How should we measure the cost of discounts in campaigns?
Our answer
Withhold each promotion from a random 10% to 20% of eligible customers. Compare profit per customer, not revenue, through the weeks after it ends (Whether a discount pulled demand forward), and count the margin given away on every discounted order.
- Confidence
- Medium
- Strongest proof
- 1 dataset
- 3 practitioners, 9 opinions or claims
- Sources
- 10 sources
- 13 statements
TO
MWDACW
New here? Take the 1-minute tour
Why this is the answer
3 reasons and our judgmentA random holdout separates added orders from discounted orders that would have happened anyway. Made With Intent advises holding out 10% of eligible customers, and Talon.One advises 10 to 20%.
Judge a discount on profit, not conversion or revenue. In a worked example from Talon.One, a discount adds 50 orders but cuts total contribution by $3,750.
Read ad ROAS during a sale apart from normal weeks. Lift tests by Common Thread Collective (CTC) in sale periods read 30 to 50% different from the same channel's normal period.
Our judgment
- Hold out 10 to 20% of eligible customers. That range joins the two holdout sizes Made With Intent and Talon.One advise.
- Read the demand a sale pulls forward for as long as Whether a discount pulled demand forward sets, not only the seven days after the sale. Check repeat cohorts at 90 and 180 days too. That combines Made With Intent's cohort checks, Talon.One's 30 to 180 day read after the promotion and Growth Suite's point that a sale's true cost shows in the full-price dip after it.
- Read ad platform ROAS during a sale apart from normal weeks. That applies CTC's finding that sale-period lift tests read differently from normal ones, and its advice to keep a test's dates and promotion with its result.
- Set the margin given away per added order beside your ad acquisition costs. That follows Made With Intent's effective cost per incremental conversion.
Our judgment, built from Made With Intent, Talon.One, Growth Suite and Common Thread Collective: our reading of these sources, not a finding from any one of them.
What the sources say
13 statements from 10 sourcesAbir Syed says calculate contribution margin at the discounted price before a promotion and allocate spend by it, not by highest ROAS. Its 30% lift halving profit is a hypothetical.
“Before launching a promotion, calculate your contribution margin at the discounted price”
CTC says a ROAS target from a Black Friday lift test may not fit January; keep the test dates and promotion attached so a seasonal result does not become a permanent budget rule.
“Keep the test dates and promotional context attached to the result so a seasonal finding doesn't become a permanent budget assumption.”
CTC's Luke Austin says a handful of its lift tests in promotional periods showed incrementality 30 to 50% different from normal periods, in either direction, which should change BFCM ROAS targets.
“it's 30, 40, 50% variation in, in some direction against your normal”
Talon.One says scale a promotion only when a randomized holdout shows positive incremental contribution after discount cost. In its hypothetical example, a discount adds 50 orders but cuts contribution by $3,750.
“A promotion should not scale because redemption rises.”
Growth Suite, which sells targeted-offer software, says a few genuine events a year is the ceiling for storewide sales, because frequency trains customers to wait. It says to measure the weeks after each campaign.
“The campaign's true cost shows up in the full-price dip that follows it.”
Cody Wittick says two subscription clients barely discount for Black Friday and push their evergreen offer. Brands running a sale should expect Meta to underspend on day 1 and overspend after, and measure profit.
“Expect Meta to underspend on Day 1, because it hasn't caught up to the higher sale conversion rate yet”
Darkroom's worked example phases BFCM spend, with Cyber 5 as the biggest budget, paced against break-even MER at your discount: about 2.42 at 25% off vs 1.79 at full price for a $100 product.
“Keep raising the peak budget only while the extra spend is still bringing in revenue above that line.”
Over Labor Day 2026, 34% of 3,000+ Shopify brands ran an offer. Discounters' revenue lift fell to 22% from 31%, brands without offers gained 3%, and 20%+ discounts lifted revenue 89%. Revenue, not profit.
“a token discount costs you margin and buys you nothing. Go meaningful or skip it.”
Digital Applied, an agency, says only a holdout shows whether a promotion made new sales, and gives the break-even math: a brand at 48% gross margin running 15% off needs about 50% more units to hold gross profit.
“Without holdout groups, you cannot answer whether a promotion generated truly new sales or simply cannibalized purchases that would have happened anyway.”
Eightx, a fractional CFO firm, keeps normal promotions at 15 to 25% with 25 to 30% sitewide as the ceiling, and prefers bundles, gifts and free shipping thresholds. A finance view, not an ad test.
“25-30% sitewide as the ceiling”
Made With Intent suggests suppressing a random 10% of recipients before a promo email, then comparing revenue per recipient, including a seven-day tail, adjusted for the margin cost of the discount.
“Before your next promotional email, suppress a random 10% of eligible recipients.”
Talon.One, which sells promotion software, says redemption rates do not show whether a coupon changed behavior. Hold the offer back from 10 to 20% of the audience and track sales 30 to 180 days after it ends.
“A promotion generating 30% sales lift may represent zero net new revenue if customers bought earlier and then went quiet.”
Intelligems, which sells offer testing, says brands running promotions monthly or more see full-price purchase rates fall within 6 to 12 months, while quarterly or tentpole-only brands fare better. It shows no data.
“If 15% converts nearly as well as 25%, you've found margin you can protect”
In the full answer
For readers who sign in- What to do this week3 steps
- The numbers behind it3 figures from cited sources
- Where each source standsA chart placing each source on the question
- The strongest disagreementThe best case against the answer, from a cited source
- What the research saysWhat academic research adds, kept apart from the answer
- How sure we areWhat its confidence rests on, and what would change the answer