Retention / Referral
Referral Program Strategy for Ecommerce
// your happiest customers are your cheapest acquisition channel
By Fracto Solutions
October 31, 2026
The short answer
A referral program rewards existing customers for introducing new ones, turning word of mouth into a measurable, low-cost acquisition channel. It works because referred customers arrive with a trusted recommendation, so they convert and retain better. Use a two-sided reward (something for the referrer and the friend), ask happy customers at the right moments, and promote it where they’ll see it.
+16%
857 days
0.94% v 0.94%
Why do referral programs work?
Because a referral lets you acquire a better prospect, not because it transforms an ordinary one. That distinction comes from the research itself, and it changes how you should run the program. The best-documented effects are real, measurable, and narrower than the industry states them.
The evidence that matters here is two papers on one large German retail bank. Schmitt, Skiera and Van den Bulte followed 9,495 customers acquired in 2006 over a 33-month window, in a program paying referrers a EUR 25 voucher. Referred customers were worth about EUR 40 more on a six-year lifetime value basis, roughly 16%, and EUR 49 more (about 25%) inside the window. Their acquisition cost was about EUR 20 lower. Their churn hazard was about 18% lower, with 82.0% still active at 33 months against 79.2% of non-referred customers.
Two findings from that work are the ones a merchant would actually act on, and they are the two the industry leaves out. First, the margin advantage expires: it shrinks by 23.1 cents per 1,000 days, and the paper states plainly that it “disappears after 857 days,” roughly two and a half years. Second, retention is the half that lasts: “there is no such erosion for customer retention.” So the economic case is front-loaded on margin and durable on retention, which argues for a different payback model than a flat per-referral reward implies.
We should also retract something. An earlier version of this page said referred customers perform better “because they arrive with a trusted recommendation.” That causal clause was ours. The researchers describe their proposed mechanisms as “only possible explanations for the differences we document,” and their own conclusion is that the program’s value lies in letting firms selectively acquire more valuable prospects. The effect is largely selection. That is still a good reason to run a program; it is not the reason usually given.
On trust, the familiar figure is real but measures something narrower than performance: 84% of more than 29,000 respondents across 58 countries rated recommendations from friends and family the most trustworthy form of advertising in Nielsen’s global survey. Trust is not conversion, and we should not present it as though it were.
Is a referral program right for your store?
It needs satisfied customers, margin headroom, and a selective approach. The research is more specific than “any store with happy customers,” and the specificity is useful, because it tells you which customers to recruit as referrers.
Start with the caveat in Schmitt and colleagues’ own abstract: “the size of the value differential varies across customer segments; therefore, firms should use a selective approach for their referral programs.” A program aimed at everybody will pay for referrals of very different quality at one flat rate.
The follow-up study is more pointed still. Van den Bulte and colleagues looked at 1,799 referrer-to-referral dyads at the same bank and found that referrers with under one month of tenure produced referrals whose margin was below that of non-referred customers. Their managerial recommendation is to recruit referrers “among their customers who have been acquired at least six months ago.” If you prompt every new buyer to refer on day one, that is the group the research says will cost you.
The same paper found something that reframes the retention argument entirely: referred customers “exhibit lower churn only as long as their referrer has not churned.” Churn was about 40% lower while the referrer stayed and considerably higher once the referrer left. The retention benefit is attached to the relationship, not to the referred customer. Keeping your referrers is therefore part of keeping their referrals.
One useful targeting signal, with a caveat: Gershon and Jiang, working with a mobile cashback firm’s 41.2m-customer dataset, found referred customers refer substantially more themselves, sending invitations at 8.72% against 2.51%, with 19 of 20 companies in a separate dataset showing the pattern. The authors state the field comparisons are not causal and referral status is not randomized, so use it to decide whom to ask rather than as a proven multiplier.
Should the reward be two-sided?
Reward the friend at minimum. The two-sided reward is the industry default, and in the only field experiments that tested it directly it did not beat rewarding the friend alone. We had this the wrong way round and are correcting it.
Gershon, Cryder and John ran eight experiments including two with real companies. At GiftAMeal, 6,364 customers were emailed across control, sender-benefiting, recipient-benefiting, shared (two-sided) and donation conditions. Conversion was 0.94% for recipient-benefiting and 0.94% for shared, statistically indistinguishable at p = .99, against 0.39% for sender-benefiting and 0.08% for control. At Game Access, with 1,438 emails delivered, recipient-benefiting converted at 4.77% against 1.71% for sender-benefiting. The paper’s own summary is that the shared incentive “did not outperform the recipient-benefiting incentive.”
So the half that does the work is the friend’s. Rewarding only the referrer is the structure that measurably underperforms, and adding a referrer-side reward on top of a friend-side one has not been shown to add anything. That does not make two-sided wrong, and there is a reason to keep it: the experiments also found the ranking can reverse when sharing takes real effort, which suggests the referrer-side reward matters more as the ask gets harder. The full comparison is in how to structure referral rewards.
How should I size the reward?
Against the total value of a referred customer, not against acquisition cost alone. In the one case that has been measured, the reward exceeded the acquisition-cost saving and the program still paid, through margin and retention rather than through cheaper acquisition.
The arithmetic is worth doing in those terms. At the German bank the EUR 25 voucher sat against an acquisition-cost saving of about EUR 20, so on a pure cost-per-acquisition view the program lost money. It paid because the referred customer delivered roughly 25% more margin inside the window and churned about 18% less. If you size the reward only against your paid CAC, you will underpay for a customer whose value shows up later, or overpay for one whose margin edge expires in two and a half years.
Reward type
How it works
Watch for
Friend-side discount
The new customer gets the incentive
The structure with the best experimental record
Two-sided discount or credit
Both sides get something
Industry default; no measured advantage over friend-only
Store credit
Redeemed against a margin-bearing basket
Deferred revenue until redeemed
Cash or gift
Direct reward to the referrer
Costly, no repeat nudge, no measured advantage
Tiered or milestone
Bigger rewards for more referrals
Structurally the FTC's points-for-prizes example; disclosure applies
The third column is the one that changed. The friend’s side carries the measured effect; the referrer’s side carries the compliance burden.
When and where should I ask for referrals?
After a genuinely good experience, from customers with some tenure behind them, and ideally from customers who arrived by referral themselves. The timing advice is practitioner judgment; the targeting advice has research behind it.
We have no published test of referral-ask timing against satisfaction signals, so treat “ask at peak satisfaction” as our recommendation rather than a finding. What is supported is who to ask: customers with at least six months of tenure, per Van den Bulte and colleagues, and customers who were themselves referred, per Gershon and Jiang’s invitation-rate gap. Those two filters are cheap to implement and both come from real data.
On placement, cover the post-purchase flow, the account area, order confirmations and a findable on-site entry. The detail is in how to promote a referral program, which also covers the share mechanism, because that is where the legal exposure sits.
What do I have to disclose?
That the referrer is being paid. An incentivized referral is an endorsement with a material connection, and this is the section the referral-marketing literature almost universally omits, including an earlier version of this page.
Under the FTC’s revised Endorsement Guides at 16 CFR Part 255, effective July 26, 2023, an endorsement includes social posts and verbal statements a consumer would take as reflecting someone other than the sponsor. Section 255.5(a) treats payments, free or discounted products, early access and even a chance at a prize as material connections that must be disclosed clearly and conspicuously, and the Guides’ Example 11 is the affiliate and referral case specifically. Example 9 covers earning points exchangeable for prizes, which is what a tiered referral program is.
There is a second, sharper rule if you connect referrals to reviews. The FTC’s Rule on the Use of Consumer Reviews and Testimonials, 16 CFR Part 465, took effect on October 21, 2024 and carries civil penalties of up to $53,088 per violation. Section 465.4 prohibits incentives conditioned, expressly or by implication, on a review expressing a particular sentiment. An earlier version of this page recommended triggering referral asks after a five-star review; routing a reward to customers selected for having left a positive review is precisely the design that provision reaches.
What are the privacy rules on a referred friend?
The friend is a new contact, not your customer, so the soft opt-in does not cover them. Who sends the message matters as much as what it says, and the architecture most referral programs ship is the wrong one.
Under Article 13(1) of the ePrivacy Directive, marketing email needs prior consent. The soft opt-in at Article 13(2) only applies to contact details obtained from that person in the course of a sale to them, which is never true of a referred friend. The UK ICO is direct about the mechanism: you must comply if you send a marketing message “or if you ‘instigate’ someone else to send it,” responsibility attaches where “you must have encouraged them to send it,” and on asking customers for their friends’ contact details the guidance says “we would therefore advise against this type of viral marketing.”
If you do receive the friend’s details, you are the controller for that data and GDPR Article 14 applies: you must tell the friend “from which source the personal data originate,” which means naming the referrer, and do so within one month at the latest. The compliant architecture follows from all of this: supply the referrer with a link and let them send it from their own email or messaging app, do not have your system send to the friend, do not pre-fill the message body, and do not retain the friend’s contact details.
That last point is not theoretical. In Jensen v. Capital One (W.D. Wash., February 2025) a court held that Section 230 did not immunize a company whose refer-a-friend feature generated a pre-filled message, finding it “the sole author of the content of the text” and that it “substantially assisted its customers in formulating, composing, and sending commercial text messages.” An in-app notice asking the customer to confirm consent did not save it. The one-tap, pre-filled share that referral advice recommends is the design that got litigated.
How do I keep the program honest and measure it?
Reward on a completed first purchase, cap per account, and judge the program on incremental customers against reward cost. On fraud specifically, we can describe the problem but not quantify it, and we should say so.
We have looked for a published referral-fraud incidence rate and found none; everything available is vendor content with no methodology. The nearest primary signal is negative evidence worth citing as such: Gershon and Jiang’s 41.2m-customer analysis excludes suspected fraud and affiliate-referred customers without publishing the exclusion rate. Fraud is real enough to require exclusion in real referral data, and its magnitude is undocumented. Treat safeguards as cost control rather than as a quantified risk.
One quality control does have a published basis and belongs in the program rules: recruit referrers with at least six months of tenure, because newer referrers produced below-par referrals in the dyad study. That is a more useful safeguard than most anti-fraud checklists.
On measurement, the test is incremental new customers and their quality against reward spend, not raw sign-ups. And because the margin advantage in the one long study expired at roughly 857 days while the retention advantage did not, measure both separately rather than collapsing them into a single lifetime-value number. It is the same measure-and-improve discipline behind our growth-strategy pillar.
We turn your happy customers into a measurable, low-cost acquisition channel.
We build a two-sided referral program with margin-aware rewards, ask at the moments customers are happiest, make sharing effortless, add basic fraud safeguards, and measure genuinely incremental new customers, so word of mouth becomes a channel you can grow on purpose.
It’s the same measurement-first discipline behind our conversion optimization and lifecycle work, where we rebuild the buying path and measure the result .
// we would rather show up in your reporting than in your inbox
Frequently asked questions
Do referral programs actually work?
The best evidence says referred customers are genuinely more valuable, with specific magnitudes: about 16% higher six-year value, roughly 25% higher margin inside the study window, about 18% lower churn hazard, and about EUR 20 lower acquisition cost, across 9,495 customers at one German bank. Two caveats matter: the margin advantage is estimated to disappear after 857 days, and the authors attribute the effect largely to selecting better prospects rather than to the referral itself.
Should referral rewards be two-sided?
Reward the friend at minimum. In the only field experiments that tested the three structures directly, two-sided and friend-only performed the same (0.94% against 0.94%, p = .99, across 6,364 emailed customers), while referrer-only lagged at 0.39%. Two-sided is the industry default rather than the best-performing design, though the referrer side may matter more when sharing takes real effort.
Who should I ask to refer?
Customers with at least six months of tenure, and ideally customers who arrived by referral themselves. Referrers with under one month of tenure produced referrals whose margin was below that of non-referred customers, and referred customers send invitations at 8.72% against 2.51%. Both findings come from real data; the common advice to prompt every new buyer immediately does not.
Do I have to disclose that a referrer is paid?
Yes. An incentivized referral is an endorsement with a material connection under the FTC Endorsement Guides, 16 CFR Part 255, where discounts, free products and even a chance at a prize must be disclosed clearly and conspicuously. If you tie the reward to a review expressing a particular sentiment you also engage 16 CFR Part 465, effective October 2024, at up to $53,088 per violation.
Can I email the friend my customer refers?
Not on your customer’s consent. The friend is a new contact, so the ePrivacy soft opt-in does not apply and prior consent is required. The UK ICO holds you responsible where you instigated the message and advises against collecting friends’ contact details at all. The compliant design is to give the referrer a link they send themselves, without pre-filling the message or storing the friend’s details.
Sources
Schmitt, Skiera & Van den Bulte (2011), “Referral Programs and Customer Value,” Journal of Marketing 75(1):46-59. One large German retail bank, anonymized, paying referrers a EUR 25 voucher. 5,181 referred and 4,633 non-referred customers acquired in 2006, analyzed sample 9,495 after removing influence points; observation window January 2006 to September 2008 (33 months). Referred customers were worth about EUR 40 (16%) more on a six-year lifetime value basis and EUR 49 (about 25%) more within the window; acquisition cost was about EUR 20 lower; churn hazard was about 18% lower, with 82.0% of referred against 79.2% of non-referred still active at 33 months. The margin advantage shrinks by 23.1 cents per 1,000 days and the paper states that “the expected contribution margin advantage of a referred customer disappears after 857 days,” while on retention “there is no such erosion.” The abstract adds that “the size of the value differential varies across customer segments; therefore, firms should use a selective approach for their referral programs.” Wharton research on referral program strategy for ecommerce
Van den Bulte, Bayer, Skiera & Schmitt (2018), “How Customer Referral Programs Turn Social Capital into Economic Capital,” Journal of Marketing Research. Same bank, now at dyad level: 1,799 referrer-to-referral dyads (3,598 customers) against 3,663 non-referred. Referred customers churned about 30% less after controls (9.7% against 14.5% by the end of the window), but the paper finds they “exhibit lower churn only as long as their referrer has not churned.” Churn was about 40% lower while the referrer remained a customer and markedly higher once the referrer left. Referrers with under one month of tenure produced referrals whose margin was below that of non-referred customers, which is why the authors advise recruiting referrers “among their customers who have been acquired at least six months ago.” Peer-reviewed study (SAGE) on referral program strategy for ecommerce
Gershon, Cryder & John (2020), “Why Prosocial Referral Incentives Work: The Interplay of Reputational Benefits and Action Costs,” Journal of Marketing Research. Eight experiments including two field experiments. GiftAMeal: 6,364 customers emailed across control, sender-benefiting, recipient-benefiting, shared (two-sided) and donation conditions. Conversion was 0.94% for recipient-benefiting and 0.94% for shared (p = .99), against 0.39% sender-benefiting and 0.08% control. Game Access: 1,500 assigned, 1,438 emails delivered; conversion 4.77% recipient-benefiting against 1.71% sender-benefiting and 0.61% control. The paper states that the shared incentive “did not outperform the recipient-benefiting incentive.” Harvard Business School research on referral program strategy for ecommerce
Gershon & Jiang, “Referral Contagion: Downstream Benefits of Customer Referrals,” Journal of Marketing Research. Mobile cashback firm, 41.2m registered customers 2012 to 2022, with a 500,000-consumer regression sample, plus a 20-company dataset and a field experiment. Referred customers themselves referred substantially more: invitation rates of 8.72% against 2.51%, and 31% to 57% more referrals, with 19 of 20 companies showing the pattern. The authors state the field comparisons are not causal and that referral status is not randomized, so treat this as a targeting signal rather than a proven effect. Peer-reviewed study (SAGE) on referral program strategy for ecommerce
Nielsen Global Survey of Trust in Advertising. 84% of more than 29,000 respondents across 58 countries rated recommendations from friends and family the most trustworthy form of advertising; fieldwork 18 February to 8 March 2013. Nielsen’s later US work puts trust in recommendations from people you know at 89% on a sample of about 2,000. Note that both measure trust, not purchase outcome.
Shopify, “Referral Marketing Statistics.” Vendor guidance, cited only for the figures it carries and attributes, each of which lacks a stated sample: referral traffic converting at 2.5% to 3.5% against 0.5% to 1% for paid social (Elogic Commerce); a “healthy referral rate” of 1% to 3% (unattributed); and merchant anecdotes from the Shopify Masters podcast. It does not address friend-side rewards, referral fraud, FTC disclosure or program promotion. Shopify referral marketing statistics
FTC Guides Concerning the Use of Endorsements and Testimonials in Advertising, 16 CFR Part 255 (revised Guides, 88 FR 48102, effective July 26, 2023). §255.0(b) defines an endorsement to include social media posts and verbal statements consumers would take as reflecting someone other than the sponsor. §255.5(a) treats payments, free or discounted products, early access and a chance at a prize as material connections that must be disclosed clearly and conspicuously. Example 11 covers affiliate and referral compensation specifically, and Example 9 covers earning points exchangeable for prizes, which is structurally a tiered referral program. The Guides are administrative interpretations of FTC Act Section 5 rather than regulations. FTC Endorsement Guides
FTC Rule on the Use of Consumer Reviews and Testimonials, 16 CFR Part 465 (89 FR 68034, effective October 21, 2024). This is a rule, so violations carry civil penalties of up to $53,088 each under FTC Act §5(m)(1)(A) (16 CFR 1.98(d), for penalties assessed after January 17, 2025). §465.4 prohibits providing compensation or other incentives conditioned, expressly or by implication, on a review expressing a particular sentiment. §465.1 defines the compensation broadly enough to cover discounts, coupons and store credit. Relevant here because a program that routes a referral reward to customers selected for leaving a positive review moves from Part 255 into Part 465. FTC rule on consumer reviews and testimonials
ePrivacy Directive 2002/58/EC, Article 13. Article 13(1): electronic mail for direct marketing “may only be allowed in respect of subscribers who have given their prior consent.” Article 13(2)’s soft opt-in applies only where the contact details were obtained from that customer “in the context of the sale of a product or a service,” which is never true of a referred friend. EU ePrivacy Directive
UK ICO, electronic mail marketing guidance under PECR. You must comply “if you send a marketing message, or if you ‘instigate’ someone else to send it,” and the guidance is explicit that you are only responsible where “you must have encouraged them to send it.” The soft opt-in for previous customers “does not apply to prospective customers or new contacts,” which by construction is the referred friend. On asking customers for their friends’ contact details the ICO states: “we would therefore advise against this type of viral marketing.” UK ICO guide to email marketing (PECR)
GDPR Article 14 (information where data are not obtained from the data subject). When you receive a friend’s email or phone number from a referrer you are the controller for that data. Article 14(2)(f) requires you to tell the friend “from which source the personal data originate,” and Article 14(3)(a) requires it “within a reasonable period after obtaining the personal data, but at the latest within one month.” GDPR Article 14
Jensen v. Capital One Financial Corp., 2025 WL 606194 (W.D. Wash., February 25, 2025). A putative class action over a “Refer a Friend” text where the company generated the referral link and a pre-filled message and directed the customer to send it. The court held that Section 230 did not immunize the company, finding it “the sole author of the content of the text” and that it “substantially assisted its customers in formulating, composing, and sending commercial text messages.” An in-app notice asking the customer to confirm they had the recipient’s consent did not change the outcome. Cited because it is the exact pre-filled, one-tap share design that referral advice usually recommends.