An affiliate program is the rare acquisition channel that costs nothing until it works. You pay on results, after the sale, at a rate you set. That is also why almost every store eventually tries one, and why so many of them leak money quietly for a year before anyone looks closely.
This playbook is about running one on Shopify without the two failure modes that cause most of the damage: attribution you cannot trust, and terms you never wrote down.
Affiliate is not seeding
Worth separating, because the words get used interchangeably and the economics are opposite.
Seeding is free product with no obligation. You send it, you hope for content, you pay nothing further. The cost is up front and certain; the return is uncertain.
Affiliate is commission on sales the creator drives. You pay nothing until an order lands. The cost is uncertain but always justified, because it only occurs alongside revenue.
Most mature programs run both, in sequence: seed broadly, watch which creators actually post and actually convert, and graduate that handful into an affiliate arrangement. Seeding is the audition; affiliate is the contract.
The attribution problem
This is the part that decides whether the program is worth running.
The default approach — give each creator a discount code — is broken in a specific and expensive way. A code is a string. Strings get shared, screenshotted and submitted to coupon aggregators. Once that happens:
- Your margin goes to customers who were already buying, because the code now sits between your storefront and anyone who searches “[your brand] discount code” before checkout.
- The creator is credited for those orders, because the code is the attribution.
- You pay commission on them.
That third point is what makes it worse than an ordinary code leak. An ordinary leak costs you the discount. An affiliate leak costs you the discount and a commission on revenue the creator did not influence — and the creator is not doing anything wrong, because from their side the number simply went up.
You also cannot fix it cleanly. Disabling the code punishes the creator who is actively promoting it; leaving it up funds the leak.
A link does not have this failure mode. A customer cannot type a URL into the discount field at checkout. Someone who finds the link on a coupon site and clicks it is, at that point, a real click through the creator’s link — which is the thing you agreed to pay for. The attribution stays honest because the mechanism is the destination rather than a string.
What the customer’s offer should be
The creator’s audience needs a reason to use the link rather than going to your homepage.
Ten to twenty percent is the standard band. Below ten and it does not move behaviour. Above twenty and you are training the audience to wait for a creator link rather than buy at full price — and you are stacking that discount on top of a commission.
Do the arithmetic before you publish a rate. A 15% customer discount plus a 15% creator commission is 30% off contribution on every affiliate order. On a 40% gross margin catalogue, that is most of the margin gone. It can still be worth it for new-customer acquisition, but it should be a decision rather than a discovery.
The alternative worth considering: no customer discount at all, and a higher commission. Some creator audiences convert on the recommendation rather than the deal, and you keep your pricing integrity. Test it with a few creators before assuming it will not work.
Giving each creator a link
One link per creator, carrying three things:
- The customer offer, applied automatically. Nothing to type.
- The creator’s UTM, so the traffic is attributable in analytics as well as in orders.
- Limits, so a single partnership cannot cost an unbounded amount.
The offer should stay live while the audience browses. A creator talks about one product; their audience arrives interested in it and frequently buys something adjacent. A link that only discounts the single product mentioned converts worse than one that keeps the creator’s pricing active across the storefront, and it reads as meaner.
For creators promoting a specific launch, the link can open with that product already in the cart. For creators talking about your brand generally, it should land them on the storefront with pricing applied.
Writing the terms
Almost every affiliate dispute traces back to something nobody wrote down. Cover these:
- Commission rate, and whether it is calculated on order value before or after discount and shipping. Say which.
- Attribution window. Thirty days is a sane default. State it.
- What counts as qualifying. Are subscription renewals commissionable, or only the first order? Are orders from existing customers included?
- Payout schedule, and that it runs after the returns window.
- Refunds and chargebacks. Commission is reversed on refunded orders.
- Branded-term bidding. Prohibited — see below.
- Termination. How either side ends it, and what happens to pending commission.
None of this needs to be long. A page is enough. The point is that it exists before the first payout rather than after the first argument.
The branded-search trap
The most common way affiliate programs lose money: a creator runs paid search on your brand name, intercepts customers who were already navigating to you, and collects commission on orders you would have received anyway.
From the reporting it looks like a star performer. In reality you are paying a commission to buy back your own traffic.
Prohibit branded-term bidding explicitly in the terms, and check periodically — searching your own brand name from a clean browser is usually enough to catch it.
Tracking what matters
Per creator, not in aggregate:
- Orders and revenue. The obvious pair.
- New versus returning customers. An affiliate driving existing customers is cannibalising, not acquiring. This is the number that separates a real partner from an expensive coupon.
- Refund rate. A creator whose audience returns at twice your baseline is overselling, and the gross number is lying to you.
- Net commission against contribution. After the customer discount, the commission, and refunds — is the order still profitable?
- Repeat rate of acquired customers. The real return is whether the customers they send come back. A creator with lower volume but higher repeat rate is worth more than the raw numbers suggest.
Expect a power law. A small number of creators will drive most of the revenue, and a long tail will drive approximately nothing. The job of the program is to find the first group quickly and stop spending admin time on the second.
Paying out
Monthly, covering orders older than your returns window.
Paying on order is tempting and creates a permanent clawback problem — you spend the relationship asking for money back, which sours something that should be low-friction. Paying after the window closes means the number you send is final.
Make the schedule explicit up front. Creators are generally relaxed about waiting if they know the rule, and reliably unhappy about a payout that arrives on an unpredictable date.
When to adopt a platform
Links and a spreadsheet work well up to roughly a few dozen active creators. Beyond that, the admin becomes the constraint: tiered rates, creator-facing dashboards, automated payouts and tax handling are genuinely worth paying for at scale.
The mistake is adopting the platform first. A platform does not solve attribution — it just reports whatever mechanism you gave it, and if that mechanism is discount codes, you get well-presented numbers that are still inflated by leakage. Fix the mechanism, then buy the admin.
Recruiting creators
The program is only as good as who is in it, and the best sources are rarely the obvious ones.
Your own customers first. Search your customer list against social platforms, or simply ask in a post-purchase email whether anyone creates content. Customers who already bought and liked the product convert their audiences at rates cold outreach never matches, because the endorsement is real and reads that way.
Creators who already mention you. Brand-mention monitoring surfaces people talking about you without any arrangement. They have already done the hard part unprompted; the conversation is about formalising something rather than persuading someone.
Graduated seeding recipients. The strongest pipeline. Seed broadly, watch who posts and whose posts convert, and offer affiliate terms to the handful who did both. You have evidence rather than a guess, which is the whole argument for running seeding upstream of affiliate.
Competitor affiliates. Creators promoting an adjacent, non-competing brand in your category are already comfortable with affiliate mechanics and already have the right audience. Cold, but a much warmer cold than a general creator search.
What underperforms: buying lists, mass DM campaigns, and any affiliate directory where the sign-up is open. Open programs attract coupon-site operators rather than creators, and coupon sites are precisely the traffic that inflates attribution without adding customers.
Onboarding: what a creator actually needs
Most programs hand over a link and go quiet. The ones that work send a short brief covering:
- The link, and how it behaves. That the discount applies automatically, that it works across the storefront, and that their audience does not need to type anything.
- What to say about the offer. Not a script — a sentence they can adapt. Creators consistently under-explain the offer because they assume their audience will work it out.
- Products worth featuring. Your best converters, and anything with margin room. Left alone, creators default to the cheapest item.
- What is off-limits. Branded-term bidding, coupon-site submission, and any claim you cannot substantiate.
- When and how they get paid. Removes the most common source of follow-up messages.
A page. The return on writing it once is that you stop answering the same four questions individually, and creators who know what good looks like produce more of it.
Commission structures
Flat percentage. One rate for everyone. Simple to explain, simple to administer, and the right default until the program is big enough that the differences matter.
Tiered by performance. A base rate that increases past a volume threshold. Rewards the small group that drives most of the revenue and gives mid-tier creators something to aim at. Worth introducing once you have enough creators for the tiers to be meaningful — below about twenty, it is administrative overhead for no behavioural effect.
Flat fee per order. Suits narrow catalogues where order values cluster tightly. Easier for creators to reason about — “$8 an order” is more motivating than “12%” — and it protects margin on higher-value orders.
Hybrid: fee plus commission. A modest upfront payment plus a lower ongoing rate. Appropriate for larger creators who will not work on pure performance, and a way to buy a guaranteed post without paying full sponsorship rates.
New-customer bonus. A higher rate on first-time buyers than on returning ones. This directly targets the cannibalisation problem: it pays more for the thing you actually want, and less for orders that would likely have happened anyway.
Whatever you pick, model it against contribution margin including the customer discount before publishing. Rates are easy to raise and very hard to cut.
Reading the numbers honestly
A few patterns worth recognising early, because each one looks like success in a default report.
The brand-term harvester. Exceptional conversion rate, almost entirely returning customers, revenue concentrated on your best-selling product. This is someone intercepting existing demand. Conversion rate is high precisely because those customers were already buying.
The coupon-adjacent affiliate. Volume with no corresponding social reach. If a creator drives 200 orders and their posts get 40 engagements, the traffic is not coming from their audience.
The high-return creator. Strong gross numbers, refund rate well above baseline. Usually overselling — promising fit, results or quality the product does not deliver. Costs you the return shipping and the customer.
The genuine long-tail. Small, steady, low refund rate, high new-customer share. Individually unimpressive and collectively often the healthiest part of the program. Do not prune these in favour of a single high-volume affiliate who is harvesting brand terms.
The habit worth building: review new-customer share and refund rate alongside revenue every time, never revenue alone. Almost every affiliate problem is visible in those two columns before it shows up anywhere else.
A 60-day rollout
Days 1-10 — decide the economics. Commission model, rate, customer offer, attribution window. Model the combined cost against contribution margin. Write the terms.
Days 11-20 — recruit five creators. Not fifty. Five, from your own customer list or from people already mentioning you. A small first cohort means you find the problems in your links, terms and tracking while the cost of fixing them is nothing.
Days 21-35 — run and watch. Per-creator links out, brief sent, first content live. Watch new-customer share from the first orders — it is the earliest signal that attribution is behaving.
Days 36-45 — first payout. Run it on schedule even if the amounts are trivial. The first payout is where you discover whether your process works, and doing it while the numbers are small is much better than discovering it at scale.
Days 46-60 — expand or fix. If new-customer share is healthy and refund rate is normal, recruit the next cohort. If either is off, fix the terms or the creator mix before adding anyone.
Sixty days is enough to know whether the program works. It is not enough to know which creators are keepers — that takes a couple of quarters, because the signal is repeat rate of the customers they sent.
Where affiliate sits against your other channels
Affiliate is often adopted as a replacement for paid acquisition and is better understood as a complement, because the two fail in opposite conditions.
Against paid social. Ads buy attention from people who have never heard of you; affiliate buys endorsement from someone their audience already trusts. Ads scale linearly with budget and stop the moment you stop paying. Affiliate scales with relationships and keeps producing from content that stays up. Ads give you precise control over targeting; affiliate gives you none, which is why creator selection carries the weight that audience targeting does elsewhere.
Against your own email. Email is nearly free and reaches people who already bought. Affiliate reaches people who have not. Any month where affiliate revenue is dominated by returning customers is a month where it duplicated your email programme at a commission — which is exactly what the new-customer share metric is there to catch.
Against seeding. Covered above, but the sequencing is the point: seeding is how you discover who converts, affiliate is how you scale the ones who do. Running affiliate without a seeding pipeline means recruiting on guesswork.
The honest framing for a small store: affiliate rarely becomes a primary channel. It becomes a reliable ten-to-twenty percent that costs nothing when it is not working, which is a genuinely good property that no paid channel shares.
Common failure modes over a longer horizon
Programs rarely fail on day one — they degrade. The recognisable patterns:
Rate inflation. A large creator asks for more, gets it, and the exception becomes the expectation as word spreads. Because rates cannot practically be cut, the program ratchets upward until the economics stop working. Defend against it with a published tier structure so a higher rate is something earned rather than negotiated.
The stale roster. Fifty creators enrolled, six active. The dormant accounts cost nothing directly but they distort every average you look at and hide whether the active core is growing. Prune on a schedule.
Attribution drift. Someone adds a second tracking mechanism — a code alongside the link, or a platform pixel — and the two disagree. Now every payout is a reconciliation. Pick one source of truth and keep it.
Discount dependence. A year in, a meaningful share of your audience has learned to search for a creator link before purchasing. The affiliate offer has become the price. This is the slow version of the coupon-site problem and the reason to keep the customer-facing discount modest.
Owner attention. The program works, so nobody looks at it. Six months later a brand-term harvester has been paid every month since spring. It needs an hour a month, permanently.
Getting the first ten right
The single most useful piece of advice: start much smaller than feels worthwhile.
Five to ten creators is enough to expose every problem in your links, terms, tracking and payout process, and small enough that fixing those problems costs nothing. Every mistake you make at that scale is cheap; the same mistakes across fifty creators mean fifty awkward conversations.
Concretely, before the eleventh creator you want to have: run a real payout, seen at least one refund flow through the commission calculation, checked new-customer share on actual orders, and searched your own brand name to confirm nobody is bidding on it. All four are cheap checks that become expensive to retrofit.
What not to do
Do not open the program publicly. An open sign-up form attracts coupon operators, not creators. Recruit deliberately.
Do not pay on order. Pay after the returns window. Clawbacks poison an otherwise easy relationship.
Do not skip the terms because it feels informal. The first dispute will be about attribution windows or refunds, and having written it down in advance turns an argument into a lookup.
Do not use discount codes as the attribution mechanism. The entire cost structure of the program depends on attribution being honest, and codes make it arguable the moment one leaks.
Do not judge creators on revenue alone. New-customer share and refund rate are what separate an acquisition channel from an expensive way to discount your existing customers.
Further reading
- Influencer seeding and gifting playbook — the top of the funnel that feeds an affiliate program.
- Shopify discount codes playbook — why codes leak and what to use instead.
- Checkout Links for influencers — the link layer for creator programs.
- Track influencer sales on Shopify — attribution mechanics in detail.
- How to create influencer tracking links on Shopify — implementation walkthrough.
- UTM tracking best practices — per-creator attribution setup.
