Which Partner Program Metrics Matter, and Which Don't?

The default partner dashboard is built to make a program look busy: clicks climbing, partners joining, signups up and to the right. None of those numbers tell you whether to keep funding the program. A handful of others do, and they are less flattering.

Here is the short list I would review every month, followed by the numbers worth ignoring and the traps in the ones you do use.

What does the program really cost per customer?

Start here, because everything else is context for this number. Cost per acquired customer is everything you spend on the program divided by the new customers it produced in the same period.

Count more than commissions. Add the platform fee on commissions paid, the software subscription, any bonuses, contests or co-marketing money, and a realistic share of the hours your team spends on recruiting, support and review. Staff time is the item people leave out, and it is often the biggest.

A worked example, with made-up round numbers. In a month the program produces 30 new paying customers. Commissions came to $6,000. Software and platform fees were $1,000. You spent about 40 hours on the program, which you value at $50 an hour: $2,000. Total cost $9,000, so $300 per customer.

Is $300 good? It depends on what a customer is worth. If the average customer pays $200 a month and stays a year, you earn $2,400 and the program is cheap. If they pay $30 a month and stay four months, it loses money on every sale no matter how many partners you sign. That brings us to the next number.

How long until a customer pays back?

Payback is the months of gross margin it takes to earn back the cost of acquiring the customer. Using the same example, if a customer yields $150 a month in margin, $300 is paid back in two months. If margin is $20 a month, it takes fifteen, and with a typical churn rate many customers will leave before that.

Payback matters most when you pay commissions for a long time. A recurring commission of 20% for 24 months on a customer who churns after six costs you less than you planned; the same arrangement on a customer who stays three years costs more. If you are choosing a commission structure, how to set commission rates partners respect goes through the math in detail.

How concentrated is the revenue?

Rank partners by the revenue they brought in over the last quarter, and look at the share held by the top one, the top three, and the top ten percent.

If one partner produces 60% of the program's revenue, you do not have a program, you have a relationship with one company that has a program wrapped around it. That is fine to know and dangerous not to. If that partner shifts focus, renegotiates or goes quiet, the quarter changes overnight.

Some concentration is normal. Partner programs usually have a long tail where a few partners do most of the work, and the right response is not to force the numbers to even out. It is to treat your top partners as accounts, with their own check-ins, and to keep recruiting enough that a loss would hurt rather than end the program.

How many partners are actually active?

Total partner count is the most misleading number in this business. A program with 400 partners and 15 who have sold anything in the last 90 days is a program with 15 partners.

Define "active" once and write it down: for example, at least one conversion in the last 90 days, or at least one click in the last 30. Report the active count, the share of signed partners it represents, and how it moves. Activation also deserves a number of its own: of the partners who joined three months ago, what share have a first sale? What should happen in a partner's first 14 days is about moving that number.

Pricing matters here as well. Plans are often tied to active partners rather than signed ones, so a long tail of dead accounts is also a cost question.

How long from signup to first sale?

Time to first conversion tells you more about your onboarding than any survey. If the median partner who ever sells does so within 30 days, then a partner at day 90 with nothing is probably not coming. If it is 120 days, you have a long cycle and patience is warranted.

Segment it by how the partner came to you. Partners you recruited personally may behave very differently from partners who found you in a directory. That is useful, because it tells you where to spend recruiting time. It is also a small number once you split it, so do not read too much into a month when only six partners qualify.

Which refund and quality numbers should you watch?

Revenue that is later refunded or canceled is not revenue, and the problem is often concentrated in one or two partners. Calculate the refund rate and early cancellation rate per partner, then compare each to the program average.

Keep your sample sizes in mind. Three refunds out of ten sales is 30%, and tells you almost nothing. Wait until a partner has a few dozen conversions before treating the rate as a finding. Below that, treat it as a reason to ask a question.

A related quality number is the lead-to-customer rate for lead-based programs. A partner who sends 100 leads that produce 2 customers and one who sends 20 that produce 5 are very different partners, even though the first looks better on a volume ranking. If your lead flow is uneven, the common fraud patterns show which ratios deserve attention.

One more habit helps with all of these: write down your target before you open the report. If you expected a cost per customer near $250 and see $300, you have learned something. If you had no expectation, any number looks acceptable, and the review turns into reading rather than deciding.

Do you need separate numbers for each program?

If you run more than one program, yes. A referral program for customers and an affiliate program for publishers behave nothing alike, and blending them hides both. Customer referrals usually have a high conversion rate on a small audience; content partners have the reverse. Cost per customer averaged across the two describes neither.

The same goes for deal registration and lead pass motions. Their cycle times are measured in months, so a monthly review will show deals created long before it shows revenue. For those, track pipeline created, pipeline won and the time from registration to close, and give them a quarter before judging. Not sure which motion fits? Affiliate, referral, lead pass or deal registration lays out how they differ.

Splitting also protects you from a common misread. If the blended number improves because a cheap, high-volume program grew while the expensive one stagnated, the average hides a problem. Compare each program with its own history, not with the others.

What if the numbers look bad?

Resist the urge to change everything at once. Work through the likely causes in order of how cheap they are to fix.

If activation is low, the problem is onboarding or recruiting the wrong people. Look at what the partners who did sell have in common, and what the silent ones were told when they joined.

If cost per customer is high but concentrated in a few partners, look at their commission terms: a rate set for one product may be too generous for another. Tiers, product-level conditions and limits exist for exactly that. If it is high everywhere, the issue is the rate or the product's margin, and a program may not be the right channel yet.

If refunds cluster, tighten what you pay for and when. If one partner dominates, recruit. And if the program is small, say 10 active partners and a few sales a month, accept that the numbers are noisy. A month's change of one sale is not a trend. Wait for a quarter before concluding anything.

Whatever you change, change one thing and note the date, so that three months later you can tell what moved the number.

Which numbers can you safely ignore?

Raw clicks. They are useful for diagnosing one partner and useless as a program score. Conversion rate from click to sale, by contrast, tells you something about the traffic and your landing page.

Total commissions earned, as an achievement. Higher commissions mean the program works only if revenue rose by more. A program that doubled commissions and tripled revenue is healthy; one that doubled both without growing margin is not.

Leaderboards of partners by sales alone, unless you intend to build the program around a few partners. Ranking by raw sales rewards size, and size isn't the same as efficiency.

Vanity totals such as lifetime partners signed or lifetime clicks. They only go up.

What does attribution do to every number above?

Every metric in this post depends on partners getting credit for the right customers. If attribution is wrong, the numbers are wrong in a particular direction: usually in favor of whoever touched the customer last. A partner credited with a customer you would have won anyway makes your cost per customer look better for the partner and worse for you.

There is no way to get attribution perfectly right, but you can know where it is shaky. Coupon codes shared on deal sites, two partners claiming the same buyer, and traffic that browsers stop tracking all push the numbers one way or another. What happens when a customer clicks two partner links explains the credit rules you can set.

Judge the program on the totals as much as on partner-level credit. If revenue from partners rose 20%, did overall new revenue rise? A program can look great on its own dashboard while moving credit around inside a company that would have grown anyway. That is the question a partner-level report cannot answer, and it is the one the finance side will ask.

What should the monthly review look like?

Keep it to one page and the same page every month, so you can see trends. Cost per customer. Payback. Active partners and activation rate. Top-partner share. Refund rate, by exception. Then three questions: which partner should we talk to this week, which should we stop spending time on, and what one change would we test next month?

Inside Affiliate Tabs, the program overview, partner performance and cost of program reports cover most of this, and CSV exports cover the rest. If you are small, a spreadsheet with those rows is enough. The discipline of looking at the same page monthly matters more than the tool that produces it.