How to Run a Customer Referral Programme With Automated Reminders

Offer a simple reward that both people get, ask when customers are happiest, and automate three reminders: an ask a day or two after a good job, one nudge a few weeks later, and a "your reward is ready" message when the friend books. A spreadsheet or referral app tracks the codes; your email or text tool sends the reminders.

The automation is the easy part. Most referral schemes fail on the offer and the moment: a reward too small to bother with, a code that's awkward to use, or an ask buried at the bottom of an invoice email among five other messages. And reminders need an off switch. Two asks per job is plenty; a customer who doesn't share after that shouldn't hear about it again until their next visit.

Follow me on Instagram@sagnikteaches

Work out the reward before you build anything

Our illustrative business is a bicycle repair shop with two mechanics and a steady stream of commuters and weekend riders. Its owner started with a sum rather than a guess.

Connect on LinkedInSagnik Bhattacharya

• Average job: $110.
• A new customer typically comes back about twice more in their first year, so a first year is worth about $330 in takings.
• Last year's local ads cost about $600 and brought roughly 15 new customers the shop could trace, so around $40 each.

So a reward costing less than $40 per new customer beats the ads, and a referred customer tends to arrive already trusting the shop. The owner chose $15 off the next job for the referrer and $15 off the first job for the friend: $30 in total, paid only when the friend's job is done and paid for. Because it's a discount rather than cash, the real cost is lower than $30, since part of it is margin on work the shop does anyway.

Subscribe on YouTube@codingliquids

The shop's previous attempt is a useful warning. For a year it printed "refer a friend, get 10% off" at the bottom of every invoice. That's $11 on an average job, offered only to the referrer, with no code and nothing for the friend. It produced two referrals that anyone could trace. The new scheme's first month beat the whole of that year, with the same customers.

Where the referrals get tracked

Every referral scheme needs three things: a unique code or link per customer, a way to spot the code when a friend buys, and a record of who's owed what. There are three common setups:

The bike shop's booking system had a free-text "how did you hear about us?" field, which it renamed "referral code (optional)", and an automation tool that could read completed jobs. That made the third setup the natural choice. Always check your own software first; paying for a referral app your booking system already duplicates is a common waste.

The reminder timeline

Here's the full schedule the shop settled on. Each line is one automated message with a clear reason to exist and a stop rule.

Why two days after the job for the first ask? Long enough that the bike has been ridden and the customer knows the brakes work; soon enough that the visit is fresh. Asking in the payment email is too early and too cluttered. The fifth message deliberately isn't automated: a top referrer deserves a line from a person.

Building it with a spreadsheet and Zapier

This version uses a Google Sheet and Zapier; Make and other automation tools can do the same. Set up the sheet first, with one row per customer:

Customer name · Email · Phone · Code · Code issued · Last job date · Referrals completed · Reward owed · Reward expires · Opted out

Then the automations:

1. Issue the code. When a job is marked complete in the booking system, create or update the customer's row and give them a code, for example the first four letters of their surname plus the last three digits of their booking number. A Formatter step can join the pieces together, and Formatter steps don't use tasks.
2. Send the ask. A Delay step waits two days (delays don't count as tasks either), then the email tool sends message 1 with the code.
3. Send the nudge. A second Delay waits until day 21, a lookup checks the sheet for referrals, and a Filter step stops the Zap if one exists or the customer opted out. Only then does message 2 go.
4. Spot the friend. When a new booking arrives with something in the referral code field, look the code up, check the friend is genuinely new, and once their job is paid, add a referral and a reward to the referrer's row.
5. Send the reward message straight away, with the expiry date.
6. Check expiries weekly with a scheduled Zap that finds rewards expiring in 14 days and sends message 4.

Two limits to know. Zapier's Delay step can hold a run for up to about a month, so waits longer than that need a different approach, such as a scheduled daily check of the sheet. And task counts add up: at about 60 completed jobs a month, each run of the main Zap uses a handful of tasks for the sheet update, emails and lookup, so expect a few hundred tasks a month. That fits Zapier's Professional plan (750 tasks, $19.99 a month billed annually) but not the Free plan's 100 tasks and two-step limit.

Test it end to end with your own email address as the customer and a colleague's as the friend before any real customer gets a message. The shop's first test found that the nudge fired even after a referral, because the lookup checked the wrong column. Found in testing, that's a two-minute fix; found by customers, it's an embarrassing email.

Writing the ask with AI

Generic referral emails are easy to ignore. The difference AI makes is personalisation at scale: it can turn a mechanic's job notes into one friendly, specific line. A prompt for an AI step, or for drafting the templates yourself:

Write a short referral email from a small bicycle repair shop.
Customer first name: [name]. Bike and work done: [job notes].
Their code: [code]. Offer: $15 off their next job and $15 off their
friend's first job, once the friend's job is done. Expires [date].
Rules: under 90 words, warm and plain, one specific line about their
bike from the notes, no exclamation marks, no pressure, and a
one-click way to opt out of referral emails.

Before, the shop's old message, sent with every invoice:

REFER A FRIEND AND SAVE! Tell your friends about us and get 10% off your next service. Terms apply.

After, an AI-drafted version for one customer (illustrative):

Hi Jo, hope the gravel bike's feeling sharper now the new brake pads have bedded in. If you know anyone whose bike needs some love before winter, give them your code, KEEN482. They'll get $15 off their first job, and once it's done you'll get $15 off your next one. It's valid until 31 January. (Don't want these? Reply "stop" and we won't send them.)

Review the first 20 AI-written messages before letting the step run unsupervised. In the shop's first batch, the AI turned the job note "customer declined new chain, worn 0.75" into "your new chain", which was the opposite of what happened. The fix was an instruction to mention only work marked "done" in the notes, and to skip the personal line if the notes were unclear.

The nudge and the reward message

The nudge should be shorter than the ask and offer a new reason, not repeat the old one. The shop's version:

Hi Jo, a quick one: the clocks change soon and we get busy with lights and winter tyres. If a friend's been putting off a service, your code KEEN482 still gets them $15 off. That's the last reminder from us.

"That's the last reminder from us" does real work. It tells the customer you won't keep asking, which makes the message feel less like marketing and more like a heads-up.

The reward message is the one that builds the habit, because it proves the scheme works: "Your friend Sam's bike is sorted, thanks for sending them our way. You've got $15 off your next job, valid until 30 April. Just mention it when you book." Send it the same day the friend's job is paid.

Fairness, fraud and privacy

• Pay rewards only after the friend's job is paid, not when they book. That removes most gaming.
• Block self-referrals: same email, phone or address as the referrer doesn't count. The lookup step can check this.
• Decide whether rewards stack. One reward per job, or several? Write it down.
• Don't collect friends' details. Let customers share the code themselves. Messaging people who never gave you their details can break marketing and data-protection rules.
• Keep referrals and reviews separate. Rewarding referrals is normal; rewarding reviews, or only asking happy customers for them, can break platform rules. Review gating explained covers why.
• Ask public sharers to say there's a reward. If someone posts their code on social media, advertising rules in many places expect them to mention they get something for it. A line in your terms asking them to do so is sensible.

The same caution applies if you're tempted to have AI write testimonials to go with the scheme; where the legal line sits on AI testimonials explains the risk.

Twelve weeks at the bike shop

The figures below are illustrative, from the first three months of the scheme:

That's about $25 of discounts per new customer, against roughly $40 for the ads, before counting the friends' future visits. About a third of the referrals came after the nudge rather than the first ask, which justified the second message. Seven of the 23 came from just three customers, who each got a handwritten note from the owner.

Reading the numbers and deciding what to change

• Few codes used at all: the ask isn't landing. Check open rates, try the ask by text instead of email, or mention it in person at collection.
• Codes shared but friends don't book: the friend's offer is too weak, or friends can't find how to book. Check the booking page works from a phone.
• Rewards earned but not used: the reminder before expiry needs to be clearer, or the expiry is too short for how often customers come back.
• A handful of customers produce most referrals: normal. Look after them personally, and consider a bigger reward from the third referral.

Look at the numbers monthly for the first quarter, then quarterly. If you segment customers by how often they visit, you can time the ask better for each group; segmenting your email list with AI shows how, even for a small list.

Referral schemes for a groomer, bakery, café and florist

• A dog groomer: hand over a small card with the owner's code at collection, when the dog looks its best, and send the ask by text the next day with a photo of the groomed dog. Friends book by phone, so the code is written in the booking diary.
• A bakery with an online shop: a referral app connected to checkout tracks codes automatically. The reminder schedule is the same, but the ask goes in the order-delivered email, not the order confirmation.
• A café: if the loyalty app has referrals built in, use them; bonus stamps for both people is enough. The reminder is a push notification, sparingly.
• A florist with corporate clients: discounts feel odd between businesses. A thank-you arrangement for the referrer's reception desk when a new corporate account starts works better, and the "ask" is a personal note from the owner, drafted with AI but signed by hand.

Referral schemes sit naturally alongside two other automations: asking for reviews after a good job, covered in getting more Google reviews with automated requests, and a loyalty programme for regulars, covered in building a loyalty programme with AI personalisation. Keep the three messages on separate days so customers aren't asked for everything at once.

Referral programmes: questions owners ask

Can I ask customers for their friends' email addresses?

It's better not to. Messaging someone who never gave you their details can break data-protection and marketing rules in many places, and it feels intrusive. Give customers a link or code to share themselves, by text, email or in person. The friend then contacts you, which also means the lead arrives warm.

Should the reward be cash, a discount or a gift?

For most small businesses a discount or credit on the next visit works best: it costs you less than its face value, and it brings the referrer back. Cash is simple but costs the full amount and can feel transactional. A small gift or a donation to a cause suits trades where discounts feel wrong, such as funerals or weddings.

How long should referral codes stay valid?

Long enough for a friend to need you, which depends on how often people buy. A bicycle service might be needed within three months; a wedding florist might wait a year. Pick an expiry that matches your typical gap between purchases, say it clearly, and send one reminder shortly before a reward runs out.

Do I need special software to run a referral scheme?

No. A spreadsheet, a unique code per customer and your existing email or text tool will run a small programme. Dedicated referral apps make sense for online shops, where codes can be tracked at checkout automatically. Check whether your booking, till or loyalty system already has a referral feature before paying for another tool.

Further reads

• How Much Does It Cost to Automate Lead Follow-Up With AI? — What automating the follow-up to referred leads costs.
• What Can AI Do for Your Email Marketing? 10 Practical Uses — Other emails worth automating once referrals are running.
• How to Turn Customer Reviews Into Marketing Copy With AI — Use happy customers' words in your marketing, too.
• Getting More Reviews After Every Job With AI Follow-Ups — Pair the referral ask with a review request after each job.
• When Does Zapier Get Too Expensive? Finding the Tipping Point — Know when your reminder automations outgrow Zapier's plans.
• How to Build a Welcome Email Sequence With AI — Welcome referred customers properly on their first visit.
• How Photographers Can Fill Mini-Session Slots With AI Marketing — A three-week, waitlist-first campaign for mini sessions: slot maths, past-client segments, copy prompts and honest ways to fill the last slots.
• ChatGPT Prompts for Pet Sitters and Dog Walkers That Win Clients — Twelve ready-to-paste prompts for pet sitters and dog walkers, from profile bio to daily updates, each with what to feed in and what to check.
• How to Build an Upsell and Cross-Sell Email Sequence With AI — Find what customers really buy next, plan a short triggered sequence, draft it with AI from a proper brief, and stop it the moment someone buys.
• AI Lead Generation for Small Businesses: What Actually Works — Which AI lead generation methods bring small businesses real customers, which ones burn your reputation, and how to measure the difference.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: ReferralCandy pricing page; Zapier help on task usage and Delay by Zapier; Zapier pricing (facts sheet). Example figures are illustrative.