized sales represents the portion of “promoted sales that would have occurred anyway, even if no promotion had been present“. As such, to satisfy this definition, subsidized sales must be some value between 0 and promoted sales. In other words, semantically speaking, either (1) all of your promoted sales were incremental (making 0% subsidized), (2) none of your promoted sales were incremental (making 100% subsidized), or (3) some portion of your promoted sales were incremental and some subsidized.
In my opinion, it doesn’t make sense to say you had negative subsidized sales. Nor does it make sense to say that you had more subsidized sales than promoted sales. Subsidized sales should always be zero or a portion of promoted sales (promoted sales is always a non-negative value).
Given this argument, I believe the calculation for subsidized sales should include some conditional logic, as follows:
If Incremental Sales > Promoted Sales, then Subsidized Sales = 0 (0% Subsidized)
If Incremental Sales < 0, then Subsidized Sales = Promoted Sales (100% Subsidized)
Otherwise, Subsidized Sales = Promoted Sales – Incremental Sales
Let’s consider an example.
Suppose I have a product that has sold $200 in actual sales of which $100 is promoted sales.
Case 1: Suppose base sales were $50, resulting in $150 of incremental sales.
Case 2: Suppose base sales were $250, resulting in -$50 of incremental sales.
Case 3: Suppose base sales were $150, resulting in $50 of incremental sales.
Using the formula by itself, subsidized sales for these 3 cases would be calculated as follows:
Case 1: $100 – $150 = -$50 (but what does this mean?)
Case 2: $100 – (-$50) = $150 (but what does this mean?)
Case 3: $100 – $50 = $50 (okay, this is useful, 50% of promoted sales were subsidized)
Using the formula (along with the conditional logic), subsidized sales for these 3 cases would be calculated as follows:
Case 1: $150 (incremental) > $100 (promoted), therefore subsidized sales = 0 (0% of promoted sales were subsidized)
Case 2: -$50 (incremental) < 0, therefore subsidized sales = $100 (100% of promoted sales were subsidized)
Case 3: $100 – $50 = $50 (50% of promoted sales were subsidized)
To me, this seems like a better way to think of subsidized sales.
Having said that, I am relatively new to the CPG data analytics field, so perhaps I’m completely overlooking or overthinking this one.
Nestor,
I absolutely agree that the concept of negative subsidized sales makes no business sense.
I would have two concerns about your conditional formula though:
1) Practically speaking, incremental sales + subsidized sales will not longer equal promoted sales. So you would want to think about the consequences of that in terms of communication – will it cause confusion down the road?
2) Negative subsidized sales are a flag that something odd is going on in your data. So it’s probably better to dig in and sort that out then just mechanistically fix that one measure. If subsidized sales doesn’t make sense, then the relationship between at least some of the other sales measures involved (total, promoted, base, incremental) must also be off in some way and you’ve only addressed subsidized sales.
Just like negative incremental sales, if negative subsidized sales is small relative to the aggregate positive values for that measure, just don’t worry about it. It won’t impact your overall conclusions. Consider it noise.
You may have seen my recent post titled “CPG Data 911”. Some of those suggestions would also apply when you have negative subsidized volume.