Running Your Own Affiliate Program: The Merchant’s Guide to Affiliate Marketing

Almost every guide to affiliate marketing is written for the affiliate. Pick a niche, build an audience, get links, earn commission. That is a perfectly good article and it is not the one you need if you are the one paying the commission.

The merchant side has different questions with different answers. What can I afford to pay? How do I know the sale was really theirs? What stops someone claiming commission on a customer who was already checking out? What happens when a coupon site inserts itself into every transaction I was going to get anyway?

This is that guide.

In-house or a network

The first decision, and the one most write-ups get wrong by framing it as a cost comparison.

What you can afford to pay

Commission rates get copied from competitors, which is how programs end up unprofitable. The number comes out of your own margin arithmetic.

Work from contribution margin, not revenue.

  Sale price                        100.00
  - cost of goods / delivery         30.00
  - payment processing                2.90
  - support and fulfilment cost       8.00
  ------------------------------------------
  Contribution margin                59.10

Now decide what share of that margin you will trade
for a sale you would not otherwise have had.

  20% of margin  → commission  11.80  (11.8% of sale price)
  30% of margin  → commission  17.70  (17.7% of sale price)

The two figures people confuse:
  "20% commission" usually means 20% OF SALE PRICE
  which here is 20.00 — a third of your entire margin.

For recurring products, work from expected lifetime value
minus servicing cost, not from the first payment. And decide
in advance whether commission is paid on renewals, because
that single clause changes the economics completely.

Two structural decisions follow, and both are easier to set now than to change later.

• First payment only, or recurring. Recurring commission is attractive to affiliates and it compounds against you forever on customers who would have renewed regardless. A common middle position is recurring for a fixed period — twelve months, say — then it stops.
• Flat or tiered. Tiering rewards volume and adds administrative complexity. For most small programs a single flat rate is correct, because the affiliates who would reach a higher tier are the ones you should be negotiating with individually anyway.

How tracking actually works, and where it breaks

This is the part general affiliate guides skip entirely, and it is the part that determines whether your payouts are correct.

The standard mechanism:

  1. Visitor clicks yoursite.com/?ref=jane
  2. You set a first-party cookie: ref=jane, expires in N days
  3. Visitor buys, at some point inside that window
  4. At checkout you read the cookie and attribute the sale

Where it breaks, in rough order of how much money it costs:

  Cookie cleared or blocked      → attributed to nobody
  Different device               → phone click, desktop purchase, lost
  Long consideration cycle       → cookie expired before they bought
  Another affiliate link clicked → last click overwrites first
  Ad blocker or privacy browser  → tracking may never fire
  Checkout on a different domain → cookie not readable there

The consequence worth internalising: your tracking systematically undercounts affiliate contribution. Affiliates know this and will tell you about it, frequently while asking for a longer cookie window. They are not wrong. The question is how much of an unmeasurable gap you are willing to pay for.

Cookie window and attribution model

Terms that prevent the predictable problems

Your affiliate agreement is the only mechanism you have for controlling behaviour after someone joins. Six clauses that address the things that actually go wrong.

1. Brand bidding. State whether affiliates may bid on your brand name in paid search. If you do not, someone will, and you will end up paying commission to appear above your own organic result while also paying for the click.
2. Coupon and discount conduct. Whether affiliates may publish codes, whether they may create codes, and whether commission is payable on orders using a code they did not issue.
3. Self-referral. Explicitly prohibited, and enforced technically as well as contractually.
4. Cooling-off and clawback. Commission approved after the refund window closes, not on order. Otherwise you pay out on orders that get refunded.
5. Disclosure requirement. Affiliates must disclose the relationship. This is not politeness — regulators hold advertisers responsible for how their affiliates promote them.
6. Termination and unpaid balance. What happens to accrued commission when you remove someone. Vague terms here produce disputes and occasionally public ones.

Clause five carries real legal weight and deserves its own treatment, which it gets in affiliate compliance: what you are actually responsible for.

Paying people, which is more annoying than it sounds

Running in-house means you handle payouts. Decisions to make before your first affiliate earns anything.

• Minimum payout threshold. Paying out £4 costs more in fees and admin than it is worth. Fifty is a common floor.
• Schedule. Monthly, on a fixed date, after the refund window. Predictability matters more to affiliates than speed.
• Method and who absorbs the fee. International transfers are expensive. State in the terms whether the fee comes out of the commission.
• Tax paperwork. Commission is income to the recipient. Depending on your jurisdiction and theirs, you may have reporting obligations, and affiliates may need to invoice you with VAT. Ask your accountant before the first payment, not after the first tax year.
• Records. Keep per-affiliate, per-order records. You will need them for a dispute and possibly for an audit.

Recruiting the first ten, which is the whole game

Programs fail at recruitment, not at technology. A perfectly configured program with four inactive affiliates produces nothing.

The sources that actually work for a new in-house program, in order:

1. Existing happy customers. They already use it, already recommend it, and converting an advocate into an affiliate is a single email. This is the highest-yield source and the most neglected.
2. People who already mention you. Search your brand name. Anyone who wrote about you without being asked is a warm lead.
3. Integration and service partners. Agencies, consultants and complementary tools whose clients need what you sell.
4. Affiliates of adjacent, non-competing products. Findable by looking at who links to products your customers also buy.
5. Inbound, from a proper program page. Slow, and it compounds. Publish commission rate, cookie window and terms openly — programs that hide the rate get fewer serious applicants.

What does not work: buying lists, mass outreach to anyone with a blog, and expecting a directory listing to produce partners. The distribution of affiliate performance is extremely skewed — a small number will produce almost everything, and finding them is relationship work rather than volume work.

When you should not run a program at all

Written by a site that gives away affiliate software, so weigh it accordingly — and I would still tell you this before you install anything.

• Your margin cannot absorb it. If a meaningful commission takes you to break-even, an affiliate program converts profitable sales into unprofitable ones. Run the arithmetic above first.
• Nobody will own it. An affiliate program is a channel with partners, not a feature you switch on. Unmanaged, it attracts coupon and brand-bidding traffic because those are the affiliates who need no cultivation.
• Your product has a long, consultative sale. Attribution over a six-month cycle with multiple stakeholders is close to meaningless, and a referral or partner agreement fits better than commission-per-click.
• You have no customers yet. Affiliates promote things that convert. Asking people to send traffic to an unproven page wastes their effort and your credibility.
• You are in a regulated category where you are liable for how others advertise you. Doable, and it needs the compliance work in place before launch rather than after.

A launch sequence that works

1. Calculate the rate from contribution margin. Write down the number and the reasoning.
2. Write the terms, covering the six clauses above. This is a document, not a checkbox.
3. Install tracking and test it properly — click a link, clear cookies, buy on another device, apply a coupon, request a refund. Confirm each case attributes the way you intended.
4. Recruit ten people you already know before publishing a program page.
5. Give them assets that work — accurate copy, current screenshots, a landing page that converts. Affiliates promote what makes them money.
6. Review monthly. Which affiliates produce, what the refund rate looks like per affiliate, and whether anyone is bidding on your brand.

Step three is the one people skip and the one that produces angry emails in month two. Test the refund case in particular — a program that pays commission on orders that get refunded is a program that funds returns fraud.

My verdict

An in-house affiliate program is genuinely worth running when you already have people who would recommend you, margin that can absorb a commission, and one person willing to own it. Under those conditions it is one of the few channels where you pay only for outcomes.

It is a poor idea as a growth tactic bolted on because a competitor has one. The failure mode is not dramatic — the program simply fills up with coupon sites and brand bidders, and twelve months later somebody works out that most of the commission went on sales you would have made anyway.

Decide the rate from your margin, write terms that address coupons and brand bidding before launch, test the refund path, and recruit ten people you can name. That is the whole job, and the software is the easiest part of it.

What you are legally responsible for once affiliates start promoting you is covered in affiliate compliance and when you actually need software for it.

Frequently asked questions