Key takeaway
A health score is an average of evidence collected at different times. Product telemetry may be minutes old, a support ticket may be three weeks old, and a sponsor comment may date from the last business review. Averaging those inputs without dates produces a number that looks current and describes no particular week.
Freshness is a property of each signal
A health score is an average of evidence collected at different times. Product telemetry may be minutes old, a support ticket may be three weeks old, and a sponsor comment may date from the last business review. Averaging those inputs without dates produces a number that looks current and describes no particular week.
Freshness review treats each signal as a dated claim. Before weighting anything, the CSM records the event time, the retrieval time, the timezone, and the source system for every input. The purpose is not to perfect the score. It is to prevent a stale input from dominating a decision that will be defended in a customer conversation.
Set source-specific age thresholds
Different sources decay at different rates. Usage data older than seven days should not be described as current adoption. Support history older than thirty days should not be presented as an active risk. Stakeholder sentiment older than one review cycle should be treated as reported history, not present attitude.
Write the thresholds into the health definition, not into individual judgment. When a signal exceeds its threshold, the platform marks it stale and the CSM must either refresh it or state the limitation in the account note. A threshold everyone can inspect is fairer than a rule applied only when someone remembers.
Separate event time from load time
Most confusion comes from conflating when something happened with when the warehouse received it. A nightly job that fails on Friday and backfills on Monday makes a Saturday event look new. The review should record both timestamps and display the event time in the customer timezone and the analysis timezone.
When the two diverge by more than the source threshold, treat the input as late rather than fresh. Label the gap. A late signal can still inform the review, but it cannot be used to claim that a condition is current, and it should not trigger an automated playbook on its own.
Keep disagreement visible in the review
A single account often carries contradictory evidence. Usage grows while the executive sponsor raises a concern about value. Two business units report opposite experiences. The review should present both states side by side with their dates rather than collapsing them into one color.
The CSM writes a short conflict note: signal A, signal B, what each measures, and which decision depends on the difference. The account owner resolves the conflict in the review and records the rationale. That note is more useful to the next reviewer than any score, because it shows where the judgment actually happened.
Record the scoring version and exclusions
Health calculations change. A denominator gets redefined, a segment is excluded, a weight is adjusted after a bad quarter. Without a version, a chart that moves in March cannot be compared with a chart from January. The review records the calculation version beside every classification.
Exclusions deserve the same treatment. If inactive seats were removed from an adoption rate, the note says so and states the size of the excluded group. Exclusions are legitimate when declared. They become misleading when the resulting trend is presented as comparable to a prior period that used a different base.
Route stale signals before the meeting
The CSM should not discover a stale source during the review. Two business days before the session, the platform lists every account where a required signal exceeds its threshold. The CSM refreshes what can be refreshed and flags what cannot, with a reason.
Unrefreshable signals get an owner and a date. Telemetry blocked by a customer firewall goes to the technical account manager. A sponsor conversation that has not happened goes to the account owner with a proposed date. The review then starts with known gaps rather than discovering them in front of the customer.
Translate freshness into customer language
Internal freshness rules should not leak into customer conversations as jargon. Instead of saying the adoption figure is stale, say the last complete usage export covers the four weeks ending on the fifteenth and the current month is still reconciling. That is specific, checkable, and honest.
The same discipline applies to success plans. When a health classification changes, the customer should hear which evidence changed, on what date, and what the team plans to check next. A classification without its evidence invites the customer to treat the score as an opinion.
When a source goes dark
A source can stop updating without raising an error. The dashboard keeps showing the last successful load while the freshness marker quietly ages. Add a heartbeat check that compares the latest event time with the expected cadence, and treat a missed window as missing data rather than as a stable value.
When a source is dark, the CSM records the outage, the affected accounts, the fallback source, and the date the gap began. Customer communication states what is known and what is not. A health conversation built on a silent outage is worse than one that admits the number is unavailable this week.
Review the review itself each quarter
Once a quarter, sample ten accounts and check whether the freshness rules were applied. Count how many classifications rested on a signal beyond its threshold, how many conflicts were recorded, and how many stale inputs were refreshed before the meeting rather than after.
Use the sample to adjust thresholds and data contracts, not to grade individuals. If the same source is chronically late, the problem is the pipeline. If the same step is skipped under pressure, the problem is the checklist. Both are fixable, and both stay invisible when only the final score is reviewed.
Common questions
What timestamps must a health signal carry?
Event time, retrieval time, source system, and timezone. Event time drives the freshness rule; retrieval time explains late or backfilled data.
Who sets the freshness thresholds?
The customer success operations lead with the account leadership team, documented in the health definition and reviewed each quarter against sampled accounts.
How should a stale signal be handled in a review?
Refresh it before the meeting when possible. Otherwise label it stale, name the gap owner and date, and do not describe it as current evidence.
Does a stale signal invalidate the health classification?
Not automatically, but the classification must state which inputs were stale and how the owner's judgment accounted for the gap.
Sources and context
Sources provide public context. Their inclusion does not imply endorsement of this site or a guaranteed business result.
Apply this to your customer journey
Define a support role with visible boundaries.
Bring one recurring workflow. We will help map the approved inputs, review owner, evidence, and escalation points.
Book a free consultation