Prism

Metrics Library

Prism Customer Success Handbook — Section 1.8

Two rules that govern this page

1. Every metric has an action-if-off-target. A metric with no defined response is a vanity metric — it gets reported, nodded at, and changes nothing. If we can't say what we'd do when the number moves the wrong way, the metric doesn't belong here.

2. Outcome metrics and diagnostic metrics are kept separate. Outcomes are what CS is accountable for. Diagnostics are leading indicators that explain why an outcome is moving. Confusing the two is a classic failure: a leading indicator that becomes a target gets gamed, and stops being informative exactly when it's needed most.

Ownership: the CSM owns every metric below — there is only one. The Director of CS reviews outcomes monthly and the full set quarterly. A per-metric owner column would be noise at this company size.

Outcome metrics

What CS is accountable for. Reviewed monthly with the Director; assessed properly at quarter end.

Metric Definition & formula Source Target If off target
Gross revenue retention Renewed ARR ÷ ARR available to renew, in the period. Excludes expansion. (v2 — see Not measured below) Accounts, renewal outcomes 90%+ Churn review by root cause (4.5 Churn Review Template); check whether losses cluster in a segment, cause, or cohort
Renewal rate (count) Accounts renewed ÷ accounts up for renewal Accounts 90%+ Compare against GRR — a high count rate with low GRR means we're losing large accounts
Time to value Median days from handoff to milestone 4 (Recurring) Onboarding playbook (v2) Under 90 days Identify which milestone is the bottleneck; the fix is usually structural, not motivational (3.6 New Customer Onboarding)
Onboarding success rate Accounts reaching Value Realization ÷ accounts started, per cohort Onboarding playbook (v2) 85%+ Onboarding failures are churn 12 months early. Examine where cohorts died — milestone 3 is the usual answer

Diagnostic metrics — leading indicators

These explain the outcomes above, early enough to act. Reviewed weekly or monthly. Not targets to hit — signals to read.

Metric Definition & formula Source Cadence If it moves wrong
At-Risk rate Accounts in At-Risk state ÷ active accounts (1.6 Health Scoring Methodology) Health snapshots, playbooks Weekly Rising rate: check whether it's real deterioration or trigger noise (2.3 Operating Rhythm, system review)
At-risk ARR Sum of ARR for accounts in At-Risk state Accounts, health Weekly Weight triage toward the concentrated accounts (2.4 Account Prioritization Framework)
Weighted revenue forecast Σ (account ARR × forecast category weight), by month (2.6 Renewal Forecasting) Forecast categories Monthly Falling forecast with flat health means judgment is seeing something the data isn't — investigate what
Silence rate Accounts with no logged two-way contact in 90 days ÷ active accounts Activities Monthly Expected to be high for SMB by design. Rising for MM/Enterprise means cadence is slipping (2.2 Meeting Standards & Working Norms)
Single-threaded rate Accounts with one live contact ÷ active accounts, by segment Contacts Monthly Each one is a champion departure away from crisis (3.4 Champion Departure). Target deliberate widening on the largest
Milestone 3 by day 30 New accounts live on the portal within 30 days ÷ new accounts Onboarding (v2) Monthly The single best predictor of first-year retention for this product. Falling means onboarding intervention is coming too late
Accounts with no champion Accounts with no contact designated champion Contacts Monthly Forecast cannot be Commit for any of these (2.11 Exception Handling)

Operational metrics — is the system working?

These measure the CS function itself rather than the customers. Read at the quarterly system review (2.3 Operating Rhythm), where they drive changes to triggers, weights, and playbooks.

Metric Definition & formula Source What it tells you
Forecast accuracy Actual outcome vs. forecast category at T-60, by category Forecast history, renewals Commit accounts that churned mean the evidence standard wasn't applied. At Risk accounts that all renewed mean the category is over-applied and its weight is too low (2.6 Renewal Forecasting)
Playbook completion rate Playbooks completed ÷ playbooks opened, by type Playbooks Consistently low for one type means that playbook is wrong for the situation it fires on
Playbook abandonment Playbooks stale over 30 days ÷ active playbooks (3.0 Playbook Index) Playbooks Either the playbook is wrong or capacity is genuinely short. Both matter; neither is visible any other way
Trigger precision Playbooks that fired on accounts subsequently found fine ÷ playbooks fired Playbooks, outcomes A trigger that cries wolf gets ignored, which is worse than no trigger
Unknown-at-T-60 count Accounts inside the renewal window still forecast Unknown Forecast, renewal dates Target: zero. Any instance is a process failure to name, not a data gap (2.6 Renewal Forecasting, 3.7 Renewal Motion)
Escalation comms compliance Escalations where the required update cadence was met ÷ escalations (2.7 Escalation Management) Activities, escalations The commitment most easily broken under load, and the one customers remember
Three-week roll count Items deferred three consecutive weekly reviews (2.4 Account Prioritization Framework) Prioritization Chronic deferral means the tier is wrong or capacity is short

Churn analysis

Reviewed quarterly (2.3 Operating Rhythm), using the post-mortem template (4.5 Churn Review Template).

Metric Definition Why it's split this way
Churn by root cause Count and ARR, grouped by the six root-cause categories in 3.8 Churn & Offboarding Stated reason is nearly always "price." Root cause is what's actionable
Preventable vs. not Churn attributable to something CS or the product could have changed, vs. their business changing A function that counts every loss as a failure can't learn from the ones that were. Only preventable churn generates a process change

Not measured — and why

Stating this explicitly matters as much as the metrics themselves.

Not measured Reason
Net revenue retention Requires expansion and contraction history the data model doesn't hold. Rejected in 1.2 CS Operating Model as fabricated-looking rather than approximated. Becomes computable once expansion is tracked as a first-class record (v2)
NPS / CSAT Survey coverage across a tech-touch SMB book would be too sparse to read. A number derived from a handful of responses invites confident conclusions the sample can't support
Ticket CSAT Same sparse-coverage problem; rejected as a health input in 1.6 Health Scoring Methodology for the same reason
Ticket volume Deliberately unscored (1.6 Health Scoring Methodology) — ambiguous in both directions. High volume can mean engagement or dysfunction; low can mean smooth or dead
Activity counts (calls made, emails sent) Measures effort, not outcome. In a solo-CSM model with no one to compare against, it produces activity theatre. What matters is whether milestones moved and accounts renewed
Time in lifecycle stage No lifecycle stage field exists yet (1.3 Customer Data Model) (v2)

On activity counts specifically: the temptation is real, because they're easy to compute and always available. The reason they're excluded is that they answer "was the CSM busy?" — a question nobody should be asking. The operating rhythm (2.3 Operating Rhythm) already guarantees the work happens on a schedule; counting it adds nothing and distorts behavior.

Review cadence summary

When What's reviewed
Weekly At-Risk rate, at-risk ARR — inputs to the risk review and triage (2.2 Meeting Standards & Working Norms, 2.4 Account Prioritization Framework)
Monthly Outcome metrics with the Director; all diagnostic metrics; forecast pass
Quarterly Operational metrics and churn analysis at the system review (2.3 Operating Rhythm). Output is concrete changes to triggers, weights, or playbooks — or an explicit decision to change nothing

Not yet computable

Metrics above marked (v2) depend on records that don't exist in the CSP: onboarding milestones as tracked records, lifecycle stage, expansion and contraction history, and forecast history with reasons. All are in BUILD_BACKLOG.md.

Until then, this page defines the target measurement model. The metrics that are computable today — At-Risk rate, at-risk ARR, health distribution, playbook counts, renewal volume — run off the Executive Dashboard.