A 12-Month Funnel Model, and Where It Actually Breaks

Every affiliate site's first year produces the same question, over and over: is this month's number a traffic problem or a funnel problem? The model below answers it. It walks a hypothetical new site through twelve months, introducing one common failure at a time, and shows what each one looks like in the numbers before anyone notices it. Every figure in it is illustrative — a constructed model, not a measured site — built to show the shape of four specific breaks: bad traffic, geo mismatch, tracking loss and bonus-hunter churn. The multipliers and the site are invented. The four failure types and the checks that catch them are not.

The baseline this model assumes

A healthy month, in this model, converts 5% of visits into outbound clicks, 35% of clicks into registrations, and 55% of registrations into a first deposit. Average tracked revenue per depositor is an illustrative EUR 90. None of those numbers is a claim about any real site; they exist so the model has a "clean" line to compare against when a problem shows up. The site starts at 900 monthly visits and grows to 7,000 by month twelve — a plausible early-stage trajectory for a niche content site, not a forecast.

Twelve months, one site, four breaks

Note the FTD column throughout is what actually happened, not what a click-cookie dashboard would show. Months 7 to 9 are the exception worth remembering: the deposits are real and counted here, but a tracking gap means your own tracker would under-report them until check 2 catches it.

Month 2: the click-through rate is the tell, not the deposits

A paid social campaign triples the click-through rate to 22%, and registrations barely move. This is the pattern to distrust on sight. Vendor research on invalid traffic backs it up: CHEQ's State of Fake Traffic report found gambling among the industries with the highest fake-traffic rates it measured, in 2022 data. CHEQ sells fraud-detection tools, so treat the figure as a vendor benchmark, not an independent audit. A click-through rate several multiples above your site's own historical average, arriving alongside a new paid source, is still the first thing to distrust, not the last.

Months 4 to 6: clicks that were never going to convert

A guide starts ranking for a country the featured operators do not accept registrations from. Nothing about the click looks wrong in the aggregate numbers. The visitor is real, the click is real, and the click-through rate looks normal. It converts to nothing because the operator's licence, not the content, decides who can register, and a licence in one market says nothing about acceptance in another. This is not a traffic-quality problem, and no amount of on-page optimisation fixes it. It is a targeting mismatch, invisible until you look at conversion broken out by country rather than in total.

Months 7 to 9: real depositors, missing revenue

Registrations and first-time deposits both look normal in months 7 and 8, but tracked revenue lags what the volume implies. A rising share of clicks are coming from Safari on mobile. Apple's own engineering documentation on Intelligent Tracking Prevention describes exactly this mechanism: WebKit's own account of full third-party cookie blocking confirms that third-party cookies are blocked outright, and that script-writable cookies are capped at seven days. That caps a cookie-only attribution window for any player who deposits more than seven days after the click. The player exists, the deposit happened, and the operator knows it happened. Your own dashboard, reading only its own cookie, does not.

Months 10 to 11: the deposit that was the whole relationship

First deposits look strong once bonus-code content starts ranking. Revenue does not. The depositors this content attracts are shopping for the bonus, not the operator, and a shopper who came for the bonus typically claims it once and does not return. Bonus-driven abuse is not a minor nuisance in this vertical. A 2026 LexisNexis Risk Solutions survey of 993 North American online-gaming decision-makers found bonus abuse cited by 78% of respondents as a top fraud threat to their business — the single most-cited category in that survey. A revshare deal, or an EPC assumption, built on a blended average revenue figure quietly collapses once a growing share of first deposits belongs to this cohort. That average assumed some of them would come back.

The three checks that catch all four

• Check 1 — click-to-registration ratio, split by traffic source and by country. In aggregate, months 2 and 4 to 5 in the table above look only mildly off. Split by source, month 2's paid channel is obviously broken. Split by country, months 4 to 5's geo-mismatch traffic converts at close to zero while everything else looks normal. One segmentation catches two different failures, because both hide inside an aggregate number and neither hides inside a breakdown.
• Check 2 — reconcile postback-confirmed deposits against your own cookie-attributed count, weekly. A gap between what the operator's postback reports and what your own tracker shows is the tracking-loss signal, and it shows up as a gap, not as an error message. Ask whether the program supports server-to-server postbacks once the gap appears. Many in-house programs attribute on their own back end by btag rather than offering per-affiliate S2S by default. This may take a request, not a toggle.
• Check 3 — second-deposit rate at 30 days, tracked as its own number, not folded into total revenue. A healthy cohort keeps depositing. A bonus-hunter cohort does not. Watching this ratio catches the month-10-to-11 problem well before a full quarter of blended revenue makes it visible, and it is the number that tells you whether your EPC assumption in the traffic revenue estimator is still honest.

What the shortfall actually cost

Summed across the twelve months, this model's actual revenue comes to roughly EUR 25,700, against roughly EUR 40,500 if none of the four problems had occurred. That is a shortfall of roughly 36%, produced entirely by four temporary, catchable conditions — not by a change in traffic volume, which grew every month regardless of what was going wrong underneath it. That is the trap in a dead or damaged funnel: total visits is the number everyone watches, and it is the one number in this whole model that never told you anything was wrong.

Where this differs from a fully dead funnel

Our piece on a fully dead funnel covers the case where nothing converts at all and the diagnosis is binary — broken link, broken tracking, or no human traffic. This model is the more common, messier case. The funnel mostly works, and it loses a third of its revenue to problems that never show up as a zero anywhere. Read the two together if you are auditing a live site. The dead-funnel piece finds the catastrophic break; this one finds the slow leak. Both start from the same discipline described in our guide on traffic sources for iGaming affiliates. Know what a channel is supposed to convert at before you scale it. Then a deviation is visible on day one instead of in the quarterly total.

What to check this week

Pick one month of your own traffic and run check 1 against it — source and country, not the aggregate. If nothing looks unusual, that is a real answer too, worth re-running monthly rather than once. The checks are cheap. A quarter spent not running them is not.