Customer Referral Programs

Your customers already recommend you. Pay them for it.

A customer referral program rewards existing customers for bringing new ones: the customer shares a personal link or code, the new customer buys, and both sides get a reward, account credit, a discount, or cash. It differs from an affiliate program in who does the promoting: your own customers, rather than publishers and creators.

14 min readReviewed

Quick answer: what makes a referral program work?

Reward both sides. One-sided programs ask the sharer to do a friend a favour on your behalf. Match the reward to the product: credit for subscriptions, a discount for stores, cash only when neither is worth anything to the referrer. Ask at the moment of satisfaction, right after the product worked, not in a monthly newsletter.

Give them a code as well as a link: codes get spoken aloud and typed at checkout, links do not. And track it properly, or you will pay for sales that were coming anyway.

The one distinction

Referral, affiliate, or ambassador?

Who promotes decides everything else: how you recruit, what you pay, what breaks. Most companies end up running both.

Referral, affiliate and ambassador programs compared
Customer referralThis guideAffiliateInfluencer / ambassador
Who promotesExisting customersPublishers, review sites, tool directoriesCreators with an audience
Why they promoteThey like the product and want the rewardIt is their business modelMoney, product, or affinity
Typical rewardAccount credit, a discount, a free monthA percentage of the sale, often recurringA flat fee, a commission, or both
Volume per personOne to five referrals, everContinuousBursty, tied to what they post
How you recruit themA widget or an email, they are already customersYou recruit them, or they find your listingYou search for them and invite them
What breaks itA reward nobody wantsAttribution nobody trustsNo assets and no brief

The rows people get wrong are the last two: volume per person, and what breaks it. A referral program measured like an affiliate program always looks like it is failing.

Decision one

The reward decides everything else

Five structures. Pick by what your product makes cheap, then build it and watch the cost.

Give $20, get $20

Two-sided credit

Best for subscriptions and anything with a recurring bill · costs margin, not cash

A non-cash commission plan issuing Stripe credits, Shopify store credits, WooCommerce credits or a custom unit you define, with a cap on how many times one customer can earn.

20% off for them, 20% off your next order

Two-sided discount

Best for stores and one-off purchases · costs margin on two orders

A per-affiliate coupon minted automatically in Shopify, Stripe or WooCommerce, with the discount type, amount and duration set on the plan, and a commission rule that only matches above your minimum order amount.

$50 per referred customer who stays 30 days

Cash to the referrer

Best for products the referrer does not want more of · costs real money, plus payout and tax overhead

A cash commission plan with a holding period so the reward is not payable until the refund window closes, payout terms from NET 0 to NET 60, and a minimum threshold.

500 credits per referral

A unit you invent

Best for products with an internal currency · costs whatever the unit costs you to serve

A custom credit unit with its own name, calculated flat per purchase or per unit of currency spent, and capped by number of payments, by months, or by total times earned.

A bonus at the fifth referral

A milestone on top

Best for turning one-time sharers into repeat ones · costs a little, at the point it is most deserved

Affiliate milestones move a customer onto a better plan automatically at a threshold you set, and performance bonuses pay a one-off amount on top.

Non-cash rewards, custom credit units and milestones start on the Business plan. Coupon tracking is on every plan.

Build the reward

Both sides, the guard rails, and what one referral actually costs.

Structure

When does the reward trigger?

Paying on purchase, held until the refund window closes, is the default for good reason: the reward and the revenue arrive in the right order.

What one referral costs

Referrer's side
$20
Friend's side, on a $40 order
$8
Total per referral
$28
Maximum one customer can cost you
$280

Credit and discounts cost margin rather than cash, so the true figure is this number multiplied by your cost of goods, not by its face value.

Arithmetic on the numbers you set. Not a benchmark, no sourced participation or conversion figure for referral programmes is published on this page.

Decision two

Ask at the moment of satisfaction

Six placements, in the order they convert. The placement matters more than the copy.

Immediately after an order

The post-purchase confirmation

The highest-intent moment you will ever get, and the one most commonly wasted on a bare receipt. The customer has just decided you were worth paying for; the ask costs them nothing and confirms the decision.

The moment the product works

In-app, after a success event

A report finished, a milestone hit, a workflow completed. Trigger on the event rather than on a schedule, the same prompt shown on a Tuesday morning converts a fraction as well as the one shown ninety seconds after something worked.

When the thing physically arrives

The order-delivered email

For stores, this beats the confirmation email: the customer now has the product in their hands rather than a promise of it. It is also the moment they are most likely to be showing it to somebody.

When they choose you again

The renewal or invoice email

A renewal is a second purchase decision, and a quiet one. Attaching the ask to it catches customers who never open marketing email but always open billing email.

Whenever they look for it

The account or rewards page

The permanent home. It converts nobody on its own and it is where every other placement sends people, so the link, the code and the QR all need to live here rather than only in the email that prompted the share.

As they leave, not as they arrive

An exit-intent prompt on high-intent pages

The most easily overdone placement on the list. Confined to pages where someone was clearly considering something, pricing, an account page, it works. Fired on every page load it is an interruption that costs more goodwill than it earns referrals.

The referral widget and popups and banners start on the Business plan.

Decision three

Four numbers tell you if it works

Every formula written out. The fourth cannot be measured directly, and pretending otherwise is how referral programs become discounts.

1. Participation rate

participation_rate = customers who shared ÷ customers asked

The first thing to fix and the last thing anyone measures. A low participation rate is almost always the reward or the placement, not the copy, and it is the only metric here that isolates your ask from your product.

2. Referral conversion

referral_conversion = referred signups ÷ referral clicks

How well the traffic a customer sends actually converts. It should beat your site-wide rate comfortably, because it arrives with a personal recommendation attached. If it does not, the landing experience is discarding the recommendation.

3. Cost per acquired customer

referral_cac = total rewards paid ÷ referred customers acquired

Include both sides of a two-sided reward, and include the cost of the credit at what it costs you to serve rather than at face value. Compare it against your blended acquisition cost, that comparison is the whole business case.

4. Incrementality

incrementality = the share you would have won anyway

The number that decides whether the programme is real, and the one you cannot measure directly. A holdout group is the only clean answer and most companies at this size cannot run one honestly.

A referral program, one month

Illustrative
  1. Customers asked4,200

    Post-purchase prompt

  2. 21% continue· 3,339 drop off
    Opened the widget861

    Saw the reward

  3. 36% continue· 549 drop off
    Shared a link or code312

    Participation

  4. 365% continue· -828 drop off
    Referred a click1,140

    Traffic they sent

  5. 8.4% continue· 1,044 drop off
    Referred customers96

    Attributed and kept

Click to sale
2.3%
Referral to sale
8.4%
Illustrative shapes to reason about, not benchmarks to compare against. No sourced participation figure exists for this category.

What goes wrong

Six ways referral programs quietly die

Five are design mistakes. The sixth is the one the software stops for you.

A reward nobody wants

Credit toward a product the customer buys once a year, a discount on a subscription they already pay annually, swag. The test is simple and rarely applied: would you personally do the work of asking a friend for this?

Asking at the wrong moment

A referral ask in a monthly newsletter reaches people in no particular state of mind. The same ask ninety seconds after the product did something useful reaches people who have just been reminded why they pay you.

A share flow with too many steps

Every step between the ask and the share loses people, and most flows have three more than they need. One click to a personal link, a code and a QR code, with an email invite that does not require leaving the page.

No code, only a link

Codes get spoken aloud, typed at checkout, and written in captions. Links do not. A referral programme with no code cannot be used in the situations where customers actually recommend things, in person, on a call, in a video.

Paying for self-referrals

The most predictable form of leakage, and the easiest to stop. Email and IP matching between the buyer and the referrer runs on every plan and can hold the reward for review or reject it outright, with the reason shown on the row.

Launching without attribution

A programme with no tracking produces a number nobody trusts and an argument nobody can settle. It is also unfixable retrospectively, the sales you cannot attribute this quarter stay unattributed forever.

Self-referrals, caught automatically

Buyer email and IP are matched against the referrer's on every plan. A match holds the reward for review or rejects it outright, with the reason on the row. Clicks carry proxy and VPN flags for the deliberate version.

Fraud prevention runs on every plan, including entry.

Customer stories

Take their word for it

The founders and affiliate managers behind three fast-growing programs, in their own words.

Moov
EcommerceFitness

Affiliates drive a serious share of our revenue across three countries. Since we moved to Referly everything runs from one place, tracking has been spot on, and our partners get paid without me chasing spreadsheets.

Alexander Koppers

Alexander Koppers

Affiliate Manager, Moov

$6.9M

in total affiliate-driven revenue

1,154
active affiliates
19,017
affiliate sales
1,240,030
clicks tracked
OptimizePress
SaaSSite builder

We have been running affiliate programs for over a decade, and ours is one of the biggest parts of how OptimizePress grows. Referly is the first platform where an operation with thousands of partners actually feels manageable, from recruiting new affiliates to tracking every sale they send.

James Dyson

James Dyson

Founder & CEO, OptimizePress

7 figures
in affiliate-driven sales
4,554
active affiliates
10,536
referrals
Letterly
Mobile appAI writing

This platform is quite simple and easy to use, and you can start attracting people to your affiliate program very quickly. A big plus is that the founder, Ayo, is super responsive.

Ilfat Khayrullin

Ilfat Khayrullin

CMO & Co-founder, Letterly

300K+
app users
689
active affiliates
24,114
clicks tracked

FAQ

Referral questions, answered

Mostly about money, tax and whether any of it was incremental.

14-day free trial · No credit card required · Cancel anytime

  • Work back from what a customer is worth to you and what you already pay to acquire one elsewhere. A referral reward that costs less than your blended acquisition cost is worth paying; one that costs more is a discount programme wearing a different name. Two-sided credit is usually the cheapest structure that still works, because credit costs margin rather than cash.