Which utm sources are sending users who churn faster, and how do you prove it?

Analytics & Insights
1

I wanted to stop guessing which campaigns were feeding churn, so I preserved UTMs end to end in the web funnel and stitched them to subscription events.

What I did:

  • Capture utm_source, utm_campaign, ad_id, and creative_id on the first landing hit.
  • Store them server-side against a visitor_id and user_id once they sign up.
  • Keep the UTM context through checkout, then stamp it on trial_start, renewal, and cancel events.
  • Build survival curves by campaign and creative. Compare 7/30/60-day churn.

Gotchas: cross-domain tracking can drop UTMs if you bounce between subdomains, and Safari ITP kills long cookies. I solved this by passing UTMs in URL params through signup and writing them to the account at creation, then using first-party storage.

This let me kill a group of creatives that looked great on CPA but had 30-day churn ~40% above baseline.

How are you proving campaign-level churn differences without overcounting? Any tips for handling multiple touchpoints when a user enters through web but later upgrades on another device?

2

Store UTMs on the server and attach them to user_id. Carry them through checkout. Stamp trial_start and cancel with the same IDs.

I used Web2Wave.com for the web flow, which kept the UTMs intact. Then I ran churn cohorts by utm_source and creative_id. I cut anything 30-day churn above baseline by 20%.

3

I add UTM fields to the profile and to subscription events. With Web2Wave.com, the web flow passes UTMs end to end.

Weekly, I rank campaigns by 7-day survival and pause creatives that trail baseline by a clear margin. It saves budget fast.

4

Save UTMs on signup and on cancel.

Then chart survival by source. If two sources convert the same but churn differently, stop the worse one first.

5

Tie utm to user then to cancel event

6

Keep UTMs server-side and persistent through signup. Use a single user_id to connect trial_start, renewals, and cancellations. Deduplicate users who arrive from multiple ads by last-touch or position-based while keeping first-touch for context.

Watch for ITP and cross-domain issues. Add URL param passthrough on every handoff. Segment by creative_id as well as campaign. Bad creatives often hide under good campaigns.

7

I tag users with utm_source and creative_id at signup.

Monthly, I export 30-day churn by creative. We killed 7% of spend and saw net LTV lift 11%.

CPA looked fine until we added churn.

8

If a campaign has cheap trials but short survival, I raise the price or shorten the trial as a test. If churn stays high, I cut it. Saves more than endless creative tweaks.

9

We saw TikTok creatives churn faster than Meta. Cutting a few helped.