Affiliate commission tiers that partners actually reach

Affiliate commission tiers work when the first threshold sits just above where many partners already are, and they cost little when the reward applies only to customers above the threshold. In the illustrative portfolio below, 50 partners refer 230 paying customers a year at a base 20% commission. A one-time $250 bonus at 10 paying customers adds $2,000 a year; raising the rate to 30% on customers after the 10th adds $3,240; applying that 30% to every customer once the partner qualifies adds $6,840. A tier at 100 customers adds nothing because nobody reaches it, and that is exactly why it motivates no one. This post shows how to set thresholds, compare designs, avoid cliffs and announce tiers.

What are affiliate commission tiers for?

A tier changes a partner's reward once they cross a threshold. It has two jobs. The first is to keep your best partners from leaving for a competitor with a higher headline rate, by paying them more without paying everyone more. The second is to give partners in the middle a reason to send one more customer this quarter. Those two jobs pull the design in different directions: the first favors a high rung for a few partners, the second favors a low rung that many partners can see.

There is research behind the second job. Kivetz, Urminsky and Zheng's study of reward programs found that members of a coffee shop loyalty program bought more often as they got closer to a free coffee, and that giving people artificial progress, two "bonus" stamps on a 12-stamp card, made them complete the required purchases faster [1]. They also found that stronger acceleration toward the goal predicted better retention. A partner program is not a coffee card, but the lesson transfers: people work harder when the goal is close and visible, and barely at all when it is far away.

Should tier thresholds count clicks, signups or paying customers?

Count paying customers. Clicks are easy to inflate and say nothing about value. Trial signups are better but still reward volume over fit; a partner can send hundreds of trials that never convert. A paying customer is a customer whose invoice was paid, which is the same event your commission is calculated from. In Stripe, a subscription generates an invoice each billing period and sends an invoice.paid event when payment succeeds [2], so the count is a fact in your billing data, not an estimate.

Three details need a decision and a sentence in your terms. First, the counting window: lifetime, calendar year, or a rolling 12 months. Lifetime is simplest and never takes a tier away; a yearly window keeps the program active but means a partner can drop a rung on January 1. Second, refunds: a customer refunded in full during the hold period should not count, since the commission row it created was cancelled by a negative one; Stripe allows full and partial refunds, and several refunds against one charge [3]. Third, what counts as one customer: count customers, not invoices, so that an annual plan and a monthly plan each count once.

RelayWonder's tiers use paying customers as the threshold and let each rung pay a bonus, a higher recurring rate, or both. Whatever tool you use, the threshold should be a number the partner can see in their dashboard and check against their own records.

An example partner portfolio to price tiers against

You cannot price a tier design without knowing where your partners are. Partner programs are usually very uneven: many partners refer one or two customers, a few refer dozens. The portfolio below is illustrative; replace it with a count from your own data before deciding anything.

The base economics: 230 customers × $450 = $103,500 of first-year revenue, and $20,700 of commission, exactly 20%. Notice the shape. Two partners bring 80 customers, more than the 30 occasional partners combined. And there is a gap: nobody sits between 4 and 12 customers. That gap will matter when we choose thresholds.

Bonus vs higher recurring rate: the cost math of each design

Here are five designs priced against that portfolio. A bonus is a one-time payment when a partner crosses the threshold. A higher recurring rate raises the percentage, and the important choice is whether it applies only to customers referred after the threshold (forward-only) or to all of the partner's customers once they qualify (retroactive).

1. Bonus: $250 once at 10 paying customers. Reached by the 6 strong and 2 top partners. 8 × $250 = $2,000.
2. Higher rate, forward-only: 30% instead of 20% on customers after the 10th. The extra 10% is $45 per customer. Strong partners have 2 customers above 10 each, top partners 30 each: 6 × 2 + 2 × 30 = 72 customers × $45 = $3,240.
3. Higher rate, retroactive: 30% on all customers once a partner reaches 10. 6 × 12 + 2 × 40 = 152 customers × $45 = $6,840.
4. Out-of-reach rung: 40% after 100 paying customers. Nobody reaches it. $0.
5. Two-rung ladder: $50 bonus at 3 customers, then 25% on customers after the 10th. The bonus reaches 12 + 6 + 2 = 20 partners: $1,000. The extra 5% is $22.50 per customer on 72 customers: $1,620. Total $2,620.

Extra cost is only half the picture. The other half is how many additional customers the design has to cause to pay for itself. Use gross profit, not revenue: at an 80% gross margin, a customer's first year is worth $450 × 0.8 = $360 to you. Each additional customer brought in under a design is worth $360 minus the commission paid on it. Divide the design's extra cost by that number.

Three conclusions follow. Retroactive rates are expensive: the same 30% costs more than twice as much as forward-only, because you are paying a higher rate on customers the partner had already brought in at the lower rate. Bonuses are the cheapest way to reward the top of the program, because the cost is fixed per partner. And the two-rung ladder reaches 20 partners instead of 8 for roughly the cost of the single bonus, which is the design most likely to change behavior in the middle of the program.

How do you avoid tier cliffs that nobody reaches?

A cliff is a rung so far above where partners are that it changes nothing. The 40%-after-100 design is the extreme case: it looks generous on the program page and costs nothing, because no partner in the portfolio is within reach. Partners notice. A tier that no one has ever reached reads as marketing, not as an offer.

The goal-gradient finding suggests a better rule: put the first rung just above a cluster of partners, so that the goal is one or two customers away for many people [1]. In the example portfolio, the 12 regular partners each have 4 customers a year. A rung at 5 is one customer away for all twelve of them; a rung at 10 is six customers away, and the portfolio has a gap there, so it rewards only partners who were already well past it. If you want one rung that moves behavior, 5 is the better threshold for this portfolio, and 10 is the better threshold for retaining the top.

• Plot your partners by paying customers per year before choosing thresholds. Look for clusters, not averages.
• Put the first rung one or two customers above the largest cluster of active partners.
• Put a higher rung where your top partners are now, so they qualify on day one and have a reason to stay.
• Avoid any rung that no partner has reached in the last year, unless you are adding it deliberately as a long-term goal and say so.
• Show progress as a fraction, "4 of 5 paying customers", in the partner's dashboard and monthly email.

Showing progress matters because of the second half of the finding: perceived progress speeds people up [1]. Counting the customers a partner already has toward the first rung when you launch tiers is the program equivalent of the two bonus stamps. A partner who starts at "4 of 5" behaves differently from one who starts at "0 of 5", even when the remaining distance is the same.

Refunds, churn and tiers

Tiers interact with refunds and churn in ways that should be settled before launch. A refunded customer is a negative row in the ledger, and should reduce the paying-customer count as well. Disputes behave the same way: when a cardholder disputes a payment, the card network reverses it [4], and the commission on that invoice should be reversed too. If a partner crosses a threshold and then a refund drops them back below it inside the hold period, the simplest rule is that the bonus is not earned until the count holds through the hold. That is why a 30-day hold, which is RelayWonder's default, matters for tiers as well as for commissions.

Churn affects higher recurring rates more than bonuses. A 30% rate on a customer who churns after two months costs you little; on a customer who stays twelve months it costs $45 more than the base. The cost math above already uses expected invoices under 5% churn. If your churn is lower, every recurring-rate design costs more, and bonuses look relatively cheaper. Rerun the table with your own churn.

How to announce new commission tiers to partners

Announcing tiers well is as important as the design. Partners compare programs quickly, and a confusing announcement is read as a hidden cut. Write it as a short email and a matching block on the program page, with numbers in every line.

• State the rungs as sentences: "From your 5th paying customer: $50 bonus. From your 11th: 25% of each invoice for 12 months on new customers."
• State the counting rule: lifetime, calendar year or rolling 12 months, and that refunded customers do not count.
• State whether higher rates apply forward-only or to existing customers. Forward-only is cheaper; say so plainly rather than hoping nobody asks.
• Give an effective date and say that existing commission rows are not changed.
• Tell each partner where they stand today: "You have 4 paying customers. One more reaches the first rung."
• Never lower an existing partner's rate as part of a tier launch. If the base rate must change, announce it separately with notice.

The last point deserves emphasis. A tier launch that quietly lowers the base rate to fund the upper rungs feels to most partners like a pay cut with a bonus attached, and the partners who send you one or two customers a year are also the ones most likely to recommend you in a forum thread. Fund tiers from growth, not from the base.

A tier design procedure for this week

1. Export paying customers per partner for the last 12 months. Group partners into bands and look for clusters.
2. Compute first-year value per customer: (1 − (1 − churn)¹²) ÷ churn × monthly price, then multiply by gross margin.
3. Draft two or three designs: a bonus at the top, a forward-only higher rate, and a low first rung above your largest cluster.
4. Price each design against your actual portfolio: extra cost per year, commission share, and extra customers needed to break even.
5. Pick the design with the most partners within one or two customers of a rung at an acceptable cost.
6. Write the rules as sentences, including counting window, refunds, and forward-only or retroactive.
7. Announce with each partner's current progress, and review the portfolio again in six months.