Because this poses such a challenge, retailers tend to rely on proxies for promotion performance, like:
- # people buying a promoted product
- % of promoted items in a basket
- online conversion rate,
- basket value.
While these can prove useful indicators, they don’t tell a promotion’s full story. They won’t help you decide when to stop, adjust, or continue running a promotion.
Lift, on the other hand, gives you much better insight. It really helps you decide what action to take next.
We have a company that sells one item: gadgets.
We buy them for €5. We sell them for €10. We have a €5 profit margin (to keep it simple, we have no operating costs).
We decide to run a promotion for one week. We reduce the price of the gadget by 10% to €9.
As a result we sell 105 gadgets for a profit of €4 * 105 = €420.
Was this promotion a success?
The answer depends on what we would have sold without running this promotion.
This is an unobservable quantity.
If we would have sold 100 gadgets that week anyway, then had we not run the promotion we would have made a profit of €5 * 100 = €500.
Let's call this Scenario A.
Running the promotion therefore had an opportunity cost of €80. This corresponds to a negative lift of €80 under the baseline. Ouch.
On the other hand, if we would only have sold 60 gadgets that week if we had not run the promotion, we would have experienced a profit of €5 * 60 = €300.
Let's call this Scenario B.
Running the promotion had an opportunity profit of €120, which corresponds to a lift of €120 over the baseline. Success.
Calculating lift requires knowledge of the expected sales during the promotion period that would have occurred without a promotion. This expected sales absent promotions quantity is the baseline sales.
Without an accurate baseline sales estimate, it's impossible to say whether or not a promotion is a success. Revenue alone won't tell you.
Sticking with Scenario B, our gadget promotion appears successful when comparing profit obtained that week from profit we expect to have obtained that week if we did not run the promotion.
I.e we sold 105 gadgets with €4 profit per item instead of 60 gadgets with €5 per item, for an increase in profit of €120.
But this does not consider that the additional 45 gadgets we sold at reduced profit may have been sold at full profit at another time.
Our customers may have decided to stockpile gadgets that they could obtain at the cheaper price so as to avoid purchasing them from us later at full price.
We therefore want to determine the size of this effect before deciding whether our promotion was successful.
We want to see the expected effect of the promotion not only at the time it was run, but at all times.
Still with Scenario B, let’s assume that without running a promotion when we did, say week 1, we would expect to have sold 60 gadgets the week after the promotion, week 2. These would have been sold with €5 profit, for a total profit of €300 in week 2.
(... If this sounds a little confusing, it's summarised in the visual below)
If after running the promotion we only sold 25 gadgets in week 2 at €5 because 35 gadgets were bought during the promotion at the discounted price, then in week 2 we experienced a profit of only €125. This is a reduction of €175.
Including this in our analysis of the opportunity cost of the promotion leads us to revise our conclusion that the promotion caused an increase of €120 in profit, to conclude that running the promotion reduced our profit overall, with an opportunity cost of €55. Ouch again.
On the other hand, if we sold 55 gadgets in week 2 because only 5 gadgets were sold ‘early’ during the promotion, our reduced profit in week 2 is €25, and the effect is that we conclude our promotion increased our overall profit by €95. Success again.
Calculating temporal effects like stockpiling requires knowledge of the expected sales outside the promotion period that would have occurred without a promotion. This expected sales absent promotions quantity is the baseline sales.
Now our company is a little more complicated. We sell 2 products, the standard gadget from the above examples, and a deluxe gadget which we sell at €20 per item for a profit of €10.
Sticking with Scenario B, but setting aside temporal shifts for moment, let’s return to the case where during our promotion of standard gadgets we sold 105 gadgets with €4 profit per item instead of 60 gadgets with €5 per item, for an increase in profit of €120.
This was an increase of 45 standard gadgets sold.
However, if 20 of these standard gadgets were bought by people who otherwise would have purchased the Deluxe-Gadget (but couldn’t resist that extra €10 of price difference), we have to factor this into our lift calculation.
This is the switching cost.
In this case we know that on those 20 items, we made €4 profit, but we lost the opportunity to make €10 profit by selling a Deluxe-Gadget at their regular price.
So on those 20 gadgets we lost €120 (20 * €4 - 20 * €10). The switching cost completely wiped out the profit increase we thought we had for the promotion during the promotion period.
On the other hand, if only 2 of these standard gadgets were bought by people who otherwise would have purchased the Deluxe-Gadget then we lost only €12 (2 * €4 - 2 * €10) in switching costs. Our promotion still resulted in an increase of profit of €108 during the promotion period.
Working out the number of deluxe gadgets that customers chose not to purchase relies on knowing the number that would have been purchased had the standard gadget not been discounted. Moreover, in looking at the expected switching effects of different items, it is not the number of standard gadgets that we sell that is the driver for switching.
We sold 105, but we would have sold 60 anyway. It is only those 45 additional standard gadgets we induced to be purchased by our discount that may have caused a customer to switch from our deluxe gadget product to the cheaper standard gadget. So in this too we will rely on a high quality estimate of the lift in, and therefore the baseline of, our standard gadgets.
Let’s complicate our company in one further way. Now we sell two products that are our standard gadgets from the first examples and a Gadget Cleaner, which is sold at €4 for a profit of €2.
There may be positive cross-purchase effects: 20% of people who buy our gadget decide to also buy a Gadget Cleaner In which case we can expect the 45 additional standard gadgets we managed to sell during the promotion to correspond to an additional 9 Gadget Cleaners sold, for an increase in the profit we obtained due to the promotion of €18.
We now revise the effect of the promotion from an increase in profit of €120 to an increase of €138.Example 4: Involving Hal
Example 4: Involving Halo
Promotions are complex. Measuring them is complex. Partly because of this retailers tend to rely on performance proxies that don't tell a promotion's full story.
One objective of this piece has been to show that this is a story worth telling in full. The value that can be derived from deeper insight into promotion performance is hard to overstate.
A second objective has been to make the case for prioritising one measure above all others: lift. Even the relatively simple examples we've used to make our case are quite complex. To simplify the analysis further by disregarding factors like switching, stockpiling and basket effects would be to not really measure the promotion at all. And in reality promotions are more complicated still. It's not only necessary to account for other variables, like promotion types (i.e 3-for-2 vs 30% off) or different media (tv, leaflet), it typically needs to be done for 100s or 1000s of different promotions every week. Some large retailers have the fire power to conduct this depth of analysis in house, but for most the potential promotions optimisation offers remains untapped.
To see how you can tap them book a demo below.
Sign up to receive stats, data, and insights that will help you keep up with retail trends