Referred Customers Churn 40% Less — Until Their Referrer Cancels. Then They Churn 280% More.
A landmark bank study found a referred customer's lower churn isn't really theirs. It belongs to whoever referred them, and it reverses when that person leaves.
Every referral program pitch makes the same promise: cheaper acquisition, and a customer who sticks around better than one you found through ads. The first half of that promise is easy to verify — you're not paying for clicks. The second half is where most SaaS teams are running on folklore instead of data, and the best data that exists on the question says something nobody puts in a referral-program deck: the retention advantage of a referred customer isn't really a property of that customer at all. It's on loan from the person who referred them, and the loan comes due the day that person cancels.
What the original study actually measured
The foundational result here comes from Philipp Schmitt, Bernd Skiera, and Christophe Van den Bulte's 2011 paper in the Journal of Marketing, "Referral Programs and Customer Value." The authors tracked 5,181 customers a leading German bank acquired through its referral program in 2006, against a random sample of 4,633 customers the same bank acquired through other channels in the same period — 33 months of panel data on contribution margin and retention for both groups.
The findings were cleaner than most marketing research gets: referred customers had higher contribution margins, higher retention, and were more valuable in both the short and long run. The combined lifetime-value premium was at least 16% over a matched non-referred customer, rising to 16–25% once the follow-up study refined the estimate. That's a real, replicated number — not a vendor's rounded-up anecdote.
What makes the paper worth reading past the headline number is the second half of it: the authors didn't just document that referred customers are more valuable, they traced the premium to two separate mechanisms that behave completely differently over time.
| Mechanism | What it drives | Behavior over the customer's lifetime | What happens if the referrer churns |
|---|---|---|---|
| Better matching | Contribution margin | Advantage fades as referred and non-referred customers come to know the firm equally well | No significant change — matching quality is baked in at acquisition, independent of the referrer |
| Social enrichment | Retention / churn | Advantage holds steady for as long as the referrer stays a customer | Advantage reverses — referred customer becomes a worse retention risk than a non-referred one |
Compiled from Schmitt, Skiera & Van den Bulte (2011) and Van den Bulte, Bayer, Skiera & Schmitt (2018), both Journal of Marketing / Journal of Marketing Research.
Margin and retention look like one combined "referral premium" from the outside, but they're produced by different things and don't move together. That distinction is the whole reason the second paper exists.
The follow-up that found the advantage was never the customer's
Four of the same researchers — Christophe Van den Bulte, Emanuel Bayer, Bernd Skiera, and Philipp Schmitt — went back to the same bank in 2018 with a sharper question in the Journal of Marketing Research: not just do referred customers retain better, but why, and does the reason hold up once you account for what happens to the person who made the referral. This time they tracked 1,799 referrer-referral dyads, with the data covering who referred whom, not just two anonymous pools of customers.
The margin half of the premium held up exactly as the matching theory predicted — the gap narrowed over time and had nothing to do with whether the referrer was still around. The retention half told a completely different story. While the referrer was still a customer, a referred customer's churn hazard ran about 40% below a non-referred customer's. The moment the referrer left the bank, that same referred customer's churn hazard jumped to about 280% above a non-referred customer's — worse than if they'd never been referred in the first place.
Source: Van den Bulte, Bayer, Skiera & Schmitt, Journal of Marketing Research (2018), hazard model on 1,799 referrer–referral dyads.
The authors call this mechanism social enrichment: a referred customer's bond to your company is partly borrowed from a third party — the referrer — who's connected to both sides. That bond provides real benefits while it exists, the same way a shared mutual friend makes you more likely to stay at a party you'd otherwise have left. Remove the friend, and the bond that was quietly doing retention work for you disappears with them, leaving a customer who in every other respect looks identical to one you acquired through a banner ad.
Why this should matter more for software than for a bank
The study's own authors flag this directly: referral programs should do the most retention work for products with network externalities or real switching friction — not a checking account, which works the same whether your friend uses it too. Their examples are pointed for anyone running a SaaS referral program: file-sharing services, two-sided marketplaces, multiplayer platforms — products where the thing you're paying for gets more useful because people you know also use it. A shared workspace tool, a collaboration platform, or anything with team seats fits that description more than a bank account ever could, which means the social-enrichment effect this research documents at banking scale is plausibly even stronger in a typical B2B SaaS referral program.
That's also exactly the mechanism behind champion turnover, which we've covered separately: an account's retention often runs through one specific person's relationship to your product, invisible to every usage metric right up until that person is gone. Champion turnover is the single-account version of this problem — one person leaving puts their own company's renewal at risk. A referral program is the same mechanism running across accounts instead of within one: when a customer who referred three other companies cancels, all three of those referred accounts just lost part of the reason they were staying, and none of your health-score dashboards will show it, because none of their own usage changed.
What this changes about how you run a referral program
The practical recommendations in the 2018 paper are specific enough to implement directly, not vague cautionary advice:
- Don't let day-one signups refer anyone. Referrals generated by customers with less than one month of tenure weren't just weaker than other referrals — they were less profitable than non-referred customers entirely. A brand-new customer doesn't yet have the "better matching" knowledge of your product that makes a referral worth having.
- Recruit referrers selectively, not universally. The paper's own recommendation to the bank it studied: focus referral invitations on customers who've been around at least six months, carry above-average margin, and show low churn risk themselves. A generic "invite anyone, get $20" program recruits plenty of referrers who are themselves on the way out the door.
- Track the referrer-referral chain, not just a referral count. If your referral data only tells you a customer came through the program, and not who specifically sent them, you have no way to flag downstream accounts when a referrer cancels. The chain is the asset — the raw signup count is not.
- Treat a referrer's cancellation as a flag on every account they referred. This is the direct operational translation of the hazard-rate finding: the moment a referrer churns, every account they brought in should move into the same kind of proactive-outreach queue a departing champion would trigger, before a renewal date forces the conversation.
None of this means referral programs are a bad acquisition channel — the 16–25% lifetime-value premium is real and well-documented. It means the premium isn't a fixed trait stamped onto a customer at signup. It's a relationship between two accounts that your systems almost certainly aren't modeling as one, which turns an apparent strength of your program into an unflagged liability the day the wrong person cancels.
If you're trying to put a number on what that's actually worth protecting, our LTV calculator is a fast way to see what a referred cohort's value looks like under the "referrer still active" assumption versus the "referrer has churned" one — the gap between those two numbers is exactly what a referrer-tracking system is designed to catch before it happens. And if a referred customer does end up canceling, the reason is worth capturing specifically, not folded into a generic "not using it" bucket: our piece on why customers actually cancel covers how granular cancel-reason data turns a vague churn number into a list of fixable mechanisms, and "the person who got me to sign up isn't here anymore" is exactly the kind of reason a cancellation flow can surface that a health score never will.
Frequently asked questions
Do customers who join through a referral program actually churn less?+
Yes, on average — but the advantage is smaller and more conditional than most referral-program pitches imply. A 2011 Journal of Marketing study tracking 5,181 referred and 4,633 non-referred customers of a German retail bank for 33 months found referred customers had higher contribution margins and higher retention, combining into a lifetime value 16–25% higher than matched non-referred customers. A 2018 follow-up in the same journal found that the retention half of that advantage depends entirely on the referrer remaining a customer themselves.
What happens to a referred customer's churn risk when their referrer cancels?+
It gets worse than if they had never been referred at all. The 2018 study (Van den Bulte, Bayer, Skiera & Schmitt, Journal of Marketing Research) found referred customers whose referrer was still active had a churn hazard about 40% lower than non-referred customers. Once the referrer had left, that same referred customer's churn hazard was about 280% higher than a non-referred customer's. The direction of the effect flips entirely.
Should a SaaS company restrict who is allowed to refer new customers?+
The research says yes. The study's authors recommend recruiting referrers who have been customers for at least six months, show high margins, and are themselves unlikely to churn. Referrals from customers with less than one month of tenure were found to be less profitable than customers acquired through any other channel — not just less valuable than other referrals, less valuable than doing nothing.
Is a referral program's value mostly from cheaper acquisition or better retention?+
Both, but through two separate and independent mechanisms, which is the central finding of this research. 'Better matching' — referrers knowing which of their contacts actually fit your product — drives the margin advantage, and it fades over the customer's lifetime regardless of what the referrer does. 'Social enrichment' — the referrer acting as an ongoing social tie to your company — drives the retention advantage, and it depends entirely on the referrer staying around.
Stop losing subscribers today
One script tag. One function call. A live cancellation flow in under 10 minutes.
Start free trial →