The audit process I run with every new client: tracking, campaign structure, attribution, signal quality and coherence between marketing KPIs and real sales.
The first mistake in any struggling acquisition system is optimizing before understanding what is broken. Optimization without diagnosis is noise — it can improve platform metrics without solving the real problem and, at worst, makes something more efficient that should not exist at all.
A complete diagnosis of any digital acquisition system can be done in 48 hours. Not because it is quick — but because there is a review order that surfaces performance leaks by potential impact.
Most systems with high CAC or stalled growth have problems in one of five places: tracking, signal, campaign architecture, creative, or coherence between marketing and sales. Each has distinct symptoms and distinct fixes.
Optimizing without diagnosis is like tuning the fuel mixture when the problem is a hole in the tank. You can improve the mixture indefinitely and the car still won't reach its destination.
Check: Meta Pixel live and not duplicated, CAPI configured with acceptable event match quality (EMQ > 6.0), conversion events firing at the right points of the funnel, GTM free of conflicting rules, no event duplication inflating the metrics. Most systems that "don't convert well" have broken tracking — the algorithm cannot optimize what it cannot measure.
Check: which conversion event is set as the campaign objective, whether that event represents a real customer or just a traffic behaviour, weekly conversion volume per ad set (viable minimum: 50 per week to exit the learning phase), whether CRM data is feeding back into the platform through CAPI. A system optimizing for "thank-you page view" instead of "port-in completed" is a system optimizing toward the wrong goal.
Check: separation between prospecting and retargeting campaigns, audience segmentation by temperature (cold, warm, hot), budget allocated by performance rather than intuition, no competition between the same advertiser's ad sets for the same audience (ad set overlap), automation rules that are not interfering with learning. An account with 80% of its budget in a single unsegmented conversion campaign has no architecture — it has one ad set with a big budget.
Check: how old the live creatives are, CTR trend over the last 14 days (decay vs stability), average frequency in the main ad set, variety of formats (static, video, carousel) and message angles. A system running the same creative in the same ad sets for 3 months has accumulated creative fatigue that the CPL dashboard does not show yet — but it is already operating in degraded mode.
Check: real lead-to-customer conversion rate (not the platform's), time from lead to close, lead contactability (share who answer the follow-up), difference between leads reported by Meta and records in the CRM. A gap greater than 20% between platform leads and CRM leads is an attribution problem. A real conversion rate below 10% means the system is optimizing toward people who don't become customers.
Where the leak most often is: In 80% of the systems audited, the main leak sits in Area 02 (signal quality) or Area 05 (alignment with sales). Technical tracking is usually in reasonable shape. What fails is the signal being fed to the algorithm, and whether that signal represents real customers.
At the end of the diagnosis there are four concrete outputs:
The diagnosis optimizes nothing. It defines what to optimize and in what order, so that the optimization has the greatest possible impact.
The diagnosis is the first step. Without it, scaling only amplifies the existing problem. In 48 hours I identify what is failing and in what order to fix it.
References
Primary documentation for the platforms cited on this page:
Case figures refer to the periods and accounts described in each one; they are not promises of replicable results.
Por Cesar Baz — Performance marketing, data and AI consultant
Publicado el · Actualizado el