Approach
The backend decides. Everything else is a lead.
How we buy, test and report, so budget follows what your backend confirms, not what a dashboard claims.
Built for teams spending around $10K a month on ads or more.
- Meta claimed
- 23
- Backend saw
- 4
One day, and the 6-day and 29-day reads pointed the same way. Anonymized. Online gaming, Meta Ads. Check your own platform against your backend
Six rules, written down.
They go into the first call, the proposal and every weekly report.
- 01
Backend first
Platform numbers are a lead, the backend is the verdict. Ads Manager, Google Ads and every DSP grade their own homework, so every report we send puts the platform figure next to the number your backend or MMP sees, and the decision follows the second one.
- 02
Kill rules before launch
Before a test spends, its stop rule is written down: the metric, the threshold and the budget at which it dies. That keeps a good-looking dashboard from rescuing a test that is not paying back.
- 03
Offer, measurement, then channel
Buying more traffic never fixes a broken offer or broken tracking. If the paywall, the landing page or the events are wrong, that gets fixed before the budget goes up.
- 04
Your accounts, your data
Ad accounts, pixels and the MMP stay in your name, so the history and the learning stay with you.
- 05
Numbers with their denominator
Every result comes with its sample size and its time window. One good day is a signal, not a trend, and we say which one it is.
- 06
Inside the rules
Every campaign follows the platform's ad policies and the market's licensing rules. No cloaking, and no AI avatars presented as real customers.
From first call to weekly rhythm.
The same sequence for an app, an ecommerce brand or a licensed operator. The channels change, the order does not.
Diagnostic
A short intake form, a pre-read built from public data, and 30 minutes where Jan walks through the three biggest opportunities.
Measurement before money
Attribution map end to end: ad platform, MMP or tracker, analytics and backend. Broken events get fixed first.
Plan with ceilings
Maximum CPI or CPA per channel and country tier, derived from price, fees and conversion, not from benchmarks.
Launch with stop rules
Structure, first creative batch and a written kill rule for every test.
Weekly rhythm
New creative every week, platform and backend side by side, and a plain verdict on what to scale, fix or stop.
A pre-read built from public data, then 30 minutes with Jan on the three biggest opportunities.
Agents do the legwork. Jan signs off.
AI agents handle ad library research, first-pass audits, creative production and QA checklists. Jan reads the output, checks the numbers and decides. That is how a small studio covers this many channels without cutting corners.
An example of the method: AI agents checked 301 claims in an Apple Ads operating reference against 52 official help articles, and 26 of them were corrected.
- 301
- claims checked
- 52
- official help articles
- 26
- claims corrected
Many agents, one person accountable
Illustrative. Research, first-pass audits, creative and QA all end in Jan's review before anything ships.
What the rules caught.
Results from past and current work, anonymized, with the caveats that come with them.
- Signal$1,239true cost per depositor
An optimization event sat flat at about $6 per lead from July to August while the true cost per depositor rose from $195 to about $1,239.Past engagement, anonymized. Gaming app, DSP.
- Platform vs backend23 vs 4depositors, Meta vs backend
On the first full day of clean tracking, Meta claimed 23 first-time depositors on one brand where the backend saw 4. One day, and the 6-day and 29-day reads pointed the same way.Anonymized. Online gaming, Meta Ads.
- Kill rule0new depositors at the kill
The kill rule was written before launch. We killed the search offer test at zero new depositors while the platform rated it the account's best cost per conversion.Anonymized. Online gaming, Google Ads.
- Pre-launch QA2bid-corrupting setups caught
Caught two bid-corrupting setups before launch: conversion events firing for over 99% of users, and a purchase action that defaulted to one euro when no value arrived.Anonymized.
- Creative81 of 86installs to the lo-fi concept
For a mobile game, the lo-fi UGC-style concept took 81 of 86 installs at about $9 each, against about $21 for the polished one, on about $865 of spend. Meta pushed most of the budget to the lo-fi one, so the fair comparison is CPI, not volume, and 5 installs on the losing side is a small sample. The client's buyer ran the account. Jan produced both concepts with AI and read the results.Past engagement, anonymized.
- Creative$2.88 vs $5.76cost per install, two batches
Over five days, a conventional creative batch delivered 74 installs at $2.88 against 29 installs at $5.76 for the high-energy batch, across two apps. Small sample, and we say so.Past engagement, anonymized.
Start with a free 30-minute diagnostic
Tell us about the account. Before the call we prepare a short pre-read from public data, such as ad libraries, the store listing and the tags on your site, as a PDF or a video. On the call Jan walks you through the three biggest opportunities and the next step.
Prefer email? hello@digitalnafta.com
- 01You fill in a short form: spend, channels, goal.
- 02We prepare a pre-read from public data.
- 0330 minutes with Jan, no sales sequence after it.