benchmarks
What a Dollar of Discount Actually Returns
Across 38 stores, every $1 in prevention discounts returned $10.50 in added revenue, all-time. The monthly range, and what attribution doesn't prove.
ChatGPT
Perplexity
Gemini
ClaudeAll-time, across 38 stores, every $1 given away in prevention discounts is connected to $10.50 in added revenue. That's $1,357,503 in added revenue against $129,415 in discounts, across 12,950 attributed orders, measured 2026-08-25.
10.5x is a real number, but leading with it alone and stopping there would be the kind of cherry-pick that loses trust the moment someone checks a single month. So here's the fuller picture, including the part that makes the headline less impressive and more believable.
The range, not just the average
Month to month, the multiple has moved between 6.7x and 13.7x. That's a wide band, and it's the honest version of this number — the 10.5x all-time figure is what you get from averaging a metric that swings by a factor of two depending on the month.
| Measure | Value |
|---|---|
| All-time return | 10.5x |
| Monthly range | 6.7x – 13.7x |
| Stores | 38 |
| Attributed orders | 12,950 |
| Added revenue | $1,357,503 |
| Discounts given | $129,415 |
All figures measured 2026-08-25.
If a single month landed at the low end of that range, 6.7x is still a strong return. If it landed at the high end, 13.7x is not the number to expect every month going forward. Both are true at once, and both are part of the same dataset.
This is attribution, not measured lift
The most important sentence in this post is this one: these are orders that carried our discount code, which is not the same as orders that would not otherwise have happened.
Attribution answers a narrower question than it sounds like it does. It tells you: of the orders that used this discount, how much revenue did they add relative to what the discount cost? It does not tell you how many of those shoppers would have completed the purchase anyway, discount or not. Some of the 12,950 attributed orders are shoppers who were already going to buy. We don't have a way to separate that share out of this figure, and we're not going to imply we do.
Proving actual incremental lift — the number of orders that genuinely would not have happened without the intervention — requires a holdout: a group of shoppers who never see the offer at all, so you can compare what they do against what everyone else does. That's a different measurement, run separately, and it answers a different question than the one this post is answering.
Why we're publishing the attribution number anyway
Attribution is still useful. It tells a merchant, in plain terms, what a discount dollar is connected to — and $10.50 connected to every $1 given away, even acknowledging that not all of it is causal, is a meaningfully different starting point than assuming a discount is pure margin loss.
The honest framing is the useful one here: this is what the money looks like when you follow the discount code, not a claim about what would have happened in a world where the offer never appeared.