THE METHOD, FROM THE INSIDE
The three rules, week by week
The home page has the three rules that decide where your money goes. This page is what doesn't fit there: what happens each week of a typical quarter, what the report you receive looks like, what gets decided in each review and where the money leaks when there is no rule. It's written for a business that already sells — its own catalogue or its own service — not for someone still choosing what to sell.
First things first: who this is for
Nothing here picks products or builds a business from scratch. It works with what you already have: your catalogue or your service, your prices, your margin. The rules decide which ad and which angle deserve budget, when to cut and when to raise. Nothing more and nothing less. If you want someone to tell you what to sell, I'm not your supplier.
The three rules, with their why
Rule 1 — The cut at €40–50 of spend
Every test has its verdict written before it's switched on: at €40–50 of spend, 0 or 1 conversions means it's turned off; 3 or more makes it a candidate for scaling. With 2, the rule doesn't decide on its own: cost per purchase against your margin decides, which is rule 3.
Why that figure and not another: it's enough spend to discard and still not enough to confirm. A test that shows a signal isn't declared a winner — it becomes a candidate and gets watched with more spend on it. All the value of the cut runs the other way: killing fast and cheap what shows no signal, so that nobody — me included — can defend their own idea for three weeks at your expense.
Rule 2 — The 20% scaling every three days
When there's a signal, the budget rises in 20% steps, never in one jump and never on the same day the first good result shows up. A big leap resets the campaign's learning: the money you already spent teaching it who to sell to is thrown away, and you're back at the start paying more for it. The three days exist so every raise rests on history, not on one good day.
Rule 3 — The hierarchy: CPA against your margin is in charge
CTR above 2% and CPC below €0.15 are my own cold-start references from Spain. They're for smelling a problem early and knowing what to fix: a low CTR points at the creative; clicks that don't end in purchases point at the page or the price. What they are not is a cut criterion. The case published on the home page ran with a CPC double my reference and was profitable, because its cost per purchase fit inside its margin. When a reference collides with the CPA, the CPA wins. Always. And every reference is also compared against your own account's history, not against some absolute from the internet.
A typical quarter, told in advance
The calendar below is a guide, not a promise: decisions run on accumulated spend, not on dates, so your budget sets the real pace. But a quarter with the rules in place looks like this:
Weeks 1-2 · Nothing switches on until the tracking reconciles
An audit of your account before touching anything. Pixel and Conversions API verified event by event, and Meta's numbers reconciled against your real sales. The cut and scaling rules are handed to you in writing. The first creatives go into production. Launching before the tracking reconciles is spending in order not to learn, and I don't do it.
Weeks 2-6 · The test cycle
The first angles launch and every test runs to its cut. Things get killed here, and it's the phase that looks ugly on the dashboard: spend going out and results not yet coming in. The report tells it as it is, not dressed up. On the home page you have the real distribution of one case — three nearly flat weeks and almost all the result at the end: a real quarter looks more like that than like a straight line.
Weeks 6-10 · Scaling and creative rotation
What shows a signal rises in 20% steps every three days. At the same time the next creatives are produced, because creatives wear out and the moment to produce the next one is while the current one still performs, not once it has died on the dashboard.
Weeks 10-13 · The decision of the quarter
With three months of data, the decision is made with numbers on the table: continue, adjust or stop. If the honest thing is to stop, I say it before you have to. Stringing you along would cost me the only thing this website is trying to build: that you can hold me to the rules.
The weekly report, column by column
It's the same dashboard I look at, not a client version. Every week, these columns and what gets decided with each one:
| Column | What it is | What decision hangs on it |
|---|---|---|
| Spend | What went in this week and the accumulated spend per test | Triggers the cuts: decisions run on spend, not on the calendar |
| Real results | Purchases or enquiries according to your system, not only Meta's dashboard | The reconciliation. If Meta and your sales don't say the same thing, the gap gets explained before anything is decided |
| Cost per purchase (CPA) | Spend divided by real results | The number in charge. Against your margin it decides kill, hold or scale |
| ROAS | Revenue divided by ad spend | Context for the CPA. It doesn't deduct product, shipping or fees: it isn't profit, and the report says so |
| Real margin | What each sale leaves after costs | The line the CPA is compared against. Without this column, ROAS is decoration |
| CTR · CPC · CPM | Diagnostic references | They say what to fix — creative, page or price — and cut nothing on their own |
| Frequency | How many times the same person has seen the ad | Warns of creative wear before you see it in the CPA |
| Decision of the week | Kill, hold, iterate or scale, and why | The row you audit me with: every decision has to be traceable to the rules on this page |
What gets decided in each review, and who decides what
Every weekly review ends in one of four decisions per test: kill, hold, iterate the creative or scale. That part I execute, with the rules in writing — which is why you can check every decision against the criterion instead of taking my word for it.
What I don't decide on my own: raising the monthly budget ceiling, touching prices or offers, or changing the strategic angle. That comes to the review with the numbers on the table, and you decide it. My job is to get you to that decision with data and with the bad news included, not with a report painted green.
Where the money leaks when there is no rule
This table is the part of the method I've used most on other people's accounts. Each row is a real hole, how it sounds from the inside while it's happening to you, and the rule that plugs it.
| The hole | How it sounds when it's happening to you | The insurance |
|---|---|---|
| Spend without a rule | 'I'll give it one more day, it might pick up' | The cut signed before switching anything on. The campaign doesn't argue: it's measured against the rule |
| Brute-force scaling | 'It works, double it now' | 20% steps with three days of history. A big leap resets the learning and you pay to get back to where you already were |
| The wrong metric | 'The CTR is great, but I'm not selling' | The hierarchy: CPA against your margin is in charge. The rest diagnoses, it doesn't celebrate |
| Broken tracking | 'Meta says twelve sales and my system says seven' | Events verified one by one and a weekly reconciliation between dashboard and real sales. The gap gets explained, not hidden |
| The invisible setting | 'The report is green, but every month I reach fewer people' | Every setting audited against its cost, not against habit. The placements experiment on the home page came out of exactly this |
| The worn-out creative | 'It worked for weeks and then it died' | Frequency is watched in the report and the replacement is produced while the current one still performs |
What this method won't solve for you
- It doesn't promise results. It promises process: written rules, reconciled tracking and an auditable weekly review. The market, the algorithm and your competitors obey nobody.
- It doesn't fix a margin that can't take it. If your margin can't carry a realistic acquisition cost, no rule saves it. That gets looked at in the diagnostic, and if that's you, I tell you there and charge you nothing.
- It doesn't eliminate losing tests. Tests lose. They're budgeted to lose: they're the price of the information.
- It isn't multichannel. Everything here is measured on Meta. For Google, LinkedIn or TikTok I have no data of my own and I won't pretend otherwise.
If you want these rules running on your account
The full prices are published, with the quarter's sums done. And the diagnostic is free, whether we fit or not.
See the prices in detail · Request the diagnostic — six fields, two minutes