A free trial opens a gap between "a customer showed up" and "a customer paid", and that gap is where affiliate programs quietly lose their best partners. The affiliate sent the traffic, the trial started, the dashboard stayed empty for two weeks, and then they stopped sending traffic.
Tracking is not the hard part. Polar, Stripe, and Paddle all deliver an event when a trial starts and another when the first invoice clears. The hard part is that a trial is a referral without revenue, and most programs never decide in writing what they pay for a referral. Here is how the money moves between a $0 order and the first real charge, plus the trial policies worth publishing.
TL;DR
- A $0 order is a referral, not a conversion. Record which affiliate sent it, pay nothing that day, and let the first paid charge inherit the affiliate from the subscription id.
- Commission belongs on the first invoice that actually clears. Money that never arrived in your account is not commissionable, no matter how enthusiastic the trial was.
- Polar, Stripe, and Paddle report a settled $0 order with different fields, and only some payloads carry the referral reference at all. Verify the path in test mode before you promise recurring pay on trials.
- Freemium is not a trial. A free account that never converts has no invoice to multiply, so decide up front whether you pay per signup, per upgrade, or not at all.
- Trial abuse is cheap to run and cheap to block: self-referral by matching email, disposable signup domains, and duplicate signups from one address.
Why a trial breaks the usual commission rule
A normal affiliate conversion has one event, one amount, and one affiliate: the customer clicks, buys, the webhook arrives with the referral attached, and the platform multiplies.
A trial splits that into two events with an awkward gap between them. Event one is the trial start: an order or checkout that settles at $0, carrying the referral code but no commissionable amount. Event two is the first paid invoice, one to four weeks later, which carries money but often no referral code, because billing systems do not keep checkout metadata attached to a subscription forever, and renewal invoices rarely carry it at all.
So the affiliate id has to travel across that gap on the only thing both events share: the subscription id. Every serious tracking layer keeps a stored mapping for exactly that reason. Without it, the first paid invoice of a trial looks like an unattributed payment from a customer nobody referred, and the affiliate who did the work gets nothing.
That failure mode is predictable: the best referrer leaves after five trials convert and none of the revenue lands, and a working tracker reads as a broken dashboard.
What your billing system calls a trial
"Trial" is not one thing across providers, so it helps to know what the event stream looks like before writing policy.
A card-required trial usually starts with a real order object that settled at zero: a $0 invoice, a $0 checkout session, a completed transaction with a zero total. That object is the best-attributed event you will get: it was created during checkout, while the referral metadata was still on the session.
A card-free trial sometimes creates no order object at all. The subscription appears, and the first event you ever see is the paid invoice when the trial ends, usually with no referral field. The same is true of trials started from your dashboard, an API script, or your sales team: no checkout happened, so no metadata was captured.
That second case is why some programs "lose" trial conversions while others do not. It is not a bug in the affiliate tool: the event never carried attribution, and no downstream system can invent it. If a sales motion exists at all, read Partner-Led Growth: Affiliate vs Referral vs Reseller before deciding how much weight affiliates carry in it.
How Polar, Stripe, and Paddle report a trial
The providers differ in which event starts a trial, what it contains, and how the later paid charge is matched.
| Provider | Event | What it carries | What a trial program should do with it |
|---|---|---|---|
| Polar | order.created with paid false | referral code, subscription id, no payable amount | Keep the referral, pay nothing |
| Polar | order.paid | net amount after discount, pre-tax | Pay commission on the first one tied to a mapped subscription |
| Stripe | checkout.session.completed in subscription mode, amount_total 0 | referral in session metadata, subscription id | Store the referral against the subscription id |
| Stripe | invoice.payment_succeeded with billing_reason subscription_cycle | amount paid, subscription id, no referral metadata | Inherit the affiliate from the subscription mapping |
| Stripe | invoice.payment_succeeded, first invoice | skipped on purpose | Prevents paying the same checkout twice |
| Paddle | transaction.completed with a zero subtotal | subscription id, referral in custom data | Store the mapping, pay nothing yet |
| Paddle | transaction.completed with a paid subtotal | subtotal in minor units, no customer email | Pay commission, and note that email-based checks cannot run here |
| Generic webhook | your own event shape | whatever you send | Only works if you pass a subscription id |
Two details matter more than they look.
First, skipping an event is sometimes correct. Stripe skips a subscription's first invoice because the checkout session already recorded that payment, and counting both would double-pay it. Polar ignores an order that is explicitly unpaid or pending and waits for the paid event on the same order id. Paddle requires a completed transaction, since its paid event can fire before settlement finishes. All three choices exist to stop one sale becoming two conversions.
Second, refunds can beat the sale. Providers do not guarantee delivery order, so a refund can arrive before the order it refers to. A tracker that only handles refunds it can find will pay commission on a sale that was already refunded. Sled writes a tombstone for the refunded order id, creates the conversion already marked refunded when the order lands, and skips the first-commission email for money that was clawed back. The full story is in When the Refund Beats the Sale.
Referrals versus conversions: the $0 order
The distinction that settles most trial arguments: a referral is a claim on a future conversion, a conversion is a payable event.
When a tracked $0 order arrives with both a referral code and a subscription id, store the pair and pay nothing. Commission is a percentage of an amount, the amount is zero, and for once the mathematical answer and the commercial answer agree. What you cannot do is discard the mapping, because it is the only thread back to the affiliate once the real invoice arrives.
Two rules make that mapping behave:
- First referral wins. If a $0 event is redelivered, or a later trial order tries to reassign the same subscription, the original affiliate stays attached.
- A trial's first paid charge is not a renewal. The affiliate is inherited from the trial mapping, not from a prior paid conversion, so a dashboard that labels it "recurring" is misdescribing what happened. Sled keeps that flag false so the first payment reads as the conversion it is.
The dashboard consequence is one category affiliates need to see: referred, not yet converted. Without it, a trial-based program looks broken for the first month of every affiliate's life.
Freemium is a different animal
A freemium product has no trial end date, so the two-event model collapses into one event and a long tail of hope.
With freemium, the meaningful event is the upgrade. A free account pays you nothing, and the tracker has to attach the affiliate to the account at signup so an upgrade three months later lands on the right person. That is the same subscription-mapping idea applied to accounts: capture the referral at signup, carry it to the first paid plan, inherit it for later renewals.
The trap is paying on signups. A per-signup bounty turns affiliates into traffic buyers with no quality filter, and free signups are the cheapest thing in SaaS to generate. You get hundreds of accounts that never upgrade, some of them fraudulent, and a commission bill that scales with your worst channel.
Where a freemium program should be generous is on the affiliate's own tooling. Sled's 10-day free trial is the full feature set with no affiliate cap, including conversion tracking and the read-only MCP endpoint that lets an affiliate's agent pull their own numbers. A tool that charges before you can prove tracking works cannot be evaluated, which is why the trial exists. The same reasoning is why the paid plan stays flat at $19/mo with no cut of commissions, as Stop Paying a % of Commissions to Your Software argues.
Trial commission policies you can publish
There is no single correct trial policy, but there are six defensible ones.
| Policy | When the affiliate is paid | What it rewards | Where it breaks |
|---|---|---|---|
| Pay on first paid invoice | After the trial converts and the first charge clears | Real revenue, nothing else | Weeks of silence, and nothing at all if trials rarely convert |
| Flat bounty at trial start | When the tracked $0 order is recorded | Volume of qualified trials | Pays for trials that never convert, so it attracts low intent traffic |
| Half bounty at start, half on first payment | Split across both events | Effort plus follow-through | Needs reconciliation unless the tool stores the mapping |
| No commission on trials, paid plans only | On the first paid invoice, trials invisible | Simple books and simple copy | Identical arithmetic to option one, only the wording differs |
| Bounty plus reduced recurring rate | Flat amount on conversion, lower percentage on renewals | Large upfront wins with a predictable tail | Lifetime math gets hard to explain, and affiliates will ask |
| Reduced flat rate for the first N months | On every invoice in the trial-to-paid window | Predictable cost while a customer proves retention | Punishes affiliates for your own churn curve |
Two of these describe identical arithmetic: "pay on first paid invoice" and "no commission on trials" differ only in wording. Choose whichever sentence reads better to your audience, then spend the effort where it changes behavior, which is the split option. It is the only policy that pays something inside the trial window, and the only one that keeps momentum alive while the customer decides.
Whatever you choose, make the terms say what the dashboard does. A program that pays a dollar per tracked trial start while showing a column of $0 conversion rows is asking for support tickets. Affiliate Program Terms: Free Template for SaaS has the clause structure for this.
Trial abuse: the checks worth running before you pay
A trial is the cheapest fraud surface in affiliate marketing: under a bounty policy the affiliate does not need a sale to be paid, and under a per-signup freemium policy they do not even need a card.
The three checks that catch most of it:
- Self-referral by email. When the customer email equals the affiliate email, the conversion is the affiliate buying their own product. Sled rejects those by default, and an owner can disable the rule for a genuine edge case.
- Disposable signup domains. Throwaway mailboxes make it free to create ten affiliate accounts and refer yourself ten times. Blocking known disposable providers at affiliate signup is on by default and costs a legitimate affiliate nothing.
- Duplicate identity patterns. One person starting several trials from one connection shows up in click data. Sled stores a salted hash of the visitor IP with each click instead of the address itself, which is enough to spot repetition.
One honest limitation: Paddle payloads carry a customer id and no customer email, so an email-based self-referral comparison cannot run on Paddle conversions. The rule still covers your Polar and Stripe traffic, and it is worth knowing which provider has the gap rather than assuming uniform coverage. For the wider catalogue of tricks and counters, see 10 Affiliate Fraud Tactics (And How to Block Them).
When the trial converts, then refunds
Trials interact badly with refunds in one specific way: the commission is often issued before the refund window closes, so a program paying on conversion can pay out on a customer who leaves eleven days later.
Handle it like any refund. The refund event flips the conversion to refunded, the amount stops counting toward the affiliate's balance, and if the refund arrives before the order, tombstone logic stops the conversion from ever becoming payable. Past invoices are a different question: a subscription cancel does not claw back commissions earned on invoices that cleared, and it should not, because retroactive clawbacks are the fastest way to lose a partner. If you pay a bounty at trial start, put a hold period in writing that matches your refund window, so the bounty becomes payable only after the money has stayed with you.
What to write in your affiliate terms about trials
Four sentences cover almost every dispute. Adapt them, but keep the specificity.
- What counts as a conversion: the first paid invoice from a referred customer, after any refund window has closed.
- What does not count: trial starts, $0 orders, free-plan signups, and invoices later refunded or charged back.
- How attribution is decided: the referral recorded at signup or trial start travels with the subscription for its lifetime, first referral wins, and a later link does not reassign an existing customer.
- How renewals are paid: the same rate applies to each invoice that clears on that subscription until it is cancelled, and cancellation does not reverse commissions already earned.
Add one operational number: your attribution window, in days. Sled tracks clicks for 90 days by default, generous for products with a long evaluation cycle and tightenable per program.
Edge cases that show up in month three
Trials produce situations that are obvious in hindsight and confusing in the moment.
- The trial that starts twice. A customer trials, lets it lapse, then trials again with another link. First referral wins, so the original affiliate keeps the eventual conversion. Say so in the terms, because the second affiliate will ask.
- The upgrade that skips checkout. An in-app upgrade may carry the subscription id but no referral metadata, which is fine as long as the subscription was mapped at signup or trial start.
- The affiliate you remove. Removing an affiliate stops new conversions and future earnings while historical conversions and payouts stay intact. Do not rewrite the past.
Each has a defensible answer and none has a self-evident one, which is why the programs that avoid the argument wrote theirs before it happened.
FAQ
Should affiliates get paid on free trials? Not on the trial itself, because there is no revenue to take a percentage of. Pay on the first invoice that clears and let the affiliate inherit that customer through the subscription id, so the work is rewarded when money arrives. To keep momentum inside the trial window, a split bounty pays a small amount at tracked trial start and the balance at first payment.
How does a trial conversion get attributed if the first paid invoice has no referral code? Through the subscription id. When a tracked $0 order arrives with both a referral code and a subscription id, the pair is stored, and every later event on that subscription resolves to the affiliate. Renewals use the same path, which is how recurring commission works on ordinary subscriptions too. Without a tracked checkout there is nothing to inherit from, and that invoice lands unattributed, so test your own flow before promising otherwise.
Do freemium signups count as affiliate conversions? Treat them as referrals, exactly like trial starts. Record the affiliate against the account, pay nothing at signup, and pay when the account upgrades. A per-signup bounty on a free plan pays for volume with no quality signal, which is the input a fraud ring wants.
What stops an affiliate from claiming trials with their own link? Email matching is the first filter: when the customer email equals the affiliate email, the conversion is rejected by default. Disposable-domain blocking at affiliate signup removes the cheapest way to create multiple identities. Paddle payloads carry a customer id and no email, so that comparison cannot run there: the guard is provider-dependent.
Running trials, freemium, or both on Polar, Stripe, or Paddle? Sled stores the referral against the subscription, pays on the first invoice that clears, blocks self-referrals, and keeps renewals attributed for a flat $19/mo with no cut of your commissions. Start free trial, then read Affiliate Commissions on Usage-Based Pricing for the variable-bill version of the same math.