Four months in: what building an honest crypto rebate program has taught me

Posted by the operator of RebateMax

The first version of this site went up in May 2026, and the first article followed on 2 July. That makes it four and a half months old as I write this, which is too young for conclusions and old enough for a record. So this is not a victory lap and it is not a confession. It is a walk through what one decision, made before the first page existed, has turned into.

The decision was this: every number on the site would be one a reader could check, and the numbers that went against us would be published at the same size as the ones that flattered us. Everything else, the scoreboard, the receipts, the free signals, the research desk, the corrections page, grew out of that one commitment, usually because it forced my hand.

The industry runs on marketing math

If you have ever searched for an exchange referral code, you have seen the shape of it. A page promises "50% off fees". It does not tell you what the fee was, what it becomes, or how much of the affiliate's commission is quietly kept. The number is chosen to be large, not to be true.

I had the same commission available to me that every other affiliate has. BingX pays sixty points of the fees a referred trader generates. I could have returned twenty and called it fifty against an inflated schedule, and I would have kept four times as much per trader. The founding post walks through that arithmetic in full, and it has not changed: sixty comes in, fifty goes back to you as a permanent discount at the exchange level, ten stays here. 60/50/10. On LBank the deal is stated the same way: 0.003% taker from your first order, no volume ladder, against a standard 0.06%.

Stating the split as three integers was the founding act. What I did not understand in May was how much else it would oblige.

What one honest number forces you to build

Once the split is on the page, the reader has a right to ask the next question. If the fee math is real, is the rest real? So the signal channel could not be a highlight reel; it had to be a record. The record could not be marketing; it had to be graded against exchange data, with losses shown. The "is this legit" question could not be answered with reassurance; it had to be answered with checks a stranger can run.

That is how the site got its shape. The scoreboard grades every signal we forward against the exchange's own candles and publishes two accountings side by side, one that scales out at each target and one that holds every position to its stop or final target. Both lines are on the page whatever they show, because taking either one down would make the fee math a lie by association.

The receipts show entry, exit and the result on individual trades, the losing ones included, because a page of winners is not a record. The corrections policy exists because the moment you publish numbers you also publish your mistakes, and the only honest response to a mistake is to fix it in public and leave the trail.

None of this was in the original plan. Each piece appeared because the previous piece made it necessary. That is the strongest argument I have found for starting with one true number: it does not let you stop.

Publishing the losses

I want to be specific about what it feels like, because most people who consider doing this never do.

The first time the hold-to-exit line went negative, I looked at it for a long time. Every instinct says: change the accounting convention, widen the window, pick the metric that reads better. All of those are available, and readers would never know.

The line stayed. Since then it has moved, and it has mostly moved against us, and the page moves with it every night. I said in August that if both accountings went negative they would be published the same way, on the same page, at the same size. That promise still holds and I expect to be tested on it.

Here is the part nobody warns you about: after the first few weeks, publishing losses stops being brave and becomes ordinary. The page updates. The number is what it is. The relief of not having anything to hide is worth more than any single month of better-looking figures.

Correcting in public

In August, an audit of the full signal record found that a source channel had been reposting the same signal, and each repost had been counted as a separate trade. Nearly a third of the published scoreboard was double-counted.

The fix ran the day the audit finished. 481 rows were voided, the published count went from 917 trades to 654, and the scoreboard, the receipts and a blog post that had been live for hours were all republished with the lower numbers the same hour. Every voided row keeps its original outcome, so nothing was deleted, and the post that carried the wrong figures still carries a note quoting them.

Two lessons came out of that week. The first is that a careful grading process does not save you from a careless front door; the grading was right and the record was still wrong because the intake was wrong. The second is that the first correction proposed would have erased real signals along with the duplicates, and it was only caught because it was checked against the whole history before it ran. A correction can be its own incident. Check it like one.

Done means true, not "ran"

Most of this site runs on automation: the grading, the market desk, the alerts, the receipts, the announcements. Automation makes a record possible at this scale, and it carries one rule that took me longer to learn than it should have.

A job that runs is not the same as a thing that happened. A process can complete cleanly and still produce nothing, and the more of a site runs on automation, the easier it is to mistake a green tick for a result. So the standard here is not "the job ran". It is "the outcome is observed true in the world": the post was announced, the coin was measured, the signal reached the channel. Every process that publishes, grades or announces something has a line saying what must be true afterwards, and those lines are tested around the clock, independently of the processes they check. The memecoin desk measures a coin only after its contract address has been verified, never on a ticker symbol, because the same symbol exists on dozens of chains and the most liquid namesake is rarely the real one.

If you run anything on automation, it is the question worth asking of every step: does it check that the step ran, or that the thing happened? The second question is the only one a reader cares about.

What honesty costs

I will not pretend the honest playbook is the fast one. It is not.

A page that says "50% off" draws more clicks than a page that says "sixty, fifty, ten, here is the schedule". A signal channel that only posts winners grows faster than one that grades everything. A site that hides its corrections looks more polished than one with a corrections page. Fewer people arrive here than would arrive at the marketing version of this site, and I knew that going in.

What I believe, and I am labelling it as a belief because four and a half months cannot prove it, is that the people who do arrive are the ones who checked, and the ones who checked stay. A trader who has verified the discount on their own fee statement does not leave when a competitor shouts a bigger number, because they know what the bigger number means. The honest version of this business is not a smaller version of the dishonest one. It is a different business, with a different customer, on a longer clock.

What has been built

Here is the site as it stands, four and a half months in. All of it is public and all of it is checkable.

• Two venues, one rulebook. BingX at half your fees back, LBank at 0.003% taker, both stated as arithmetic against the exchanges' own schedules, with a fee calculator that runs your numbers rather than ours.
• A graded record. Every forwarded signal graded against exchange data, two accountings, losses shown, nightly refresh, and a free signals group where the grading happens in the open.
• Receipts and corrections. Individual trades with entry, exit and result; every correction made in public and left visible.
• A research desk. Token coverage that measures a coin only after its identity is verified, and publishes what happened to it afterwards, whether or not that flatters the call. The memecoin desk is the newest part of it.
• Fifteen articles before this one, written the way I would explain each subject to someone who checks the arithmetic, from how BingX fees actually work to what memecoin trading really costs.

What comes next follows the same logic. A guide to verifying your rebate on your own statement, because the legit page promises checks and that is the one that matters most. LBank covered the way BingX has been. The memecoin desk's first sixty days, published whatever they show. And the record, growing, on the same page.

What I would do differently

I would have written the outcome checks before the automation, not after. It is the cheapest discipline on this site and the last one I built.

I would have checked the front door before publishing a single graded number. Grading was never the problem; intake was.

I would have spent the first two months on the few pages a reader arrives at with a question, the fee explainer, the safety review, the record, rather than on the machinery around them. The pages are what people read.

I would not change the split. Sixty, fifty, ten is the reason every other page on this site has to be true, and I have not found a better forcing function than that.

The promise

Four and a half months is not long. What I can offer is not a track record yet but a set of commitments, each of which is already on the site and each of which you can check.

There is no paid tier, and there never will be. Every signal is graded and every loss is published. Corrections happen in public and history is never edited. No win rate is ever quoted, because a win rate without the size of the losses is marketing math. And when these numbers age, whatever they show, they will be published at the same size, on the same page.

That is the heart of it. Not a bigger number. A number you can check.