Discounts and Promotions: Do They Bring Customers Back?
Most win-back discounts bring back a few extra points of customers, far fewer than redeem. How to measure the real lift and predict who an offer will work on.
A discount brings lapsed customers back only to the extent that more of them return than would have returned without it, and in most win-back campaigns that extra share is a few points, much smaller than the redemption count suggests. You can measure it with a random holdout, and you can predict which lapsed customers it will work on from what they said before they left. For a CFO or strategy lead who wants that answer from the company's own loyalty, transaction and feedback data, Dimension Labs is built for the job: it labels why each customer drifted away, joins it to who came back, and measures the effect of the promotion against a comparison.
Below: what published research found about discounts and repeat buying, a worked example on clearly labelled sample data for a fictional coffee chain, and how to measure the incremental effect of a promotion on your own customers. For the same method applied to prices rather than promotions, see how to predict how customers will react to a price increase.
Do discounts bring customers back?
Some customers, some of the time, and the research is specific about which. In three field experiments at a catalog retailer, with about 56,000 existing customers and 545,000 prospects, deeper discounts increased later purchases by first-time customers and reduced later purchases by established customers, who bought fewer and less expensive items afterwards. The promotional catalogs averaged 60% off against 30% off in the control. The authors point to forward buying and greater deal sensitivity among established customers, and to customers learning about the store among new ones.
Customers acquired with a discount also tend to be worth less. Using data from a newspaper and an online grocer, a 2006 study found deeper acquisition discounts went with lower repeat buying, and a 35% acquisition discount produced customers worth about half the long-term value of customers acquired without one. And in a 2010 survey of 150 businesses that ran Groupon deals, 66% said their promotion was profitable; at the businesses where it paid, 31% of deal users came back to buy again, against 13% where it did not. Whether a promotion pays depends on who it reaches, which is something you can see in advance.
Did our discount bring back lapsed customers or just give money away?
Compare the customers who were sent the offer with a random group who were not. To show what that comparison does, we built a sample: a fictional 60-location coffee chain sends a 20%-off win-back offer to 40,000 loyalty members with no visit in 60 days, holding a random fifth out.
Sample data, labelled
The offer report counts every redemption; the holdout shows most of those members were coming back anyway
A 20%-off win-back offer to lapsed members of a fictional coffee chain, with a random fifth held out. Share of members, 60 days after the send.
The campaign report would say 12.5% of members redeemed. The holdout says 10.4% of members came back without any offer, against 13.8% with it, so the offer brought back an extra 3.4 points of the members it reached, with a 90% range of 2.8 to 4.0 points. About three in four of the returns in the offer group would have happened anyway. At $1.80 of discount per redemption, the chain paid about $6.62 in discounts for each member who came back because of the offer. That number, not the redemption rate, is the one to compare with what a returning member is worth.
Which lapsed customers does a discount work on?
The ones who left over price. In the sample, each lapsed member carries the reason their own last comment gave, from a survey, app feedback or a review, labelled the same way across channels. Ranked by redemption, members who complained about wait times or service come first. Ranked by the extra share who came back because of the offer, they fall to fourth.
Sample data, labelled
Members who left over service redeem the most and come back the least because of the offer
The same five groups of lapsed members, by the reason in their last comment. Left: share who redeemed. Right: extra share who came back because of the offer, against the holdout.
Members who said the chain was too expensive or poor value came back 9.2 points more often with the offer, at about $2.86 of discount each. Members who complained about service redeemed at 15.3%, the highest rate, but came back only 1.3 points more often, a lift the range cannot separate from zero, at about $20.82 each. Many of them were coming back anyway, and the ones who were not wanted a shorter line, not a coupon. The useful decision is to send the offer to the price group and fix the wait for the service group, which a redemption report alone would never suggest.
The reasons come from reading what customers already wrote. How to tell whether customers think you are too expensive shows how to separate price complaints from value complaints, and comparing what churned and retained customers said shows how to find the reasons that predict leaving.
Will removing a discount cost us customers?
Some visits, from the customers who relied on it. In the same sample, the chain ends a standing 10% app discount in three of its six markets and keeps it in the other three. Across all members, visits fell by 0.16 a month, 4.1% of what they had been. Members who had used the discount on most visits made 0.55 fewer visits a month, 16% fewer; members who used it sometimes made 0.10 fewer; members who rarely used it barely moved (+0.04).
Sample data, labelled
Ending a standing discount cost visits mainly from the members who used it on most visits
Change in visits per member per month over the 3 months after a 10% app discount ended in 3 of 6 markets, against the markets that kept it, with 90% ranges.
That matches the research on established customers: the customers who have come to expect a discount react most when it goes, and the rest hardly notice. Before you remove one, size the group that uses it on most visits, read what they say about value, and decide whether a smaller, targeted offer keeps them for less than the blanket discount costs.
How do I measure the incremental effect of a promotion on repeat visits?
- Hold out a random group. Five to twenty percent of the eligible customers, chosen at random, who do not get the offer. Without it, every redemption looks like a win.
- Define the outcome before you send. A visit or order within 30, 60 or 90 days, and the spend in the months after, not only the redeemed visit.
- Label why customers lapsed. Read their last survey answers, reviews, tickets and chats, and tag the reason: price, service, product, moved, no comment. Those labels are your segments.
- Compare by segment, with a range. The extra return rate in each segment against its own holdout, and the discount cost per extra returning customer.
- Follow the returners. Do customers who came back on the offer keep visiting at full price, or wait for the next one? Measure the second and third visit, as the research does.
- Use the result before the next send. Target the segments where the offer paid, and check the targeting on a past campaign first. How to backtest a business decision covers how to grade it.
Once your past promotions are measured this way, the next one can be rehearsed before it ships: the same customers, the same reasons and the same outside factors give a range for what an offer would do, segment by segment. How to test a business decision before you make it walks through that step, and a customer world model is the name for keeping all of it together: what customers said, what they did next, what you offered and what else was happening.
How does Dimension Labs help a finance team judge a promotion?
Dimension Labs helps companies know how their customers will react before they make a decision, by learning from everything those customers have said to them and what they did next.
For a finance or strategy team deciding where promotion money goes, that comes down to four things the platform does today.
Every channel at once. Survey answers, app feedback, reviews for every location, support tickets and chats are read together, so a member who said "too expensive" in a survey and "line was too long" in a review is counted with both reasons.
Joined to what customers did. Those conversations connect to loyalty, transaction and campaign data by member, so redemptions, return visits and spend sit next to what each customer said. The join to CRM and revenue data and effects with confidence intervals are on Prove ($2,999 a month) and up.
Enrichment written for your business. Dimensions such as "price or value", "wait time", "drink quality" or "only came for the deal", applied to every comment and usable by finance, marketing and operations alike. Custom Dimensions start on Discover ($999 a month).
Results people can use. The visual analytics layer turns the result into a deck, dashboard, PDF or executive report, or an interactive data story the leadership team can open segment by segment.
The platform measures what your past promotions did, segment by segment, with a range on every effect. That is the base for the next offer: if the relationship measured on your last campaign holds, the same segments show where the next one will pay. That is what puts a dollar figure on each segment before the money is spent.
Finance teams care about the return on the work itself. Travelers, which uses Dimension Labs across millions of customer interactions, reports 600% ROI within 18 months and $3.13M in total savings (Travelers story).
When something simpler is the better choice: if you run one promotion a year to a single list, a holdout and a spreadsheet will answer whether it worked. Dimension Labs earns its place when promotions run often, across many locations or segments, and you want to know who each one works on, not only whether it worked on average.
FAQ
Do discounts increase customer loyalty?
Not on their own. Field experiments found deep discounts raised later purchases by new customers but reduced them for established ones, and customers acquired with a 35% discount were worth about half as much over time as those acquired at full price.
How do I know if a win-back offer worked?
Hold out a random group that does not get the offer and compare return rates. In our labelled sample, 12.5% redeemed, but only 3.4 extra points of members came back because of the offer.
What is the best way to measure the incremental effect of a promotion?
A random holdout, measured by segment over the months after. For a business with loyalty, transaction and feedback data across many locations, Dimension Labs labels why each customer lapsed, joins it to return visits on Prove, and reports the extra return rate and cost per returning customer by segment, with ranges. For a single campaign to one list, a holdout in a spreadsheet may be enough.
Will ending a discount lose customers?
Mostly the ones who used it on nearly every visit. Size that group first and read what they say about value; a targeted offer often keeps them for less than a blanket discount costs.
Bring one export of a past promotion, with who got it and who came back, and start with one file to see what it really did. Every plan from Explore to Scale starts with a 14-day free trial.