Do not start an affiliate program yet

By Jimi Barkway · Published 1 September 2026 · Updated 3 September 2026 · Part of the honest manual

Every vendor in this category makes money when you start an affiliate program. We're one of them. That's why all the advice starts at "here's how" and never at "should you". And it's why this module is step zero, the one that's allowed to say no. Four questions. A no on any of them means wait. The complete guide puts this decision first and the other seven in order.

Does anyone recommend you without being paid?

This is the whole test in one question. A commission pays for word of mouth that already wants to happen. It can't make any. Say a few customers already share your product in a Slack group or a newsletter, with no reward attached. Then a program pours fuel on a fire that exists. If nobody does, you're asking strangers to stake their name on a product. A product its own users stay quiet about. Almost nobody takes that trade at 20%.

You can check this in an afternoon. Look through your signups for customers you never went out and got. No ad click, no cold email, no launch spike. Ask three of your best customers how they found you. If the answer involves a person, you've got the raw material. Say every customer traces back to something you paid for or posted yourself. Then the honest next move is to make the product easier to recommend. A system for recommendations that aren't happening can wait.

Can your margin actually carry a commission?

Do this on paper before you touch any software. It's arithmetic, not strategy. Take your monthly price. Knock off payment processing, hosting and the support cost of one customer. Look at what's left. A commission comes out of that remainder every month, for as long as you pay it. The cost article prices all four lines of it.

Two versions of the same product make the point. A $49 plan at 80% gross margin has about $39 a month of room. So a 20% commission of $9.80 leaves the unit clearly profitable. The only question is how long the customer stays. A $9 plan with a heavy support load has maybe $5 of room. The same 20% takes $1.80 of it. Meanwhile the fixed cost stays exactly the same: managing the partner, answering their emails and paying them out. Busy, and still losing money.

Churn is the multiplier people miss. A recurring commission on a customer who stays three years is a bargain. At 10% monthly churn, the same commission pays partners for customers who mostly vanish before their second invoice. Partners notice. Their earnings graph tells them your retention before you do.

Is there an audience someone else can reach for you?

An affiliate is a person with an audience shaped like your buyer. For a program to work, those people have to exist. Newsletters your customers read, YouTube channels they watch, consultants they hire, communities they argue in. For most B2B SaaS they do. But say your buyer is one procurement lead inside each of two hundred hospital trusts. There is no creator whose audience is those people. No commission conjures one.

The test is concrete. Write down where your last ten customers spent their attention before they found you. If the same few names keep appearing, that list is your future program. If it's empty, or every entry says "Google", your channel is search or sales. An affiliate program would sit there unused while you felt guilty about it.

Can you name ten people to invite?

Programs don't fill themselves. The directories are a trickle. And "affiliates find you" is something that happens to companies with traffic, years in. Your first partners will be people you invite by name. Customers who already refer, writers who've covered you, hosts of podcasts you've been on, friends with relevant lists. Ten names is the floor. In the median program only around 15% of approved affiliates ever produce a referral. So ten is where you start, not where you end. The referral comparison sets the two side by side on cost and tax.

If you sit down and can't get past three, the program has no launch cohort. And an empty program is worse than none. It shows every visitor a door with nobody behind it.

When to come back

The threshold is lower than the textbooks suggest. It isn't a revenue number. Come back when all four answers are yes. A couple of customers arrived by real word of mouth. The commission arithmetic clears with room to spare. You can point at where your buyers' attention lives. And ten names are on a list. For some of you that's true today. For others it's six months of product work away. Starting now would only record that fact in a dashboard.

What do founders ask before starting?

This module ends without a sales pitch, on purpose. If your answers were no, the most useful thing we can do is not sell you anything. When they turn yes, the rest of the manual is waiting, starting with the money.