Build a reviewable subscription-renewal risk list
Prioritize accounts for informed outreach using transparent evidence, uncertainty, account value, contract timing, and human context. Start with the what one row represents and definitions, calculate only from eligible records, and check the result against the source before using it.
Why this gets messy
Usage, support, billing, CRM, and contract data have different grains and missingness. A low health score can reflect seasonality, implementation stage, data loss, or genuine disengagement.
The short version
- Prioritize accounts for informed outreach using transparent evidence, uncertainty, account value, contract timing, and human context.
- Training features must precede the prediction date, account histories must remain together during evaluation, and model performance must be reported by relevant segment.
- A risk score should not automatically change price, service, or access; it supports outreach and investigation, not an adverse decision.
From the field · Customer-success operations manager
Prioritize accounts for informed outreach using transparent evidence, uncertainty, account value, contract timing, and human context.
Usage, support, billing, CRM, and contract data have different grains and missingness. A low health score can reflect seasonality, implementation stage, data loss, or genuine disengagement.
Columns used in the example
account_id · renewal_date · contract_value · usage_week · active_users · key_action_count · ticket_count · invoice_status · outcome · owner_note
Follow it with your own file
- Create time-windowed account features using only evidence available before each historical renewal.
- Define renewed, churned, downgraded, and unresolved outcomes without future leakage.
- Compare baseline rules and classification models with recall, precision, calibration, and segment behavior.
- Present risk factors and uncertainty to account owners, then record outreach and reviewed outcomes.
Check: Training features must precede the prediction date, account histories must remain together during evaluation, and model performance must be reported by relevant segment.
A method you can follow
- Define the decision: Create time-windowed account features using only evidence available before each historical renewal.
- Prepare the evidence: Define renewed, churned, downgraded, and unresolved outcomes without future leakage.
- Calculate the result: Compare baseline rules and classification models with recall, precision, calibration, and segment behavior.
- Review the exceptions: Present risk factors and uncertainty to account owners, then record outreach and reviewed outcomes.
Frame the decision before calculating
Prioritize accounts for informed outreach using transparent evidence, uncertainty, account value, contract timing, and human context.
The exception that can reverse the conclusion
A risk score should not automatically change price, service, or access; it supports outreach and investigation, not an adverse decision.
Clues in the source data
- The source files do not share one documented what one row represents or cutoff.
- The decision owner cannot trace the headline result to eligible source records.
- A risk score should not automatically change price, service, or access; it supports outreach and investigation, not an adverse decision.
Choose the depth of analysis the decision deserves
| Approach | What it answers | What remains at risk |
|---|---|---|
| Quick spreadsheet total | A narrow descriptive question | Definitions, duplicate rows, timing, and exceptions may be hidden |
| Reviewed analysis | The stated decision with reconciled evidence | Human judgment is still required for policy and action |
| Ongoing monitored workflow | The same decision across new periods | unexpected column changes and changed business rules still need review |
Final checks before using the result
- what one row represents and eligible population documented
- Dates, units, currencies, and identifiers validated
- Training features must precede the prediction date, account histories must remain together during evaluation, and model performance must be reported by relevant segment.
- Exceptions retained for human review
- Published result links back to its evidence
Doing these steps in Prepyr
Upload a CSV working copy in Prepyr. Describe the result you need: Build a subscription renewal risk list from customer data. Check the columns selected, preview any proposed changes, and compare the output with the source before applying or exporting it.
Prepyr steps used here: ML Classification · Model Comparison · Feature Importance · Report Drafts
A decision-ready result for “Build a subscription renewal risk list from customer data,” with its assumptions, exceptions, and source totals still visible.
Questions readers usually ask next
Can this customer-success operations manager analysis be done in Excel?
Yes for a small, stable file. Keep the source unchanged, document formulas and exclusions, and check outputs against the source.
What should not be automated?
Do not automate the final policy or business judgment. A risk score should not automatically change price, service, or access; it supports outreach and investigation, not an adverse decision.