Blog · AI and self-service analytics
What is wrong with the red-amber-green account health field most CRMs carry, what a computed health score is made of, five inputs from the ledger and the CRM with stated weights, how the weights are set from the company's own churn history rather than opinion, why the score is shown with its inputs and never alone, the identity that every account has one score from the same rule, and how the score is validated against what happened.
Most CRMs have an account health field with three colours, set by the rep when the account was created, and it is green on accounts that left last year. A computed score from the ledger and the CRM is a different thing: the same rule for every account, inputs the reader can see, weights from what actually preceded churn, and a quarterly check that it predicts anything. This guide sets out the five inputs, the weights, the display, and the validation.
| Input | Source | Scaled how | What it catches |
|---|---|---|---|
| Run-rate trend | Ledger | This quarter's run rate ÷ trailing four | Shrinking spend |
| Contact recency | Activities, tiers | Days since qualifying touch ÷ tier cadence | Neglect |
| Products used vs norm | Ledger, norms | Products used ÷ segment norm | Shallow relationship |
| Ticket signal | Ticket export | Surge ratio, or repeat rate vs base | Friction |
| Pipeline slip | Pipeline snapshots | Slip count on open deals | Stalled expansion |
Each scaled to a range, each with a threshold below which it counts against the account.
| Input below threshold | Share of churned accounts, two quarters prior | Share of retained accounts | Weight |
|---|---|---|---|
| Run-rate trend | 74% | 18% | High |
| Contact recency | 61% | 30% | Medium |
| Products vs norm | 48% | 35% | Low |
| Ticket signal | 39% | 9% | High |
| Pipeline slip | 22% | 14% | Low |
The weights follow the separation. Stated on the report; recomputed annually.
| Account | Score | Run rate | Contact | Products | Tickets | Slip | Act on |
|---|---|---|---|---|---|---|---|
| 4471 | 31 | −24% | 71 days vs 45 | 2 vs 4 | Surge 4.7× | 2 | Contact and the surge |
| 2210 | 78 | +4% | 12 days | 4 vs 4 | 0 | Nothing | |
| 9034 | 55 | −6% | 30 days | 3 vs 4 | Repeat 41% | 1 | The repeats |
The score ranks the list. The inputs say what to do at each account.
every account has exactly one score, from the same rule, on the same date
An account missing an input is scored on the rest with the missing input shown, not silently defaulted to healthy.
| Score band a quarter ago | Accounts | Churned or shrank since | Rate |
|---|---|---|---|
| Bottom 20% | 240 | 71 | 30% |
| Middle 60% | 720 | 58 | 8% |
| Top 20% | 240 | 7 | 3% |
Ten times the rate at the bottom against the top. The score is working. If the bottom band were at 9 percent, it would not be, and the weights would be revisited before the score was shown again.
Colour code kept beside the score. Two versions; the rep trusts the colour.
Weights from opinion. The score is the workshop's guess, with decimals.
Score without inputs. A number nobody can act on.
Never validated. Green on the accounts that left.
Mapped once, the ledger, the activities, the norms, the tickets and the pipeline snapshots produce the five inputs, the score and the validation table for every account. Covirage builds this from the exports as they are. The sales insights solution describes the setup, and the surge list guide covers the input that separates churn most sharply at most companies.
From history: for accounts that churned or shrank in the last two years, which inputs were below threshold in the two quarters before. An input present in 70 percent of churn cases and 20 percent of healthy ones carries more weight than one present in both equally. The weights are stated, and they are recomputed annually.
Because a score of 42 tells an account manager nothing to do. A score of 42 with run rate down 20 percent, no contact in 60 days and two repeat tickets tells them three things. The score ranks; the inputs act.
Each quarter, the accounts that scored in the bottom band a quarter earlier are checked: how many churned, shrank or stayed. If the bottom band's churn rate is not well above the top band's, the score is not working, and the weights are revisited. A score that is never validated is a colour code with decimals.