Meta ads · lead generation · health & legal practices
I find out why your ads bring the wrong people — and fix it before the form, not after.
You also get a written audit of your account, free, whether or not we work together.
01 — The problem
Three things practice owners say, almost word for word, before anyone has opened an account.
01
Your front desk is being paid to turn people away. That is an hour a day of salary spent undoing what the ad just did.
02
Cost per enquiry and cost per patient are different numbers, and only one of them pays for anything.
03
Your follow-up, they said. Sometimes true. Still not the whole answer — calling the wrong person faster is a faster no.
Recognise any of those? Twenty minutes is usually enough to tell which one you actually have.
See how I'd diagnose this →More leads isn't the answer.
It is the most expensive way to avoid asking a harder question, which is who those leads were ever going to be.
02 — The ask
This is the whole argument, and it takes one paragraph.
An ad optimised for cheap enquiries goes and finds people who are cheap to reach. Curious, comparing, not ready — and at that moment, genuinely interested. Nothing is broken. The system did exactly what it was asked.
So change what it is asked. The request inside the ad decides who is willing to answer it. "Download the guide" is answered by people who want a guide. "Tell us what happened and we'll call you back" can only be answered by someone who has something to tell you.
Same budget. Same platform. A different set of people arriving at your front desk — which is why I measure the cost of a qualified enquiry and treat cost per lead as a symptom, never a target.
Want to know what your ads are currently asking for? I'll read them back to you from the public Ad Library, before you give me anything.
See how I'd diagnose this →03 — The method
In this order, every time. The order is the method — most of what goes wrong in an ad account is a step taken before the one it depends on.
01 · Tracking
Is Meta actually counting your enquiries correctly? Event match quality, the Conversions API, deduplication.
Broken tracking looks exactly like poor performance, and it sends people off rebuilding creative that was never the problem.
02 · Economics
What one patient or case is worth, what is left after delivery, and therefore what you can afford to pay for one.
If the arithmetic says ads cannot work at your margin, you hear it on the first call, not in month three.
03 · The ask
What the ad asks for, at which stage of awareness, in the words your patients or clients actually use.
Then a grid: awareness level down one side, angle across the top. Every cell is an ad I owe you, so coverage gaps become visible.
04 · Testing
A creative test needs to spend roughly five times your target cost per acquisition before the result means anything.
So I work out how many real tests your budget supports, run that many, and log each against the hypothesis it answers.
Why the order matters. Judge performance before checking tracking and you diagnose a problem that isn't there. Buy creative before doing the economics and you may be buying it for an account that cannot be profitable at any cost per lead.
Cheap leads can be expensive.
On the worked scenario below, the campaign with the cheapest leads cost 3.3 times the target to actually acquire a customer. The one that looked most expensive came in 29% under.
04 — The system
Seven stages between someone seeing an ad and someone paying you. Ads Manager only shows you three of them.
Every stage sets the price of the next one. Meta only reports on the three in navy — so a problem anywhere else arrives disguised as an ads problem.
This is why my first answer is sometimes "your ads are fine." On one account I looked at, 96% of paid traffic never reached the form — no ad change would have fixed that, and a month of creative tests would have proved nothing.
05 — The work
Four pieces, one for each step of the method. All were produced during advanced training — three worked on real accounts, with real data and access granted for that purpose, and one on figures supplied by the course. None was a paid engagement. Client names are removed; the figures stand exactly as they do in the source files.
3.3×
Target acquisition cost, overshot by the campaign with the cheapest leads.Unit economics
76%
Of the drop in people adding to basket had already happened before the thing everyone blamed.Account diagnostic
168:8
Google reviews, nearest competitor against the client. Found in research, not in the ad account.Market research
96%
Of paid traffic that never reached the form — so week four tests the page, not another ad.Test design
Proves step 02 · Economics
Three lead-gen campaigns, judged two ways. On cost per lead the ranking is obvious and it is wrong. Campaign C bought leads 29% cheaper than the campaign hitting target, with the best click metrics of the three — and converted them at 1.5% through the back end, putting its real cost to acquire a customer at 3.3× target. Campaign B paid 34% more per lead and landed at 0.71× target.
The target is built, not assumed: gross profit per customer, minus the contribution the business wants to keep, gives target acquisition cost. Times the lead-to-sale rate, it gives the only cost per lead that means anything.
Worked scenario · figures supplied by the course, diagnosis and verdicts mine
Proves step 01 · Diagnose before rebuilding
Return on ad spend fell 43% in a month. Everyone in the room had already decided why: two bestselling products were unpublished on 2 February. The diagnostic tested that story, and it did not hold.
Creative was ruled out with evidence, not opinion — click-through actually rose 4.6%. Tracking was ruled out the same way: basket-to-purchase held steady, which a broken pixel could not do. The ads still worked. The offer had stopped.
15-page diagnostic · real account, access granted for training · every figure derived from two daily Ads Manager exports and reconciled to reported totals
Proves step 03 · The ask, then the creative
129 pages on a US clinic network, built from seven evidence streams and a 402-item archive across 163 accounts, parsed by script so anyone can re-derive every count. Every claim is tagged observed, inference or gap.
It was then audited against its own rules, and two claims already in print turned out to be wrong. Both are corrected in the open. The most useful finding was not either error but what they shared: a summary of a source had been trusted instead of the source.
The headline finding reframed the account: the market's loudest group is not trying to enter the system, it is fighting to stay in it — and nobody in the category was speaking to them.
Research pack v4.6 · real account, access granted for training · research only — it stops at the implication and never writes the ad
Proves steps 03 and 04 · Briefs and test design
Three concepts, each written as a video brief and a static brief. Every one is a complete packet: persona, desire, objection, hypothesis, hooks word for word, the beats in order, the execution pattern analysed, and the compliance lines that cannot be crossed. An editor opens it and starts cutting — nothing in it is a mood board. Not yet shot; ready to be.
One opened with a blocking condition instead of a shot list: it promised something the research could not confirm the clinic does, so it could not be filmed until someone confirmed it in writing. The practice has since confirmed it, and that brief is cleared. A client-supplied statistic was rejected the same way and stayed rejected — the brand had five contradictory versions of it live in public.
The four-week plan behind them is six bets, not twelve, each funded at five times target cost per lead so the result can be read at all, each with its stop rule written before launch.
6 briefs, written and cleared — not yet shot · 4-week plan · 7-tab testing system · real account, access granted for training
Client names, brand names and account identifiers are removed throughout, and figures are given as ratios and rates rather than anyone's revenue. I'll walk through any of these live, with the file open.
Pick whichever of the four is closest to your situation and I'll walk you through it.
Let's look at your account →Don't optimise the ad before you understand the system.
Every one of those four documents is the same habit applied to a different layer: find out what is true before changing anything.
06 — Proof
You're going to ask about results. Here's the honest version, before you have to.
Good marketing doesn't require pretending uncertainty isn't there. The first thing I tell you may be that ads are not your problem — that isn't a sales technique, it's the most common honest finding.
07 — Fit
And, just as usefully, who it doesn't.
Health and legal practices, where one case or one patient justifies a meaningful cost to acquire. If a lead is worth very little, better leads change very little.
Enough margin after delivery that there is a cost per acquisition at which ads make money. We settle this on the first call, with your numbers.
You'd rather have a finding than a dashboard, and you're willing to change the offer or the page if that's where the problem turns out to be.
— When I'd say no
If what's left after delivery costs is close to nothing, there's no acquisition cost at which ads are profitable. No amount of skill changes that arithmetic.
If monthly spend supports two or three readable tests, we can't learn fast enough to stay ahead of creative fatigue. Better to say so than to bill for it.
Buying better leads for a practice that takes two days to call back is buying nothing. Fix response time first — it's cheaper than any ad budget.
08 — The free audit
Four steps. You can stop after any of them and keep what you've been given.
Your numbers, what you've spent, who ran it before and why it ended. I'm listening for whether there's a problem I can actually solve.
No cost · no commitmentThe ad account at view performance, the pixel at view. Nothing more — an audit is entirely readable at view level, and I send you the exact click path.
You revoke it in one clickTracking health, where the money goes, what the ads are asking for, and the three things I'd change first — each tied to something I can point at.
A few days · still freeIf you want me to run it, we talk scope and price then, with your economics on the table. If not, you keep the audit and whoever runs your ads can use it.
Priced after, never beforeStep one is the only one that needs booking. Everything after it happens without you in the room.
Let's look at your account →09 — The engagement
So you can picture it rather than take it on trust. Exact volumes come off your own budget, using the arithmetic in step 04.
01
Every test named, with the angle, the awareness level, the hypothesis, the budget it gets, and the rule that decides whether it continues, iterates or stops — written before anything launches.
Agreed with you, not shown to you after02
One brief per ad: the job, the awareness level, the hook word for word, the beats, and annotated reference saying what to take from each one. Your editor or mine — either way they're not guessing.
You keep every brief03
Campaign build to a naming convention that makes results traceable, a pre-launch policy check, then a weekly pass: what's still learning, what's spent enough to judge, what gets killed, what gets scaled.
A short note each week04
The number against target, what changed and why, what I'm watching next — with the attribution window and date range stated, so two months are actually comparable.
A short deck, walked through liveHow many tests a month? That's not a package, it's arithmetic. A readable test costs roughly your target acquisition cost times five. Your budget divided by that is how many real tests the month supports — and if the answer is two, I'll say so rather than sell you twelve.
10 — The call
So you know what you're booking.
01
Current spend, what one patient or case is worth, what happens after someone enquires, and who ran the ads before.
02
Where in the seven stages above the money is most likely leaking — and what I'd want to see to confirm it.
03
Sometimes they aren't. If the answer is the offer, the page or the follow-up, you'll hear that instead.
04
Either the free audit, or a specific thing to fix that has nothing to do with hiring me.
No pitch deck, no pressure. Just a diagnosis you can use whether or not we work together.
11 — Questions
Longer than anyone selling to you will admit. Meta's delivery needs roughly fifty conversion events per ad set per week before it's even stable, and every meaningful edit restarts that clock. A verdict inside seven days is a guess dressed as a result.
What you get quickly is the audit — a few days — and honest early signals: whether tracking is intact, whether the ads are asking the right thing, whether cost per enquiry is moving. What takes a cycle or two is the number that matters: the cost of an enquiry that becomes a patient or a case.
For the audit: view performance on the ad account and view on the pixel, granted through your Business Manager as a partner. Never your login, and never admin before there's a contract — everything an audit needs is readable at view level, and asking for more is a warning sign you should notice in anyone.
If we work together I'll then ask for what the work genuinely requires: advertiser or admin on the ad account, the Page, and the catalogue if there is one. Every grant and revocation gets logged.
I quote after the audit, not before, and the quote comes off your economics rather than a rate card. What one case is worth, what you can afford to pay to acquire one, and how much testing your budget supports all change what's sensible to charge.
The thing worth weighing it against is what the current situation already costs. On one account I looked at, 96% of paid traffic never reached the form — that money was already spent, and almost none of it could turn into a patient. A practice in that position is paying for the problem every month, and it appears as a line item nowhere.
The audit itself is free and stays free whatever you decide.
It's the most common finding and often the cheapest win available. Poor event matching makes an account report worse than it's performing, and people respond by rebuilding creative that was never the problem.
I check it first precisely because it's so often wrong. Improving the data sent with each event can lift reported performance without touching a single ad.
I write the strategy, the angles, the hooks and the briefs — including annotated visual reference, so whoever edits knows exactly what to take from each one. Production depends on what you already have: some practices have footage and a designer, some need everything sourced.
We settle who does what at the kick-off call, in writing. Most scope arguments are really about something nobody named at the start.
Yes. The work above includes a US clinic network and a direct-to-consumer ecommerce account. Meta's mechanics don't change by country; what changes is the advertising rules in your category and the language your market actually uses, and both of those are research questions I'd handle the same way anywhere.
The setting changes and the argument doesn't. Health and legal practices share what matters here: a high-value enquiry, a long decision, a front desk that pays for every unqualified call, and advertising rules that punish an over-promise. The ask still decides who replies.
Still deciding? The call costs you twenty minutes and the audit costs you nothing. That's the whole risk.
Let's look at your account →A written audit, a few days, no cost, and yours to keep whatever you decide afterwards. The worst case is that you learn something about your own account for free.
I write these myself rather than run them through a template, so I take a small number at a time. And if you're thinking about the new year, starting now is what makes January a launch month instead of a research month.
Ads bringing activity but not business? Let's find out why.
Get a free strategy call →