Cost per conversation went from $4.10 to $2.60 last month. You put it in the report with the arrow pointing the right way and the client said thanks. Two weeks later on a call they said sales feels slower than it did in spring. Both of those things are true at the same time, and the gap between them is the whole subject of this post. Two ad sets. Pick the one you kill. Same client, same month, same offer. - Ad set A costs $2.10 per conversation. - Ad set B costs $4.80 per conversation. Decide now, before the next line. Most people kill B, because B is more than twice the price and the report only has one column. Here's the column that isn't in the report. Set A qualifies 12% of the conversations it brings. Set B qualifies 45%. Which puts A at $17.50 per qualified lead and B at $10.70. Kill A. Scale B. The cheap one was the expensive one, and nothing in Ads Manager was ever going to tell you that. Or as the source page puts it: the most expensive leads an agency can buy are the cheap ones that never buy anything. The formula CPQL = ad spend (from Ads Manager) ÷ qualified leads (from Chatfuel). Two numbers, two screens, one division. Run the same arithmetic one step further and you get cost per sale: spend ÷ closed deals. That second one is the number the client is actually doing in their head while you present the first one. Why CPL flatters you for free Meta's average cost per lead sits around $28 and it's been climbing, up roughly 21% year over year. Agencies love quoting that number next to their own cost per conversation, because $2.60 against $28 looks like a miracle. It isn't a comparison. Meta's $28 is a form-fill. Yours is a chat that opened. Different objects, different odds of turning into money. The structural reason this keeps happening: Meta optimizes toward what it can see, which is conversations started. Only you can see what happened after the conversation started. That asymmetry is the entire edge of running chat funnels instead of forms, and most agencies give it away by reporting the same metric Meta already optimizes.

The contract is signed, the build is ready, and the project sits for two weeks. Not because anything is broken, but because the client hasn't clicked the thing. Access is where more chat funnels die than at any technical step, and it dies for a boring reason: somebody wrote a confusing message. Put yourself on their side. A person you met three weeks ago is asking for access to the business account that runs their advertising, their page, and their customer conversations. Of course they hesitate. The fix is not persistence, it's writing the ask so there is nothing left to worry about. Three rules before any of the scripts One ask per message. If you request access, documents and a call in the same message, you get none of them. Bundle the assets into one action that feels like a single thing to approve. Frame it around their outcome, not your tooling. They don't care that you need pixel access. They care that leads get answered. Answer the safety question before it's asked. The sentence that unlocks most of these is a version of: you keep full ownership, and you can revoke everything in one click. Script 1: the ask, when the client already has a Business Portfolio Send it the same day the contract is signed, while the yes is still warm. Hi [Name], to get your campaigns live we need access to your business assets on Meta. The professional way to do this is partner access: you connect our agency to your Business Portfolio, and we only get the specific permissions we need to do the work. You keep full ownership of everything. We never get your password, you can see exactly what we have access to, and you can remove us in one click at any time. We're asking for four things: the ad account, the Facebook Page, the Instagram account, and the pixel. I'll send the request from our side so all you have to do is approve it. Our Business Portfolio ID is [ID]. If you'd rather do it yourself, tell me and I'll send the click-by-click. Script 2: the click-by-click, for clients who want to do it themselves

Meta cuts the primary text of an ad at around 125 characters, and about one viewer in a hundred taps "See more". Everything under that line is written for the one percent. The ad is the first line. This sentence is exactly as long as Meta lets your ad be before it cuts you off, and you just read all of it in one breath. That's 123 characters. That's the whole budget. If you've ever decided you can't write ads, this is good news, because at that length nobody is writing. There's no room for craft in a line and a half. There are formulas, and you can learn them in an evening. The eight formulas These are the ones that keep working in chat funnels. Read the examples as shapes, not as copy to lift. - Pain, agitate, solve. "Empty chairs on weekdays? Every quiet afternoon is rent you're paying for nothing." - Callout. "Barcelona homeowners: your flat may be worth more than you think." - Number or proof. "137 apartments sold through this one WhatsApp funnel." - Curiosity gap. "The one question that tells us if a lead will actually buy." - Social proof. "Why 2,400 students started this course with the same message." - Urgency. "12 spots left for July, and after that it's September." - First person. "I messaged them at 11pm expecting nothing. Got a reply in 5 seconds." - Question. "Still paying for leads that never answer the phone?" - Every one of them puts something countable in the opening line: a number, a price, or a named outcome. That's the part people skip. "Boost your sales with automation" isn't a hook, it's a category. "137 apartments sold through this one WhatsApp funnel" is a hook, because there's something in it you can picture. The second thing they share is voice. Write the way people type in a chat, not the way brands write press releases. Emoji are fine, one to three, working as anchors for the eye rather than decoration. The most common way I watch those 125 characters get burned is on introductions. "Welcome to [Clinic Name], your trusted partner in modern aesthetics since 2016" and the line is spent before the reader has learned one thing about themselves. Your name goes in the profile. The first line belongs to them.

Pricing the first package is where I watch most people go sideways. Either they undercharge and quietly start resenting the client, or they invent a clever per-lead scheme that blows up around month two. The shape that holds up is boring: a setup fee plus a monthly fee. What's worth understanding is what each part actually buys, and why the clever alternatives fail. What the setup fee pays for The setup fee covers the build: connecting the channel, loading the client's business context, defining what counts as a qualified lead, configuring the handoff to a human, setting business-hours behavior, and testing the whole path before real traffic hits it. Real hours, spent once. In the US frame I'm working from, setup runs $500 to $1,500 depending on how much of that list the client needs. What the monthly fee pays for The monthly fee is what makes this a service instead of a gig. It pays for operating the thing: reading conversations, tightening the qualification rules when the wrong people get through, adjusting follow-ups, and showing the client evidence of what happened. The US frame is $300 to $1,000 a month. One honest rule: if you don't intend to do that operating work, charge more for setup and skip the retainer, because a monthly fee for work nobody does is where the bad reputation of retainers comes from. Why never per lead Per-lead pricing sounds fair on the surface. You get paid for results, the client pays for value. It fails on contact with reality, for reasons that have nothing to do with fairness. You don't control the volume. Lead count is a function of the client's ad budget, and the client can halve that budget without telling you. Your revenue now lives inside someone else's ad account. The incentives point the wrong way. Paid per lead, you earn more when more people get through, while the client is paying for fewer, better conversations. Sooner or later the client does this math too. And you will fight about definitions. Is a lead someone who wrote once? Someone who got qualified? Someone who booked and didn't show up? Per-lead pricing turns every invoice into an argument about what a lead is.