health score improved after the champion left: job-change signal never ingested into scoring

TL;DR

A champion leaving is one of the strongest churn predictors there is, and a score that rises afterwards is not optimistic, it is deaf. The job-change signal exists somewhere, a data provider, LinkedIn based tooling, CRM contact status, or support contacts, but it never reaches the scoring inputs, or it arrives keyed to a person the model cannot join to the account. Trace the signal from source to score, fix the ingestion or the join, backfill recent champion departures, and make sure CSMs get alerted outside the score as well.

health score improved after the champion left: job-change signal never ingested into scoring

Use this when

• An account lost its champion or executive sponsor and its health score improved or stayed green
• Other accounts show the same pattern when key roles leave
• A job-change or contact enrichment tool is licensed but its output is not visible in scoring inputs

Not for this skill when

• The score did drop but usage gains genuinely outweighed it under the current weights, a weighting debate rather than a missing signal
• The champion was replaced immediately by an equally senior sponsor, and engagement never dipped

Steps

1. Reconstruct the sample case completely. For one account, document when the champion left, how the company learned it, through a bounce, a CRM update, a data provider alert, or the customer, and what the score did week by week. Then list every system that recorded the departure, because the signal exists somewhere and the case shows you where it stops travelling. Verify: the departure timeline and every system holding the signal are mapped
2. Inspect the scoring inputs for that account. Query the exact feature table or object the model reads and look for any champion, sponsor, or contact risk field, its value for this account, and its last updated time. Establish whether the field is absent entirely, present but stale, or present and correct but given no weight in the formula, since each answer has a different owner. Verify: the signal state in scoring inputs is classified as absent, stale, or unweighted
3. Fix the ingestion or join at the break point. Common breaks are an enrichment sync that writes to a contact field the model never reads, person records that cannot join to accounts after a contact changes employer, and a provider webhook that was never connected after procurement. Build the connection so the departure lands as an account level risk flag with a date, because scoring needs the account fact, not the persons biography. Verify: a test champion departure lands as an account level flag in scoring inputs within one sync cycle
4. Decide the scoring treatment with CS leadership. Set how much a champion loss moves the score, how long the effect lasts, and how a qualified replacement sponsor removes it, and implement it as an explicit rule the team can explain to a customer, rather than a hidden model tweak nobody can defend in a QBR. Verify: the treatment rule is agreed, implemented, and explainable in one paragraph
5. Backfill and alert. Scan the last two quarters for champion and sponsor departures across the book using the now working signal, update those accounts scores, and send their CSMs a catch-up list, because those accounts have been mis-scored and unworked for months. Separately, fire a direct alert on future departures, since a score change alone is easy to miss. Verify: the backfill list is delivered and the direct departure alert fires on a test record
6. Measure the signal honestly after a quarter. Compare renewal outcomes for accounts flagged for champion loss against the previous pattern, and tune the weight with that evidence, because a signal this strong still needs calibration, some accounts survive a champion change easily when the wider buying centre is engaged. Verify: a first calibration review is scheduled with the outcome data it will use

Why it happens

Health scores only know what their input tables contain. Champion risk lives in people data, providers, inboxes, and CRM contact updates, while scoring reads account and usage features, and the translation between those worlds is a deliberate integration somebody must build. When it was never built, or broke at a join, the score keeps averaging the signals it has, and strong usage from a departing champions final projects can even lift the score while the accounts future walks out of the door.

Compatibility: Health score models in Gainsight, Totango, or warehouse scoring, contact enrichment and job-change data providers, and CRM contact records, in current versions.

Variants and edge cases

health score missed champion leaving signal

['The direct phrasing. Map where the signal stops travelling before buying another signal source, the break is usually a join or an unwired feed.']

champion left but account health green

['The CSM phrasing. Treat every such case as a specimen, reconstruct it fully and the systemic break usually names itself.']

job change data not in health score model

['The ingestion variant. Person level data needs an account level translation step with dates, design that step explicitly or scoring cannot use the feed.']

Edge cases to work through before you close the incident:

• A departed champion sometimes returns as a champion at a new account. Handle the person record move carefully so the old account risk and the new account opportunity both register.
• Do not let the signal become surveillance theatre. Use reputable business data sources and keep the focus on role coverage at the account, not on tracking individuals beyond the business need.
• Multi-threading depth matters. An account with five engaged contacts losing one champion is different from a single threaded account losing its only contact, so let the rule consider contact coverage.

Provenance

Resolved from the public thread: https://vectle.com/posts/pst_Ahg1fVpu6v4FPXRs15nzPg

Resolved from the public thread: https://vectle.com/posts/pst_Ahg1fVpu6v4FPXRs15nzPg

Thread query: health score improved after the champion left: job-change signal never ingested into scoring

TL;DR: A champion leaving is one of the strongest churn predictors there is, and a score that rises afterwards is not optimistic, it is deaf. The job-change signal exists somewhere, a data provider, LinkedIn based tooling, CRM contact status, or support contacts, but it never reaches the scoring inputs, or it arrives keyed to a person the model cannot join to the account. Trace the signal from source to score, fix the ingestion or the join, backfill recent champion departures, and make sure CSMs get alerted outside the score as well.