Tag Governance Is the Job Nobody Wants to Own

I asked a client's team who owned their GTM container's tag governance, and got three different answers from three different people, none of whom were actually doing it. Marketing assumed engineering owned it because it's technically a code deployment surface. Engineering assumed marketing owned it because the tags served marketing's own measurement needs. The container itself had 60-plus tags, a third of which nobody could confidently say were still needed.

This ownership gap isn't an accident of this one client, it's structural. Tag management tools were built to let marketing deploy measurement changes without needing an engineering release cycle, which was genuinely useful. But that same self-service design means tags accumulate without ever passing through a process that would naturally assign an owner, the way a code review or a deploy pipeline would for anything else touching production.

What governance actually requires, concretely

• A named owner for the container, not a team — a specific person accountable for what's in it, even if multiple people have edit access.
• A quarterly audit cadence where every tag gets checked against whether it's still firing, still needed, and still correctly configured.
• A lightweight approval step before a new tag goes to production, even if it's just a Slack message, so new tags don't get added invisibly.
• A deprecation process — when a vendor relationship ends or a campaign wraps, the tag gets removed as part of that process, not left to rot.

Tag governance fails not because nobody could do it, but because it sits exactly in the gap between two teams who each have a reasonable argument for why it's the other team's job.

What actually fixed it for this client

We assigned a single named owner — someone on the marketing ops side, with a standing monthly thirty-minute review with an engineering contact — and that alone cut the container from 60-plus tags to under 40 within two quarters, just from actually looking at what was there with the explicit job of deciding whether each one should stay. Nobody needed new tooling. They needed someone whose job it actually was.