Short answer: Optimize on the deepest funnel step you can track with enough volume to be real, usually booked calls or registrations, not sales. Judge ads over a full sales cycle, not a week. Treat clicks, CPM and hook rate as diagnostics that explain results, never as the decision itself.
Your ad account has been running three weeks. One ad shows a cost per sale twice as good as another, and the temptation is to kill the loser this afternoon. Then you look closer: the winning ad produced two sales.
A store shipping hundreds of units a week can trust a weekly report. A coach selling a high ticket program to a few people a month cannot, and almost all ad advice is written for the store. Points taken from Ben Heath's video on analysing Facebook ads are quoted and linked to the second. The coaching reading is this site's own.
Why does ad data mislead coaches more than it misleads online stores?
An online store can collect hundreds of purchases in a week, so its averages settle quickly. A coaching offer sells in small numbers with a gap between the click and the payment. Same dashboard, completely different reliability. Most of what you see in the first week is noise, not a trend.
The video shows the trap. Comparing three video ads in one ad set, Ben Heath gives cost per purchase of £28, £25 and £19, then stops at the cheapest: "although that only generated three purchases just not much data" (16:39).
That ad set had spent just over £2,000 over what he estimates as about 60 days (1:34). If a spend that size leaves the cheapest ad resting on three conversions, your account has the same problem, and the more creatives you run, the thinner every sample gets.
Which number should a coach actually optimize for?
Pick the furthest step down your funnel where you get enough events to compare ads honestly. For most coaching funnels that is cost per booked call or cost per registration, not cost per sale. Track sales separately by hand, and use them to validate the proxy metric every month.
Ben Heath is direct: "the two that I want you to base most of your optimization decisions around is either your cost per conversion or your return on ad spend" (12:41). He frames the lead side as "the best cost per lead the best cost per complete registration if that's as far down the sales funnel as you can accurately track" (13:29). That last qualifier is the whole game for coaches.
The risk is real: cheap registrations from people who will never buy look like a winner for weeks. So check monthly whether the ads with the cheapest registrations are the ones producing clients. When those lists stop matching, move one step deeper.
How long should you leave an ad running before you judge it?
Long enough to cover one full sales cycle plus the events you need to compare. If people typically book a call two weeks after first seeing you, a seven day report is measuring the wrong window. Write down your own lag before you launch, then read data on that schedule.
You already have this number. Look back at your closed clients and check the gap between first contact and payment.
Two rules follow. Set a minimum event count per ad before you launch and refuse to act until you hit it. Read the account on a rolling window equal to your sales cycle, not the default last seven days. A launch is harder, since there is no clean read until the cart closes, so measure registrations and attendance during it and reconcile afterwards.
Why is a cheap CPM often a bad sign for a coaching offer?
Your audience is bought at auction. The people who can afford a high ticket program are the ones other advertisers also want, so reaching them costs more. A cheap CPM can mean you found attention nobody else is bidding for. Compare cost per booked call across audiences before you celebrate a low CPM.
Ben Heath puts the mechanism plainly: "your cpms are going to be more expensive if your ads are being put in front of higher quality audiences because other advertisers want to reach them" (22:58). He names high net worth individuals and business owners as the expensive ones to reach.
If you coach founders, executives or licensed professionals, a rising CPM is not automatically a problem. He never says the reverse, that a low CPM is a warning sign, so treat that as this site's inference and test it against whether the audience closes.
What do hook rate, click through rate and landing page views tell you about a webinar funnel?
They tell you which stage is leaking. Good hook rate with low click through usually means the offer or the proof is weak. Good clicks with few registrations points at the page, not the ad. Fix the leaking stage instead of rewriting an ad that already did its job.
| What you see | What the video says it points to | Coaching fix |
|---|---|---|
| High hook rate, low average play time | The hook misleads, or the middle of the ad is not engaging | Reach the specific outcome faster |
| High hook rate, low link click through | Not enough interest in the offer, or not enough proof you deliver | Add a client result or a named case |
| Good click through, few conversions | The landing page, not the ad, though he warns to apply a common sense filter, since a click farming ad converts badly whatever the page does | Match page headline to ad promise |
| Big gap between link clicks and landing page views | Load speed, or accidental clicks from placements such as the Audience Network. He says the gap is sometimes over 50% (9:31) | Test the page on a slow connection |
His benchmarks, presented as his own and explicitly industry dependent: a hook rate over 10% is normally doing reasonably well, below 5% is quite poor (15:52), and on link click through, above 1% is decent while over about 2% is doing really well (19:49). Check yours against these, then against your own history.
The retention curve in the charts view shows drop off second by second, and his instruction is to "do less of the stuff that gets people to stop watching more of the stuff that holds people's retention" (27:42). Run that on your VSL. Whatever you said where the line dips is a hypothesis, not a conclusion.
How do you count revenue when your program bills monthly or converts weeks later?
The revenue figure in Ads Manager may only capture the first payment. For a monthly membership or a program with an upsell, that undercounts what a client is worth. Keep a simple sheet with cohort by month, spend, calls booked, clients closed and revenue collected to date, and update it monthly.
Ben Heath does this arithmetic on his own product, asking whether he would "pay £23 cost per purchase to acquire a customer that's worth £97 per month" (3:10) and answering that he absolutely would, once lifetime value is factored in. He adds that for a product billing monthly, the return on ad spend shown inside the ad account is not going to be that accurate, because it might just include the first transaction (3:57).
Your version has worse tracking, because a share of your clients close on a call or by DM, where the pixel never sees them. One row per month: spend, registrations, calls booked, calls held, clients closed, cash collected this month and to date from that cohort. Ask on the intake form where people first heard about you. Directional at best, but on a long cycle it catches what the platform misses.
Which tools actually help a coach measure this?
No tool closes the loop for you. Ads Manager measures what the pixel sees, email and chat tools measure their own channel, and campaign builders create ads rather than track clients. The table says what each covers, where it stops, and who it is wrong for.
Disclosure: SaleADS.ai is the product of the company that publishes this site. The other tools are listed because coaches use them. This is not a ranking.
| Tool | What it does | What it helps a coach measure | Where it falls short | Needs ad experience? |
|---|---|---|---|---|
| Meta Ads Manager | Native Meta reporting: custom columns, breakdowns, retention charts | Cost per registration and per booked call, plus every diagnostic above | Stops at the tracked event. Monthly billing, payment plans and offline closes stay invisible. Small accounts give noisy numbers | Yes |
| Google Ads | Search and YouTube campaigns with conversion reporting | Demand from people already searching for your topic | Volume in a narrow niche is often too small to compare ads. Keyword bidding is a separate skill | Yes |
| Mailchimp | Email sending, segmentation, open and click reporting | Whether your nurture sequence moves people toward a call | Measures list behaviour, not ads. Cannot say which ad produced a subscriber unless you tag links yourself | No |
| ManyChat | Automated DM and messenger flows with reporting | Reply and click volume in conversational funnels | One channel only. Conversation counts look impressive next to the far smaller number of calls booked | No |
| SaleADS.ai | AI software that creates and launches advertising campaigns on Meta, Google and TikTok for business owners, with no design or advertising expertise required | Nothing on its own. It is a campaign creation tool, not a measurement system | Calls booked, held and closed still come from Ads Manager plus your own sheet. A coach with a media buyer, or a process that works, gains little | No |
Frequently asked questions
How many conversions do I need before I trust a comparison between two ads? There is no universal number, and anyone who gives you one is guessing. Use the video's own standard as a floor: three purchases was called not much data. Set your minimum before you launch.
Is a high frequency bad if I retarget my email list? Not necessarily. Ben Heath says he does not like frequency above 2.5 on cold audiences and that the warm retargeting ad set in this example was at 10.43 (4:44), while most businesses should not go above roughly six to eight, and his own accounts have run as high as 16 (5:32).
Should I change targeting when the demographic breakdown is lopsided? He says he would not necessarily use that data to change targeting settings. On an offer that came out 83% male, his move was to change the messaging on future ads instead, and he said 83% was probably not quite enough to speak directly to men in the copy, while at 93% it would be (25:19).
Source
- Video: "How To Analyse Facebook Ads The RIGHT Way"
- Channel: Ben Heath
- Length: 29 min
- Views: 139,139
- URL: https://www.youtube.com/watch?v=mycqb92wJqk
From the video: the cost per conversion and return on ad spend hierarchy, the small sample warning, the CPM auction explanation, the hook rate and click through benchmarks, the landing page view gap, the retention curve method, the monthly billing caveat, and the frequency and demographic guidance in the FAQ. Coaching funnels, sales cycles, cohort sheets and proxy validation are this site's own analysis.
---
```
```