Tracking Preflight Before You Report

Analytics properties break silently: a tag firing twice, a consent banner that blocks by default in one region, a subdomain missing from the property, an undocumented internal traffic filter. Each produces confident numbers that are wrong in one direction, and the real damage is not the broken tag but a quarter of reporting built on top of it that has to be withdrawn. Run this once per property and again after any tag manager, consent or domain change.

El archivo de la habilidad

---
name: analytics-tracking-preflight
description: Use when you are about to publish an organic number from an analytics property you have not personally validated.
---

# Tracking Preflight Before You Report

Analytics properties break silently: a tag firing twice, a consent banner that blocks by default in one region, a subdomain missing from the property, an undocumented internal traffic filter. Each produces confident numbers that are wrong in one direction, and the real damage is not the broken tag but a quarter of reporting built on top of it that has to be withdrawn. Run this once per property and again after any tag manager, consent or domain change.

## Qué necesitas antes

- admin access to the analytics property and the tag manager container
- the full list of hostnames and subdomains including checkout, help and blog where they are separate
- the consent management configuration and the current accept rate
- one week of server logs or a second data source to compare against
- the list of conversions the business actually cares about

## Método

1. List every hostname that should be in the property and confirm each one appears in a hostname report. A blog on a subdomain missing from the property is the most common single cause of organic looking smaller than it is.
2. Load a page with the network panel open and count the tag fires. Duplicate fires inflate sessions and depress conversion rate, and nothing in the standard reports will show you this.
3. Check the consent configuration and record the accept rate, then decide and write down whether your reporting is consented-only or modelled. Those answer different questions and cannot be mixed within one report.
4. Cross-check one week of organic sessions against Search Console clicks for the same pages. Expect a gap; what matters is that the gap is stable. A gap that swings week to week means something is broken, not that the tools disagree.
5. Trigger each key conversion yourself and confirm it records once with the value and channel you expect. Test the failure paths too, because a declined payment or a validation error that still fires a purchase event corrupts every number downstream.
6. Review referral exclusions and internal traffic filters. A payment provider missing from the exclusion list rewrites the channel on every converting session, which looks exactly like organic underperforming.
7. Write the known limits as a short list attached to the reporting, in the form: this number excludes X and undercounts Y by about Z.

## Qué produce esto

A one-page data quality statement covering validated hostnames, tag and conversion checks, consent basis and known undercounts, attached to every report from that property.

## Dónde falla esto

- Auditing once at setup and never again, when a consent banner or tag manager release can break comparability overnight
- Fixing a tracking fault silently, which creates a step change that gets explained months later as an algorithm update
- Comparing periods either side of a tracking fix as though they were the same measurement
- Presenting modelled conversions next to observed counts without labelling which is which

---

De la biblioteca de habilidades de QuQi - https://www.quqi.io/es/skills/analytics-tracking-preflight

Qué necesitas antes

• admin access to the analytics property and the tag manager container
• the full list of hostnames and subdomains including checkout, help and blog where they are separate
• the consent management configuration and the current accept rate
• one week of server logs or a second data source to compare against
• the list of conversions the business actually cares about

Método

1. 01 List every hostname that should be in the property and confirm each one appears in a hostname report. A blog on a subdomain missing from the property is the most common single cause of organic looking smaller than it is.
2. 02 Load a page with the network panel open and count the tag fires. Duplicate fires inflate sessions and depress conversion rate, and nothing in the standard reports will show you this.
3. 03 Check the consent configuration and record the accept rate, then decide and write down whether your reporting is consented-only or modelled. Those answer different questions and cannot be mixed within one report.
4. 04 Cross-check one week of organic sessions against Search Console clicks for the same pages. Expect a gap; what matters is that the gap is stable. A gap that swings week to week means something is broken, not that the tools disagree.
5. 05 Trigger each key conversion yourself and confirm it records once with the value and channel you expect. Test the failure paths too, because a declined payment or a validation error that still fires a purchase event corrupts every number downstream.
6. 06 Review referral exclusions and internal traffic filters. A payment provider missing from the exclusion list rewrites the channel on every converting session, which looks exactly like organic underperforming.
7. 07 Write the known limits as a short list attached to the reporting, in the form: this number excludes X and undercounts Y by about Z.

Qué produce esto

A one-page data quality statement covering validated hostnames, tag and conversion checks, consent basis and known undercounts, attached to every report from that property.

Dónde falla esto

• Auditing once at setup and never again, when a consent banner or tag manager release can break comparability overnight
• Fixing a tracking fault silently, which creates a step change that gets explained months later as an algorithm update
• Comparing periods either side of a tracking fix as though they were the same measurement
• Presenting modelled conversions next to observed counts without labelling which is which

Usa esta skill en tu propia IA

El archivo es markdown simple, con el nombre y el disparador en su frontmatter. Cuando un asistente sabe cargar skills por su cuenta, es ese frontmatter lo que lee para decidir que esta le aplica.

Más en Analítica e informes