Designing a Referral Program: How to Set the Reward, and When Not to Run One at All
The most common failure mode for a referral program is not that nobody uses it. It is that everybody who uses it is a loyal customer who was going to buy anyway — they take the discount, recommend you to friends who already knew about you, the referral count looks great on paper, and all you have actually done is shave your own margin.
Avoiding that outcome is a matter of sequence: work out where the program sits in your overall word-of-mouth structure before you decide how much the reward should be.
A referral program is an amplifier, not a substitute for word of mouth
Referral programs and word-of-mouth marketing address different links in the same chain.
Word of mouth addresses whether a stranger verifying you finds credible information. They do not know you yet, and what they rely on is genuine usage content on forums, review sections, and social platforms. For the full architecture of that layer, see The Complete Guide to Word-of-Mouth Marketing.
A referral program addresses whether an already-satisfied customer is willing to speak up. Its precondition is that somebody is already satisfied. If the product experience itself is not there, a referral program only accelerates the spread of bad news, because a referred customer arrives carrying a friend’s endorsement, with higher expectations and correspondingly sharper disappointment.
The practical test: if your existing customers rarely repurchase and your support complaints cluster around one recurring problem, the priority is fixing the experience, not designing rewards. A referral program is an amplifier — it amplifies what is good and what is not, equally.
Two-sided versus one-sided rewards
This is the first decision in designing a referral program, and the psychology of the two approaches is entirely different.
One-sided (referrer only). The referrer gets something, the referred customer gets nothing. The problem with this design is that it makes sharing slightly awkward to bring up — the referrer is effectively converting a friend’s spending into a personal benefit. Unless the reward carries a charitable or community dimension (converted to a donation, say, or exchanged for status within a community), willingness to share tends to be low.
Two-sided (both parties). The referrer has a reason to bring it up, and the referred customer has a reason to try. This is the default choice for most consumer categories. The two incentives should be different in kind:
| Recipient | What the incentive is for | Suitable formats |
|---|---|---|
| Referrer | Providing a reason to speak up, acknowledging the share | Cash back, points, credit toward a future purchase |
| Referred customer | Lowering the barrier to a first purchase | First-order discount, trial pricing, a bonus sample |
Referred customer only. Rarer in practice, but reasonable where the product already has strong word of mouth and customers recommend it unprompted. In that case the problem to solve is not willingness to share but the new customer’s threshold for trying.
Four principles for incentive design
One: tie it to a qualified transaction, not a sign-up. The reward should trigger when the referred customer completes a first qualified transaction and the return window has passed — not on registration, download, or click. Tie it to a top-of-funnel metric and invalid accounts will eat the budget.
Two: make it calculable and legible. “Refer a friend and receive NT$200 in store credit once they complete their first purchase” works far better than “3% to 8% back depending on spending tier.” Rules that require mental arithmetic to evaluate get abandoned by most people.
Three: set a cap, but do not set it low. No cap at all invites arbitrage; a cap that is too low (two referrals a month, say) shuts out your most influential referrers. The sensible setting is a ceiling most normal users will never reach but which still stops anomalous accounts.
Four: make redemption frictionless. A reward that takes three months to arrive, requires a spending threshold to use, or comes wrapped in restrictions kills the second referral. The compounding in a referral program comes from one person referring repeatedly, not from one person referring once.
Choosing the reward format. Store credit and points return spending to your own channels and cost you less than face value, but appeal only so far to low-frequency buyers. Cash back is the most attractive and the most expensive, suiting categories with high order values or margin structures that can carry it. Physical gifts sit between the two, with the caveat that a poor-quality gift can damage how the referrer sees the brand. In most cases, test with store credit first and decide whether to increase the offer based on participation.
How to think about the cost. Work the ceiling back from your own customer acquisition cost: as long as the referrers and the friends rewards together come in clearly below what you pay to win a customer through advertising, the program holds up on paper. The right figure varies enormously by category, order value, and repurchase structure, so calculate it from your own data rather than copying a competitor. For a comparison of the cost structures of word of mouth and advertising, see Word-of-Mouth vs. Paid Advertising.
Keep the sharing path short: where the friction hides
Referral programs usually fail not because the reward is too small but because at the moment someone wants to refer, they cannot.
Between the impulse to refer and a completed share, every additional step sheds a portion of people. The usual friction points:
- Having to log in before you can get a referral link
- The referral code buried three levels deep in the account section
- A shared message with no pre-filled text, leaving the customer to work out what to say
- A link that opens the homepage rather than the product being recommended
- A referred customer having to type a long code by hand to get the discount
The smoother design: at the moment a customer completes a purchase or receives the product, a “share this with a friend” entry point appears on screen or in a message; one tap generates a link with pre-filled text; the friend opens it and lands directly on the product page with the discount already applied.
Timing matters more than placement. The best moment to invite a referral is at a satisfaction peak — just after the product arrives, just after a service goes smoothly, just after someone leaves a high rating. Putting the invitation in those moments works far better than parking it in the account section and waiting for people to find it.
If your customers mainly communicate with you through LINE, a referral entry point in the message flow of your Official Account usually converts better than one on the website. For the mechanics, see The Complete Guide to LINE Community Word-of-Mouth Marketing.
Abuse prevention: work out how it will be gamed
Wherever there is a financial incentive, someone will look for the hole. The common patterns and their countermeasures:
| Abuse pattern | Countermeasure |
|---|---|
| Referring yourself through multiple accounts | Deduplicate by phone number or payment method, restrict by device and address |
| Returning immediately after referral, reward already paid | Delay the payout until the return window closes |
| Mass-posting referral codes into public discount communities | Prohibit public distribution in the terms, monitor anomalous traffic sources |
| Accounts that farm referral codes full time | Set a per-period cap, manually review accounts growing abnormally |
Write the terms in advance, put them somewhere customers can see, and reserve the right to cancel rewards in cases of abuse. Changing the rules after the fact and clawing back rewards already issued usually costs far more in public relations than it saves.
One more thing to watch: do not mix your referral program with reviews. Rewarding people for leaving reviews, or asking referrers to post positive reviews on public platforms, breaks most platform rules and runs into Taiwan’s rules against misleading advertising at the same time. Referral rewards should apply strictly to introducing new customers, never to publishing reviews. Keep that line clearly drawn. For the compliance boundary, see Word-of-Mouth Marketing Compliance in Taiwan.
Which metrics to track
Looking only at the count of successful referrals will badly overstate performance. At minimum, track these:
- Referral participation rate: what proportion of customers have ever generated a referral link (measures whether the entry point is findable and the timing is right)
- Share conversion rate: of the links generated, how many get opened and how many convert (measures the pre-filled copy and the landing page)
- Retention and repurchase among referred customers: this is where the real value of a referral program sits. If retention is no better than for ordinary customers, what you have attracted is discount hunters
- Actual cost per qualified new customer: both sides of the reward summed and divided by qualified new customers, then compared against advertising cost
- Incremental, not total: how many of these people would not have come without the program, which you can estimate from the change in new-customer mix before and after launch
For quantitative methods and tracking setup, see How to Measure Word-of-Mouth ROI.
When not to run a referral program
Referral programs do not suit every brand. Hold off in these situations:
When product satisfaction is still unstable. Low repurchase, clustered complaints — fix the product first.
Categories with very low purchase frequency. For things bought once or twice in a lifetime (large durable goods, for instance), customers have no second purchase to apply a reward to, the reward format is hard to design, and nothing accumulates.
When customers do not know each other. Referral presupposes a social connection. If your customers share no common circles, referral simply will not happen, and what you should be doing instead is public word-of-mouth content rather than private referral.
Categories involving sensitive privacy. Medical, aesthetic, financial, and legal services — customers may not want anyone knowing they used you, and pushing a referral program creates resentment.
When you have no baseline word of mouth yet. A referred customer who receives a link will mostly still go and check you on Google or the forums. If what they find is a blank page, the referral gets interrupted. In that situation the word-of-mouth base has to come first; see The Complete KOC Seeding Playbook.
A well-designed referral program is, at bottom, a way of making something customers were already inclined to do slightly easier, with a face-saving reason to bring it up. It does not create satisfaction out of nothing, but it does let existing satisfaction travel further.
If you want to check your own referral mechanics for structural holes, or fold a referral program into a coherent strategy alongside public word of mouth, talk to a NETVANA consultant.
Further reading: for quantifying referral performance, see How to Measure Word-of-Mouth ROI. For working your owned customer base, see The Complete Guide to LINE Community Word-of-Mouth Marketing. For systematic everyday-creator word of mouth, see The Complete KOC Seeding Playbook. And for accumulating reviews and repurchase in an e-commerce context, see The Complete Guide to E-Commerce Word-of-Mouth Marketing. For how commission-based promotion differs from customer referrals, see What Affiliate Marketing Is; and for what lasts longer than a referral reward, see How to Run a Brand Ambassador Program; and for the owned list your referral invites depend on, see Connecting Email Marketing to Word of Mouth. For finding who to invite into a referral program, see What Net Promoter Score Actually Measures.
