Choosing an affiliate network or agency: spot shaving in your data, test partners with controlled traffic, read payment terms, and compare offers properly.
You send 4,000 clicks to a new offer. The partner's dashboard shows 3,870 clicks, 212 registrations and 58 first-time deposits. Every number looks fine. Three months later you move the same traffic to a different partner running the same advertiser, and the deposit count goes up by a fifth on identical volume. Nothing about the first dashboard ever looked wrong. That is the whole problem with choosing an affiliate network: the failure mode is quiet, it is gradual, and by the time it is obvious you have already paid for it with months of media budget.
Shaving is the word affiliates use for a partner withholding conversions or revenue that belongs to you. It is real, it is not universal, and it is not the only way a partnership costs you money. Bad payment terms, an undefined revenue base, a manager who never answers, and tracking you cannot verify all take a similar bite without anyone doing anything technically dishonest.
This is a working method for evaluating a network or an agency before you commit volume. What each layer between you and the advertiser adds and takes, the warning signs that show up before you send a click, how to test for a shave using controlled traffic, what to ask a manager, and why the headline percentage on an offer tells you almost nothing about what you will be paid.
Every affiliate deal sits at one of three distances from the money. Understanding which one you are in explains most of what happens to your payouts.
Direct with the advertiser. You have a contract with the brand itself. No intermediary, no shared attribution, the shortest possible reporting chain. In exchange you carry all the overhead: your own tracking, your own negotiation, your own chasing when an invoice is late, and a separate integration for every brand you run. Advertisers also apply their own minimums, and most will not open a direct account for an affiliate with no history.
Through a network. The network aggregates hundreds of advertisers, gives you one dashboard, one payment, and self-service access to offers you could not open individually. The trade is that the network sits between you and the conversion record, takes a margin you usually cannot see, and can be several hops from the advertiser itself. Offers on a network are sometimes resold from another network, which is where reporting drift starts.
Through an agency. An agency works with a smaller set of advertisers and platforms, but works them harder: negotiated caps, custom payout bumps, and usually operational support in the form of tracking setup, creative and landing pages. Fewer offers, deeper relationships, more human involvement. The trade is that you are dependent on their judgment about which offers to put in front of you.
| Direct advertiser | Network | Agency | |
|---|---|---|---|
| Offer breadth | One brand | Very wide | Curated, narrower |
| Payout per conversion | Highest gross | Lower, margin taken | Negotiated, often above network |
| Reporting distance | One hop | Two or more hops | Usually one or two |
| Infrastructure you provide | All of it | Most of it | Little to none |
| Access barrier | High | Low | Medium, relationship-based |
| Who fixes broken tracking | You | Support ticket | Named manager |
| Negotiating leverage | Yours alone | Pooled but impersonal | Pooled and represented |
The useful question is not which layer is best. It is which layer you need right now. An affiliate with a proven funnel, in-house tracking and enough volume to hit an advertiser's minimum should go direct on their top two offers and stay diversified everywhere else. An affiliate with good traffic and no infrastructure loses money going direct, because the effective payout after the time spent building and maintaining tracking is worse than a supported deal.
Every hop between your click and the advertiser's database is a place where a conversion can be lost, delayed, deduplicated or reattributed. Most of that loss is not malicious. Redirect chains drop mobile users. Attribution windows differ between the platform and the brand. Deduplication rules quietly kill a second conversion from the same user. A network that resells an offer inherits every one of the upstream network's quirks and adds its own.
The margin question is separate. Networks and agencies both take a cut; that is the business. What matters is whether the cut is fixed and disclosed or floating and hidden. A partner who says "the advertiser pays 65 and we pass you 50" is describing a fixed margin. A partner who will not discuss the upstream number at all is telling you the margin can move whenever their own margin needs to move.
Agencies add something networks structurally cannot: someone whose job is your account specifically. That is worth real money when a postback breaks on a Saturday, when an advertiser starts rejecting conversions from a geo, or when you need a cap raised before a launch. It is worth nothing if the "manager" is a shared inbox with a first name attached.
You can disqualify most bad partners in a twenty-minute conversation, before any traffic moves.
None of these is proof of anything. Two or three together, on a partner asking for volume, is enough to walk away.
The mental image most affiliates have — conversions simply vanishing — is the rarest form. What you actually see is subtler.
A persistent gap at one funnel step. Your click count matches within a few percent. Your registration count matches. But the registration-to-deposit rate sits consistently below what the same creative and geo produce elsewhere. One step is wrong and the rest are clean, because the step being adjusted is the one that costs money.
A cap that behaves like a filter. Conversions track normally until you cross some daily volume, then the rate softens. Sometimes this is genuine — advertisers do apply quality filters at scale. Sometimes the filter is arithmetic.
Quality rejections that arrive late and cluster. A batch of conversions reversed three weeks after the fact, always from your best-performing source, with a reason code that cannot be checked.
Revenue per depositor that decays faster than it should. On revshare, shaving does not need to touch conversion counts at all. Reducing reported net revenue per player by a small amount every month is invisible unless you know what the same players are worth elsewhere. This is also where negative carryover gets confused with shaving — a carried-forward negative balance is a disclosed contract term, not a hidden deduction, and you should know which one you are looking at before you accuse anyone.
The common thread: shaving produces a number that is plausible. If the numbers looked absurd, nobody would run the scheme. Which means your gut is useless here and you need a comparison against something the partner does not control.
Three tests, in order of cost and reliability.
Register accounts yourself, or through people you trust, each with a distinct sub-ID, spread across several days and devices. Complete the full qualifying action — not just a signup, but the deposit or the trade or whatever the payout event actually is. Then check whether all of them appear in reporting, and how long they take to show.
The value is that missing seeds are unambiguous. The limit is sample size. If you seed ten conversions and nine appear, you have observed a 10 percent gap on a sample far too small to distinguish from a genuine tracking failure. Seeds catch gross shaving and integration bugs. They will not detect a 7 percent trim.
Fire an event from your own landing page or pre-lander at each step you can observe, and compare shapes. You cannot see the advertiser's database, but you can see how many people clicked out, and often how many reached a confirmation state. A properly configured S2S postback into your own tracker gives you a record that exists independently of the partner's dashboard export.
Expect a natural gap between your outbound clicks and the partner's inbound clicks. Redirect drop-off, bot filtering and users who cancel navigation are real. Somewhere in the low single digits is normal. A double-digit click gap needs an explanation, and the explanation should be specific.
The strongest test available to an affiliate. Take one advertiser offer available through two partners, and split identical traffic between them: same creative, same geo, same days, same devices, alternating by click so time-of-day effects cancel out.
Here is what a real read looks like. You push 12,000 clicks, 6,000 to each partner.
| Partner A | Partner B | |
|---|---|---|
| Clicks | 6,000 | 6,000 |
| Registrations | 402 (6.7%) | 398 (6.6%) |
| First-time deposits | 121 (30.1% of regs) | 96 (24.1% of regs) |
Registration rates match almost exactly, which tells you both partners are counting clicks and signups the same way. The divergence appears only at the deposit step. Pooled, 217 deposits from 800 registrations is 27.1 percent, so each side would be expected to produce about 108. The standard deviation of a count like that is roughly the square root of 400 × 0.271 × 0.729, which is about 8.9. Each side sits around 1.4 standard deviations from the pooled expectation.
That is suggestive. It is not proof. A gap that size will occur by chance more often than people assume, and Partner B may simply be a further hop from the advertiser with genuine attribution loss. Run it again the following cycle. If the same pattern reappears at the same step, you now have something you can take to a manager — or act on without a conversation.
The cost of being wrong is worth pricing. Twenty-five missing deposits, at an average of say 180 dollars of net revenue per depositor in the first month and a 50 percent share, is 2,250 dollars in month one alone, before you count the tail of a revshare cohort. That is what a quiet 6-point gap is worth.
The percentage is applied to a base. The base is defined by whoever wrote the agreement. Two deals with different percentages are frequently not comparable at all.
Take a player who generates 10,000 dollars of net gaming revenue in a month, on 20,000 dollars of deposits.
Deal A: 60 percent revenue share. The contract defines the base as net revenue after bonus costs, payment processing, platform fees and administrative charges.
Deal B: 50 percent revenue share. The contract deducts bonus costs only.
The lower percentage pays 260 dollars more on the same player, in the same month, on the same activity. Now extend it. If Deal A carries a negative balance forward and Deal B resets monthly, a player who wins 4,000 dollars next month leaves you with a 4,000 dollar hole to climb out of under A and a clean slate under B. Over a six-month cohort the two deals are not remotely equivalent, and the 60 was never the better number.
The same logic applies when you are weighing revenue share against a one-time CPA. The comparison only works once both sides are expressed in the same units: expected revenue per click, over a defined window, net of every deduction.
So the first question about any offer is never "what is the percentage". It is: what exactly is deducted before my share is calculated, and can you show me a sample statement?
Ask these in writing, and keep the answers.
Question 12 catches more bad agreements than any of the others. A revshare deal that terminates your tail when you stop sending new traffic converts every existing player into an asset you no longer own.
A net period is not administration. It is you lending money to your partner at zero percent.
Run the arithmetic on a typical structure: monthly closing, net-30. You spend on media on 1 March. That day's earnings close on 1 April. Payment lands around 1 May. Your first day of spend is repaid roughly 61 days later. Your last day of spend that month, 31 March, is repaid about 31 days later. The average is about 46 days of float.
To run 8,000 dollars a month of media continuously under those terms, you need to carry roughly a month and a half of spend before revenue starts recycling — call it 12,000 dollars of working capital, plus a buffer for a late payment. Move to net-60 on the same closing structure and the float roughly doubles: about 20,000 to 24,000 dollars tied up to sustain the same 8,000 a month. Move to weekly settlement and the same campaign needs perhaps 2,000 to 3,000.
The partner offering 55 percent on net-45 and the partner offering 50 percent on weekly terms are not obviously ranked. If working capital is the thing limiting your scale, the weekly deal lets you run more volume, and more volume at 50 is worth more than less volume at 55. This is covered in more depth in our guide to affiliate payouts and payment rails, including minimum thresholds, reserves and the practical trade-offs of stablecoin settlement.
Minimum payout thresholds interact badly with long net terms. A 500 dollar minimum on monthly closing with net-30 means a small campaign can roll over for months. Ask what the minimum is before you assume you will be paid at all in your first quarter.
Optimising the headline number. Covered above. The percentage is the least informative field on the offer page.
Testing with real traffic and no instrumentation. If your only record of what happened is the partner's dashboard, you have no test. Set up independent tracking before the first click, not after the first suspicious month.
Concentrating everything in one partner too early. Exclusivity in exchange for a payout bump is a reasonable trade once you know a partner pays. Before that, it removes the comparison that would tell you whether they do.
Confusing a slow month with a shave. Seasonality, advertiser-side funnel changes, a new KYC step, a payment provider dropping a geo — all of these produce the same shape as shaving. Ask before you accuse. A good manager will tell you which one it was and usually knew before you did.
Ignoring the tail. On revshare, the money is in months three through twelve. An agreement that is generous in month one and silent about what happens when you pause traffic is worse than it looks.
Judging by dashboard quality. Beautiful reporting interfaces are cheap. Correct numbers are not.
You should be able to answer these without asking anyone: what your EPC is by source, what your registration-to-deposit rate is by geo, what proportion of conversions were reversed and why, whether every payment arrived on the stated date, and what a given cohort of depositors is worth in month three versus month one. If your partner's reporting does not let you build that picture, the relationship has a ceiling regardless of the payout terms. The KPIs worth tracking are the ones that let you compare partners on the same axis.
Good partners get boring. Payments arrive on the same date. Numbers reconcile. Your manager tells you about a problem before you notice it. That is the entire standard, and it is why choosing an affiliate network or agency is a decision worth re-making every quarter rather than once.
We work as an agency, which means fewer relationships handled more directly rather than a catalogue of offers you sort through alone. Partners get 50 percent revenue share, tracking infrastructure built and maintained for them rather than assembled from scratch, and sites and landing pages produced at no cost. Because we hold direct relationships with major platforms and advertisers across crypto, forex, iGaming, gaming, high-ticket and tech, caps and custom payout bumps are things we negotiate rather than things you accept. Every partner has a named human on Telegram, not a ticket queue — which matters most on the day a postback breaks.
If you want to see how the deal structure and reporting work in practice, the Telegram bot handles onboarding, and the community chat is where partners compare notes. You can also read more about how we work first.
You cannot know from one data set. Compare the partner's reported numbers against a source they do not control, such as your own landing page events, seeded test conversions, or the same offer run through a second partner with identical traffic. Shaving shows up as a persistent gap at one specific funnel step, not as random noise across all of them.
Neither is better by default. A network gives you breadth of offers and self-service speed. An agency gives you fewer, deeper relationships, negotiated caps and usually hands-on help with tracking and creative. If you already have a working funnel and want volume across many offers, a network fits. If you have traffic but no infrastructure, an agency usually pays more per click.
Most programs close a period monthly and pay somewhere between net-15 and net-45 after that close, so the money from your first day of spend can arrive up to two months later. Weekly or net-7 terms exist but are usually earned after you have a payment history. Treat anything longer than net-45 as a financing decision, not just an administrative detail.
Because the percentage is applied to a base that each partner defines. Bonuses, payment processing costs, platform fees, licence fees, chargebacks and administrative deductions can all be removed before your share is calculated. A 50 percent deal on a lightly deducted base routinely pays more than a 60 percent deal on a heavily deducted one.
Server-to-server postbacks are the only conversion signal that survives modern browsers. Here is how the plumbing works, how to test it, and how to find the conversions you are quietly losing.
Read postMost affiliates pick a payout model out of habit. Here is the arithmetic that should decide it: retention thresholds, cohort curves, capital velocity and the risks on both sides.
Read postTelegram audiences convert harder than almost any other free traffic source, and they are also the easiest to destroy. Here is how the economics actually work.
Read post