On this page
Learning objectivesPart 1 — The mental model (5 min)Part 2 — Reading a card (8 min)Part 3 — What actually changed (7 min)Part 4 — The Jobs page (3 min)Part 5 — When to escalate (2 min)Common questions, with answersRelated readingFor Training
A session plan for teaching this change. Audience: internal sales, account management, and admin-panel users. Not for clinic customers — nothing customer-facing changed.
Suggested length: 25 minutes plus questions. Prerequisite: the learner has seen the internal customer board before.
Learning objectives
By the end, a learner should be able to:
- Explain, in one sentence, what the renewal automation does and how often it runs.
- Read a card's score, tag and scorecard note correctly — including the two traps.
- Say what changed for churned cards, and what has not been confirmed about it.
- Find the Jobs page and tell whether the automation ran successfully.
- Recognise the three symptoms that need escalating rather than explaining away.
Part 1 — The mental model (5 min)
Open with the thing that confuses everybody, before it has a chance to confuse them.
Dentolize uses its own CRM to track its own customers. Every clinic that buys the product exists twice: once as their own tenant, and once as a card on our internal sales board. When we say "leads" in this context, we mean our customers, not the leads feature clinics use.
Then the job itself:
Every three days at 4am, an automation walks our customer board. For each customer it asks: is anyone home, and is their money about to run out? It writes the answer onto the card, and if renewal is close, it drags the card into the Renewal column.
Check for understanding: ask someone to say back what "lead" means on this board. Do not move on until they get it right — every later misunderstanding traces back to this.
Part 2 — Reading a card (8 min)
Walk through a real card on the board. Point at each element in turn.
The score, n / 10. It is a count of how many of seventeen activity types showed any sign of life in the last five days. Green above 7, red below 3.
The tag. Three possible values:
| Tag | Raw count | Meaning |
|---|---|---|
Not Working | 0 | Nothing at all happened. This is the at-risk signal. |
Working Low | 1–10 | Alive. |
Working | 11+ | Deep, broad usage. Rare. |
Trap 1 — teach this explicitly. Working Low does not mean "low engagement / at risk". It means anywhere from one signal to ten. Most healthy clinics live here forever, because reaching Working needs eleven of seventeen modules including lab orders, inventory and loyalty. Have the group say back: "the at-risk tag is Not Working, not Working Low."
Trap 2. The displayed score is capped at 10 but the tag uses the uncapped count. So two cards both reading 10 / 10 can carry different tags. When they disagree, trust the tag.
The note. Every run appends a scorecard listing all seventeen signals with ✅ or 🔴, headed by ❇️ (score rose), 🔻 (fell), or 🟡 (unchanged). Show a real one. Point out that this is a ready-made agenda for a discovery call — it tells you precisely which modules the customer has never touched.
The due date / clock icon. This is the subscription expiry, refreshed every run. Red when overdue.
Exercise: give each learner a card and ask them to state (a) is this customer alive, (b) which two modules would you ask about on a call, (c) when does their subscription end.
Part 3 — What actually changed (7 min)
State the before and after plainly.
Before: any card in the Churned column was skipped entirely. No score update, no tag update, no note, no renewal move. Its numbers froze on the day someone marked it churned. After: churned cards are treated like every other card.
Then the three consequences, in order of certainty:
- Churned cards get scored, tagged and annotated again. Certain. This happens.
- Revived customers become visible — a churned card whose score climbs is a customer who
never really left. This is the most valuable output.
- Churned cards with a live annual subscription may move to Renewal. Not confirmed.
Be straight about point 3. The automation refuses to move a card that is already in the same pipeline as the Renewal column, and nobody has confirmed whether Churned and Renewal share a pipeline on the production board. Teach it as "this may happen, engineering is confirming", not as fact. Learners who are told a certainty that turns out false stop trusting the training.
Also warn about the first run. The day this deploys, every churned card recalculates at once and a batch will change colour simultaneously. That is the backlog clearing, not a real-world event. Read the second run as signal.
Part 4 — The Jobs page (3 min)
Navigate to Admin → Jobs (/ZROYKuKEVCvQykPlS4kP/jobs). Show the companyLeadCron row and its six columns: Running, Started, Duration, Total, Error, Execute.
Teach one thing only: how to tell whether it ran. A recent "Started" and no error means healthy. Then teach the two red flags:
- Stuck showing Running. The lock has no expiry, so a job that died mid-run stays marked
running forever and never fires again. Silent — no alert.
- Any error text. The run aborts on the first failure, so every customer after that point
was skipped that cycle.
Note that "Total" dropped as part of this change — it now counts only non-disabled companies. Not a fault.
Do not teach people to press Execute. It is blocked in production anyway.
Part 5 — When to escalate (2 min)
Three symptoms. Escalate, don't explain:
- Job stuck as Running, or showing an error.
- A column's card count disagreeing with the number of cards in it.
- A card in Renewal whose plan is monthly — that should be impossible.
One symptom to expect and not escalate: the ordering of cards inside the Churned column resetting. Known defect, engineering is aware, cosmetic.
Common questions, with answers
"Does this send anything to the customer?" No. It moves a card on an internal board. No email, no message, no billing action.
"Are churned customers being un-churned?" No. The card moves to Renewal only when there is a contract event to act on. The churn decision itself is untouched.
"Why five days when it runs every three?" The windows overlap deliberately. It means a customer who goes quiet takes up to two runs — about six days — to drop to Not Working, rather than flickering.
"Why are monthly customers excluded?" By design. Monthly subscriptions renew continuously, so there is no single expiry event to prompt on. Those accounts are worked manually.
"How much notice do we get?" Between roughly four and eight weeks, depending where in the calendar month the job runs. Plus up to three days, since it runs every third day.
Related reading
- For Sales — the day-to-day playbook
- For Support — the question-and-answer version
- Feature Tour — screenshots of every screen mentioned here