Build a Week-1 Account Health Score
Objective: Given a raw table of Week-1 user activity, build a simple weighted health score from the lesson's five leading indicators and flag which accounts need proactive outreach.
You're a lifecycle marketer at Klaviyo. The support team wants a list of at-risk new accounts to reach out to before Week 2, not after churn has already happened.
Turn five raw leading-indicator columns into one weighted score, then set a threshold for proactive outreach.
Before you start
What you'll need
Free path (everything below is enough to finish)
A formula-based weighted sum needs no analytics platform for a 12-account pilot
Paid upgrades (optional, faster/deeper)
Once the weighting is validated on a small sample, a product analytics tool can compute it continuously without a manual export
No access? Google Sheets, re-run weekly on a fresh export
The process
1 step
Step 01 of 01
The lesson lists five leading indicators that predict retention before the curve turns down: time to first value action, features adopted in Week 1, session frequency in Days 2-7, support tickets in Week 1, and invite/share actions.
Given raw Week-1 data for 12 new accounts across these five columns, how do you combine them into a single score instead of eyeballing five separate numbers per account?
Procedure
- Import week-1-activity.csv with the five raw indicator columns
- Normalize each column to a 0-3 scale (fast time-to-value = 3, slow = 0, etc.)
- Apply weights: time-to-value 30%, features adopted 30%, session frequency 25%, support tickets -10%, invite action +15%
- Sum into one health score per account and sort ascending
Klaviyo new-account health score (excerpt) Account TTV Features Sessions Tickets Invite SCORE Acct 118 0.9 1 2 3 0 1.4 <- flag Acct 204 2.8 3 3 0 1 2.9 Acct 331 1.2 1 1 2 0 1.1 <- flag
Healthy
A small number of accounts land clearly below the threshold, giving support a short, prioritized outreach list instead of all new signups.
Unhealthy
Every account scores in the middle band, meaning the weights or scale aren't discriminating between healthy and at-risk behavior.
What this means
The score's value isn't the number itself, it's whether it separates accounts enough to make outreach worth prioritizing over just calling everyone.
So what do I do about it?
| Symptom | Action | Effort |
|---|---|---|
| Two accounts score below 1.5 out of a possible 3 | Route those two accounts to customer success for a Week-2 check-in call | 30 min |
Final deliverable
A weighted health-score spreadsheet with a defined outreach threshold and the list of accounts below it.
See a reference example
Freshworks trial cohort, Week-1 health score (excerpt) Acct 07: TTV 3.0, Features 2, Sessions 3, Tickets 0, Invite 1 -> SCORE 2.8 (healthy) Acct 15: TTV 0.5, Features 0, Sessions 1, Tickets 4, Invite 0 -> SCORE 0.6 (FLAG: outreach) Threshold set at 1.5. 2 of 14 trial accounts flagged for proactive CS outreach before Week 2.
Success criteria
You're done when you can:
- All five leading indicators from the lesson are represented in the score, not just one or two
- Weights are stated explicitly, not hidden inside an unexplained formula
- A clear numeric threshold separates accounts needing outreach from healthy ones