Claim and evidence
Send messages when a user's behavior changes (stalled flow, declining logins, repeated visit without action) rather than on a fixed schedule.
No dated example yet. Treat this tactic as a hypothesis.
Vendor claim
A tool vendor reports numbers about its own customers or product. Useful, but not independent.
Working · 2026-05 Standard lifecycle practice; sources are vendor guides.
What the numbers show
No published number for this tactic has passed our word-for-word check yet, so there is no chart.
Why it works
The message arrives when the user has a reason to care, so relevance is higher and noise lower. Behavioral signals show friction days before a risk score does.
How to run it
- List events that signal friction or drift
- Wire triggers from product analytics to the messaging tool
- Write one helpful message per signal
- Test against a scheduled-send control
Before / after
Generalised wireframe of the pattern, not a copy of any company's screen.
Ways to run it
Design options to test, not promised results.
- Option AIn-context reminder
Show an in-app reminder inside the flow where the signal fires, as HubSpot's Sidekick did while composing email.
- Option BFriction alerts
Trigger a helpful message after a stalled setup or repeated visits without action.
- Option CHoldout against calendar
Keep a control group on the scheduled send and compare outcomes for each trigger.
Dated examples
- 2026
HubSpot (Sidekick)
Appcues cites in-app reminders shown during email composition after auto-re-enabling tracking.
appcues.com · 2026-05-26
Does it still work?
Standard lifecycle practice; sources are vendor guides. source
Risks
- Signals need clean event tracking
- Too many triggers recreate blast fatigue
Sources
- fullstory.com
“leading signals, visible days or weeks before any risk score turns red.”
✓ quote checked word-for-word on the page - appcues.com
Change log
- 2026-11-20 — harvest v0.1 (retention)