utm-tracking-hygiene

name

utm-tracking-hygiene

description

Audit whether a Google Ads account's ad URLs carry consistent UTM tags so GA4 and the CRM can attribute each click back to the campaign / keyword that drove it. Checks presence, scheme consistency, and dynamic utm_term across campaigns, then recommends a consistent UTM scheme + auto-tagging. Use this skill whenever the user audits a Google Ads account, asks about UTM tags, URL tagging, campaign tracking, attribution, "can my CRM see which campaign drove this lead", auto-tagging / GCLID, lost source data, or untagged ad links — even if they don't say "UTM". Judges URL TAGGING HYGIENE only; conversion-action setup belongs to `conversion-tracking`, and GA4 configuration is a different connector.

UTM Tagging Hygiene

Goal (job-to-be-done)

Answer the client's question: are paid-search clicks tagged so the downstream tools (GA4 and the CRM) can attribute them back to the campaign / keyword that drove each contact? The unit of analysis is the campaign (its final URLs). The output is a tagging verdict + the exact scheme to standardize on, so downstream attribution — and the offline-conversion loop — can see the source.

• Who: media buyer / PPC manager / marketer. When: account audit / onboarding health-check.
• Decision it drives: add UTMs where they're missing, standardize an ad-hoc scheme, and turn on auto-tagging — so a member in the CRM can be joined back to a campaign and keyword.
• The differentiator: not "is the conversion event set up" — that's conversion-tracking. This reads the tags on the actual ad URLs and flags whether they're present, uniform, and dynamic.

Scope

• ✅ Presence + consistency of UTM params on final URLs, dynamic utm_term, auto-tagging adoption.
• ❌ Conversion-action setup (is offline/CRM import on, do conversions carry values) → complementary conversion-tracking skill.
• ❌ GA4 configuration (data streams, events, attribution model) — a different connector.

Components (read these references as needed)

• Tools / data plan: references/tools.md — the one campaign-grain query + the parsed UTM fields and final URLs.
• Framework / rubric: references/framework.md — the brain: the presence → consistency → dynamic-term ladder and the auto-tagging recommendation.
• Output schema: references/output.md — the JSON this skill emits.

Operate

Input: per campaign — the parsed UTM params (utm_source, utm_medium, utm_campaign, utm_term, utm_content) + cost_micros from one query, and the raw ad_group_ad_ad_final_urls from a second query (the utm_* and final-URL fields can't be combined — join by campaign_name). These params are parsed straight from the final URL by the google-ads connector. No business context needed beyond knowing the account uses GA4 / a CRM for attribution.

Process: apply the rubric in references/framework.md. For each campaign: (1) presence — are the UTM fields populated at all (all-blank = no tagging)? (2) consistency — is the scheme uniform across campaigns or ad-hoc/per-person? (3) dynamic term — is utm_term a dynamic keyword value (e.g. {keyword}) rather than a static string? Rank by spend so the highest-spend untagged / inconsistent campaigns surface first. Recommend auto-tagging (GCLID) plus a consistent manual UTM scheme — this also unblocks the offline-conversion loop in conversion-tracking.

Emit the JSON in references/output.md:

• synthesis — the canonical three strings: headline (the biggest tagging gap on the highest spender + the fix), diagnosis (where attribution leaks — untagged clicks the CRM/GA4 can't trace to a campaign/keyword), action (the one fix to make now, where / what / why).
• campaigns[] — one per campaign: has_utms, scheme_note, the verdict, the action, and a recommendation {where, what, why}.

A renderer (the orchestrator's formats/*) turns this JSON into the human document. Emit pure data — no emojis, tables, markdown, or colors in the output.

Voice (don't copy the rules, link them): write every narrative line per ../../../_framework/writing.md — the question the data answers yes/no; the metric carried as data, never spelled out in prose; first sentence answers it, then names the driver; one bridge line to the next section. Plain language for a non-technical owner ("the CRM can't see which campaign sent this lead because the link has no tags"), the technical setting (auto-tagging / GCLID / {keyword}) in parentheses.

Example (illustrative — NOT rules)

• Untagged: every final URL on the top-spend campaign is bare (no utm_* populated) → untagged → enable_autotagging + add_utms: GA4 and the CRM can't trace these clicks to a campaign.
• Inconsistent: half the campaigns use utm_campaign=brand and half use utm_campaign=Brand_2026 → inconsistent → standardize: the CRM splits the same campaign into two rows.
• Static term: utm_term=ppc hard-coded on every ad instead of the keyword → inconsistent → standardize toward a dynamic {keyword} value so each contact carries the keyword that drove it.