Most affiliate programs have a queue nobody designed. A form sits on a page, applications pile up, and the founder opens it every few weeks, feels a wave of guilt, and approves everyone at once because reading twenty applications is nobody's idea of a Tuesday.
That queue is where a program's economics get decided, and it is almost always the least deliberate part of the machine. The commission rate got three meetings. The approval policy got none.
This post is about the decision: what a status change controls, what to look at before approving someone, how to reject without being a ghost, and why removal should feel like a pause button.
TL;DR
- There are two doors into a program, and they behave differently. A direct join link creates an active affiliate instantly. An application through the public directory arrives as pending and tracks nothing until you approve it.
- Four statuses decide everything: active, pending, rejected, removed. Only active earns, and only active or pending balances count toward a payout.
- Approve on evidence of distribution, not on follower counts you cannot verify. A live URL and a real list beat any claim typed into a form.
- Rejection is silent in most tools, including ours. That silence is a policy choice, so write the one email that breaks it.
- Removal is reversible and should stay reversible. It stops tracking and blocks the same email from rejoining, while leaving earned money and history intact.
Two doors into the same program
Every hosted affiliate program ends up with two entry points, whether or not the owner thinks of them that way.
The first is a direct join link, something like usesled.com/yourslug/join. Someone fills in a name, hits submit, and they are in: an active affiliate with a unique code, a welcome email carrying their link, and tracking that works immediately. There is no approval step on this path by design. You invited them or you published the link on your own site, so the gate is the link itself.
The second is the public directory at https://usesled.com/affiliate/discover. An affiliate browsing programs there can apply to yours, and that application is deliberately different: it lands as a pending affiliate. A code is generated, but it does not track clicks or conversions until you approve it. Nothing about it can earn. The applicant sees an applied state in their cross-program dashboard, and you see a pending row with approve and reject buttons.
The launch guide covers how these two doors fit into setting a program up. The difference between them is the only reason the distinction matters. A pending affiliate is not a slower active affiliate, it is a record of intent with no effect on revenue, which is what you want from a stranger who found you in a list.
What the four statuses actually control
Statuses are not labels for your own bookkeeping. Each one gates a specific piece of behavior, and knowing which piece is what tells you how much thought the decision deserves.
| Status | Tracks clicks and conversions | Affiliate sees a program dashboard | Counts toward payout | How it is reached |
|---|---|---|---|---|
| active | yes | full, with link and share kit | yes | direct join link, or an approved application |
| pending | no | applied state in the unified dashboard | yes, balance is payable | applied through the directory, awaiting review |
| rejected | no | no | no | you rejected the application |
| removed | no | no | yes, earned balance is preserved | you removed an active affiliate |
Two of those rows need explaining, because they surprise people.
Pending is payable. If an affiliate applied, you approved them, they earned, and then a pending state appears for some unrelated reason, the money is not stuck: paid-out calculations include both active and pending rows, so a payout run does not quietly drop someone mid-transition. What a pending affiliate cannot do is generate new conversions.
Removed keeps history. Removal is a soft delete: the row stays, the conversions stay, the payouts stay, and the affiliate disappears from your active list and from their own program dashboard. Their tracking link stops recording. Their email is blocked from rejoining through the join form, which is what makes removal an enforcement action rather than a suggestion.

What to check before you approve
Most applications contain almost no signal. Name, email, sometimes a website, sometimes a sentence about how they promote. You are not running a background check, you are looking for one reason to say yes and one reason to wait.
The checks that pay for themselves, in order of value per minute:
Does the email match a customer? If the applicant already pays you, they are the best affiliate you can approve: they know the product, they have a reason to talk about it, and their recommendation is credible. Check the customer list before the form.
Is there a live thing to look at? A newsletter archive, a channel with uploads from this year, a blog with recent posts, a template gallery. Not a link tree, not a course sales page. Something with an audience and a publishing cadence.
Does the audience overlap your buyer? An affiliate with 40,000 followers who sell to hobbyists will produce clicks and no conversions, and then both of you conclude affiliate marketing does not work. Audience fit beats audience size every time.
Is the channel compatible with your billing model? Someone who promotes one-time purchases will underperform on an annual-only plan unless they can explain how they sell a year up front. This is the most common mismatch we see in applications for developer tools.
Can you pay them at all? Payout method and country. If their region only supports a method you disabled, approving them creates a support ticket when they hit the minimum.
Have they applied before? A rejected applicant who reapplies after building something is a good sign. The same address hitting five programs in a week is not.
The ten minute approval pass
Batching beats reacting. Applications that sit for three weeks convert worse than ones answered in two days, because enthusiasm has a half life and the applicant has forgotten the program by the time you reply.
A pass that takes ten minutes, twice a week, is enough for most programs under a hundred affiliates:
- Open the pending list and read only the "how will you promote" field. Skip the rest on the first read.
- Open the website for anyone whose answer was not an instant no. Ten seconds each.
- Search your customer list for the email. Approve those immediately.
- Approve anyone with a live channel that fits the buyer. Reject the rest with the template below.
- Stop when the list is empty, even if it took four minutes. Do not go looking for more applications.

What you buy with that cadence is a bounded decision, not diligence. The real failure mode of approval queues is not approving the wrong affiliate, it is approving nobody, which makes the program look abandoned to exactly the people willing to work for commission.
Approval policy by stage
The right policy changes as the program grows, and most programs never update it. A rule that made sense at eight affiliates is expensive at two hundred.
| Program stage | Default policy | Review depth | What goes wrong if you get it wrong |
|---|---|---|---|
| 0 to 25 affiliates | approve nearly everyone | 30 seconds, link only | a few dormant rows, which cost nothing |
| 25 to 100 | approve on channel evidence | 2 minutes, site plus audience fit | coupon and brand bidding affiliates slip in |
| 100 and up | approve on track record | 5 minutes, past programs and disclosure | your best affiliates compete with link farms |
| Directory heavy | answer within 48 hours | 90 seconds, triage only | listings look dead, applications stop |
| High average order value | manual, documented | 10 minutes, traffic source and claims | one bad claim becomes a support issue |
| Regulated vertical | verify before approving | read their existing disclosures | you inherit compliance risk you did not price |
The first row is the one founders get wrong, and wrong in the direction of caution. With eight affiliates, a bad approval costs a database row and maybe a conversion you claw back. A bad rejection costs a distribution shot you cannot replace. Approve generously early, tighten once you know which channels convert for your product.
The last two rows are the opposite failure. If a wrong claim about your product can cost you money, a licence or a customer, the review is insurance rather than politeness, and the program terms template is where the rules those claims must respect live.
Rejecting without ghosting
Here is a detail worth stating plainly, because it is invisible unless you build it: in Sled, rejecting an application sets the status to rejected and sends nothing. No email, no notification. The applicant simply stops seeing the program in their applied list.
That is a defensible default for a tool, since automated rejection emails are easy to get wrong and easy to make cruel, but it is a bad default for a program owner. The person you just rejected is usually a customer or a newsletter reader, or someone who will apply again in six months with a real audience. Silence is the only part of the interaction they will remember.
One template, three lines, sent from your own inbox:
Thanks for applying. I am not opening the program to new affiliates this month while we tighten attribution, and I would rather not promise you a commission I cannot track properly. If you are still interested in a few weeks, reply to this email and I will fast track it.
Three properties make that work. It says no. It gives a reason that is about you, not about them. It leaves a door open with a specific action attached, which turns a rejection into a warm contact you can recruit later. Do not send a rejection that implies a judgement about their audience. You do not know enough about them to make one, and that sentence will be screenshotted.
Removal is a pause, not a firing
Removal gets treated as the nuclear option and then used never, which is how programs end up with an affiliate bidding on the brand name for eighteen months while the owner refreshes a dashboard.
In Sled, removal does four things: sets the row to removed, stops the tracking link from recording new clicks, hides the affiliate from your active list and from their own dashboard, and blocks that email from signing up again through the join form. It does not delete conversions, and it does not cancel commission already earned. A restore action puts the affiliate back to active with the same code, so the pause framing is literal.
Use it for conduct, not performance:
- Bidding on your brand in paid search when the terms forbid it
- Cookie stuffing, click farms, or self-referral through a second address
- Promoting with claims you have asked them twice to stop making
- Abusive messages to you or to customers
Do not use it for an affiliate who has not sold anything yet. That is what the nudge is for, and the post on why affiliates go quiet covers the sequence. Someone who has not shared a single link has not done anything wrong; they have not started.
One operational rule: pay what is owed before you remove, or remove and pay in the same run. Removal is not a reason to hold back earned commission, and an affiliate who loses money on the way out does not stay quiet about it in the communities where you recruit.
Per-affiliate rates belong in the approval step
Set a non-standard rate at approval, while the reason for the exception is still in front of you. Approving at the default rate and promising a better deal three weeks later is how a program ends up with an affiliate earning a number nobody can explain during an audit.
Sled supports a per-affiliate override on top of the program default: the newsletter with a fixed placement fee, the agency bringing three clients, the creator whose list is small and unusually on point. Set it at approval and make the terms say the owner can set individual rates. Defaults for a new program are 30 percent, a 90 day cookie and a $20 minimum payout.
Then check the arithmetic against refunds before you agree to anything. A 50 percent rate on a product with a 20 percent refund rate and a 10 percent processing cost is not a growth program, it is a hobby with a spreadsheet. The commission rate guide does the long form.
Two kinds of pending, and why mixing them ruins reports
Two unrelated things are called pending in affiliate reporting, and confusing them produces decisions that look reasonable and are wrong.
An affiliate can be pending, meaning the person exists but the link does not track. A conversion can be pending, meaning the money is real but not yet payable: the sale happened, the commission is calculated, and it clears when the refund window allows. One is a person waiting for approval, the other is revenue in flight.
The consequence shows up in your monthly numbers. Counting pending affiliates as affiliates makes activation rate look terrible, because you divide by people who were never allowed to promote. Counting pending conversions as revenue means paying commission on money that refunds next week. Keep the columns separate, and reconcile refunds against the conversions they reverse. Sled does that with tombstone rows, so a refund arriving before the sale still lands, and the refund and clawback writeup explains why most providers send events out of order.
What to watch in the first two weeks after approval
Approval is not the end of the decision, it is the start of a short observation window. The first fourteen days tell you what the application could not: whether this person actually promotes.
Track three things per new affiliate: did they share the link, did that produce clicks, did any click convert. Shares with no clicks means a broken link or a bad placement. Clicks with no conversions means an audience mismatch. Neither means they have not started, so one nudge is appropriate, not three.
Sled groups silent affiliates, joined with no shares and no conversions, and the nudge throttles each address to once a week. Two nudges over a month is the polite ceiling. After that the correct action is nothing until they come back.
The fraud signals only appear once money is moving, which is why approval cannot screen for them: conversions clustered on one IP range, a customer email matching the affiliate's second address, refunds concentrated on one affiliate's traffic. Self-referral and disposable email blocking are on by default, which removes the laziest cases, but the fraud tactics list is worth reading before your first payout run rather than after.
Cadence and a rejection template matter more than any single call: the queue's real failure is drift, not judgement.
FAQ
Should I approve every affiliate who applies? Early on, close to yes. With fewer than twenty five affiliates the cost of a bad approval is a dormant row, while the cost of a bad rejection is losing one of the few distribution shots you have. Tighten the policy once you know which channels convert for your product.
How long should an application sit in pending? Under 48 hours if the program is listed in the directory, since the listing keeps showing it to applicants while an unanswered queue makes it look abandoned. A direct join link has no queue at all: the affiliate is active the moment they submit, which is why the link should only be published where you are comfortable with that.
Can a removed affiliate rejoin? Not through the join form. Removal blocks the same email from signing up again and shows a clear error rather than a generic failure. Restoration is available to the owner and puts the affiliate back to active with their original code, so removal works as a pause for conduct issues rather than a permanent deletion.
Do pending affiliates get paid for the clicks they send? A pending affiliate's link records no clicks and no conversions, so there is nothing to pay. Once approved, tracking starts from that moment and does not backfill earlier traffic. An existing balance stays payable, which is why payout runs include both active and pending rows.
If you would rather spend the ten minute pass approving affiliates than reconciling spreadsheets, Sled keeps applications, statuses, conversions and payouts in one dashboard, and a directory listing costs nothing extra. Start a program and the first decision is already waiting.