Cliff, Slope, or Floor: Diagnosing a Broken Retention Curve
Objective: Given a real weekly cohort table for a SaaS product, identify which of the three retention zones (cliff, slope, floor) is broken and recommend the correct fix.
You're a growth analyst at Freshworks reviewing a new self-serve add-on's first eight weeks of cohort data before the team commits next quarter's roadmap to a retention fix.
Read the cohort table, classify the drop-off by zone, and recommend the single highest-leverage fix.
Before you start
What you'll need
Free path (everything below is enough to finish)
No account friction, sufficient for an 8-week, single-product cohort table
Paid upgrades (optional, faster/deeper)
Built-in retention reports let you re-run the same cohort split by channel without a manual export
No access? Google Sheets with a manually tagged acquisition-source column
The process
1 step
Step 01 of 01
The lesson splits every retention curve into three zones: the cliff (Day 1-7, the largest absolute drop), the slope (Week 2-8, decelerating decay), and the floor (Week 8+, where the cohort stabilizes).
This cohort loses 71% of users between Day 0 and Day 1, then decays slowly and predictably after that. Which zone is broken, and does that point at acquisition quality or onboarding?
Procedure
- Import cohort-export.csv and freeze row 1
- Chart each cohort row as a line from Week 0 to Week 8
- Measure the percentage-point drop between each adjacent pair of columns
- Flag whichever single gap accounts for the largest share of total attrition
Freshworks add-on cohort, Week-over-week drop Day 0 -> Day 1: -71 pts (100% -> 29%) Day 1 -> Week 2: -6 pts (29% -> 23%) Week 2 -> Week 4: -3 pts (23% -> 20%) Week 4 -> Week 8: -1 pt (20% -> 19%)
Healthy
The largest drop sits between Day 1 and Day 7 (the cliff), and the curve flattens into a stable floor by Week 8.
Unhealthy
A 71-point drop concentrated in the first 24 hours, with almost no further decay afterward, meaning the users who churn never experienced the product at all.
What this means
When nearly all attrition happens before Day 1 and the remaining users retain well, the product itself is fine, the problem is who is arriving. This is an acquisition-quality signal, not an onboarding signal.
So what do I do about it?
| Symptom | Action | Effort |
|---|---|---|
| 71% of a cohort disappears before completing a single session | Audit the acquisition source mix for this cohort before touching onboarding copy or flow | 30 min |
| The post-Day-1 slope is already shallow and the floor is stable | Do not fund an onboarding redesign; it is not where the users are being lost | 5 min |
Final deliverable
A one-page diagnosis memo naming the broken zone (cliff, slope, or floor), the evidence for it, and the single highest-leverage recommended fix.
See a reference example
Mailchimp automation add-on, cohort diagnosis (excerpt) ZONE: Slope (Week 2-8), not the cliff Day 0->Day 1 drop: -18 pts (mild, expected) Week 2->Week 8 drop: -34 pts (broken) DIAGNOSIS: Users complete first setup but abandon before finding a second reason to return. This is an onboarding depth problem, not an acquisition problem. RECOMMENDATION: Build a Week-2 email nudge toward the second core action, not a paid-traffic audit.
Success criteria
You're done when you can:
- Correctly identifies which zone (cliff, slope, or floor) accounts for most of the attrition
- Distinguishes an acquisition-quality problem from an onboarding problem based on where the drop concentrates
- Recommends one action, not a scattershot list