10.1 The Funnel That Actually Converts

An account that collected 2.35 million views over 90 days and approximately zero revenue is the most common failure report in this business, and almost everyone who tells that story blames the content. The content was not the problem. Reach arrived and had nowhere to go: no routing, no landing page, no capture, therefore no revenue. This chapter is the chain that turns an impression into a paying chat user, the leak at every link, and the instrument that tells you which link is broken. It is the highest-value chapter in Part X, because three independent operators in the research each placed more than two-thirds of their revenue here.

The chain, link by link

Five links, in this order. Each one has a job and each one can fail on its own.

The Instagram reach baseline fell roughly 35 percent year over year [AS OF 2026-10]. Virtual accounts still post engagement rates of about 5.67 percent against 1.89 percent for human accounts, with trust scores of 42 out of 100 against 73 out of 100 (see 1.4). Read those together and the strategy writes itself: you will get more interaction than a human account and less belief, so the funnel has to work on curiosity and specificity rather than on authority. Reach is not a distribution strategy and never was.

The 2.35 million view case, step by step

This case is worth walking slowly, because it is reproduced by most new accounts and almost none of them diagnose it correctly.

1. Reach achieved. 2.35 million views over 90 days. The algorithm liked something — format, face, or hook. This link worked.
2. No routing. The profile had no functional path: no link-in-bio destination that resolved, no Story sticker sequence, no comment-to-DM trigger, no named destination anywhere in the caption. A viewer who wanted to go further had no next step that produced a measurable event.
3. No landing. There was no owned page carrying the click, the terms, or the disclosure. So nothing was recorded, and the operator could not see the drop even after it happened.
4. No capture. No handle, no email, no consent, no channel membership. The audience stayed anonymous, which means the 90 days of attention produced zero addresses that could be contacted again.
5. Therefore no revenue. Not a low conversion rate. No conversion surface at all.

The instructive part is step 3. Reach without instrumentation does not fail loudly; it fails silently, and the operator concludes that the niche is dead. Two changes fix this class of case: a landing page with one measurable action on it, and a channel that survives a platform change. Neither costs money. Both require that you decide the path before you post.

What the landing page must contain

One page, one job. A landing page that asks the visitor to do two things does neither.

1. The disclosure. A plain line saying the persona is AI-generated, visible without scrolling, before anything else is asked of the visitor. This is the cheapest compliance layer you will build; the template language lives in 11.3.
2. One primary action, stated as a verb, repeated above and below the fold. Everything else on the page is secondary and visually quieter.
3. The operator identity. A Polish company, Zordon Intelligence sp. z o.o., is the seller. Who is charging the card, and under what terms, belongs on the page, not behind it (see 11.8 for VAT, invoicing, and the EU withdrawal right on digital content).
4. The cancellation and contact path. A way to reach a human, and the renewal terms stated in plain language. The auto-renewal duties that matter for churn are covered in 11.7.
5. Attribution. A UTM tag or a sticker code so you can see which post produced the click. A landing page you cannot attribute is a landing page you cannot tune.
6. Age gate on any paid surface. Nothing paid sits behind a single click from an open feed.

Forbidden: a direct link from the social profile to the monetization platform. It is a flag, and chapter 8.5 lists it among the patterns that cost accounts. The route goes through a landing page or an intermediate channel, always, even when the shorter path would convert better this week.

Routing: four options and what each costs

Run one route at a time. Testing procedure, in order:

1. Hold the content constant. Changing the creative and the route together teaches you nothing.
2. Set a floor before you start: 2,000 landing sessions per variant and 14 days. [UNVERIFIED]
3. Read the step-by-step drop, not the endpoint. A route that halves profile-tap-to-landing but doubles landing-to-signup is often the better route.
4. Log the result with dates. Routes decay; a link-in-bio that worked in March can lose 40 percent of its tap-through after a platform change in June.

The funnel table, with honest ranges

Every range below is [UNVERIFIED]: assembled from operator reports on forums and creator communities. No platform publishes cohort data, and the operators who know their numbers do not publish them. Treat this table as a starting hypothesis and replace each row with your own figure after 30 days of instrumented traffic.

Multiply the ranges through and the mechanism behind "millions of views, no money" becomes ordinary arithmetic. Two million views at a 2 percent profile-tap rate is 40,000 taps; at 35 percent that is 14,000 landing sessions; at 20 percent, 2,800 joins; at 8 percent, 224 checkouts. The chain does not need one heroic number. It needs no link at zero.

Which link is broken

You do not need a dashboard. You need the ratio that is furthest below its own previous 30 days.

If reach moves and nothing downstream moves, the routing link is broken. Check the link itself from a phone that is not yours, in the in-app browser, while logged out.

If routing moves and the landing does not, the page is the problem: too slow, unclear, or asking for two things.

If the landing moves and signup does not, the offer or the price is the problem, not the traffic.

If signup moves and chat does not, you have an activation problem, and chapter 10.3 is the fix.

If chat happens and renewal does not, you have a retention problem, and chapter 10.4 is the fix.