Layer 02 · Engineering + Applied intelligence
The operation became events, and routine messages started going out on their own.
An entire gym modelled as a domain and connected to the group's intelligence layer: subscription, class, training sheet, attendance.
Continuous events since 25 September 2026, with class reminders, renewals and birthdays going out on WhatsApp with no human step.
ATUO group venture · Code and Soul engineering · SYNQi³ intelligence
The context
Churn does not happen at once. It announces itself in silence.
Ōkami is an integral health centre in Itapetininga: subscriptions and invoices in Odoo, classes on the board, training on the sheet, attendance at the front desk. Everything needed to know whether a student is leaving already existed, scattered across four places that did not talk to each other. The signal was not missing. What was missing was someone connecting it to an action.
And Ōkami is the group's own venture. The brand and the experience came from ATUO, the engineering from Code and Soul, and the messaging rules live in SYNQi³. It is not a supplier delivering a module to a client: it is the whole arrangement operating inside the same business, with four divisions accountable for the same result.
The question is not who cancelled. It is who is about to, and what to do now.
How we approached it
What was built, in the order it had to exist.
None of this started with the model. It started with the domain, because you cannot infer over what is not structured, and most of the work sits in layers nobody sees.
The domain, before the model
Exercise, circuit, class, training sheet: the gym's vocabulary became a data structure. While what a student does lives in free text, there is no inference possible, only search.
Odoo stopped being the end of the line
Orders, issued invoices, class reminders and attendance started leaving as events. The ERP is still the system of record; it stopped being where the data dies.
History came in with the right time
Nineteen months of invoices loaded into the lake stamped with the real business date, not the load date. It looks like a detail and is not: without that stamp every customer is born one day old and the value model projects a multiple of what exists.
The profile lives in SYNQi³
Risk is calculated daily in Odoo, by an explicit rule, with the reason recorded. Profile and messaging rules live in the group's intelligence layer, and evolve for every business inside it.
The message arrives where the student is
The daily routine emits the event, the rule evaluates and the send is recorded with no human step. For now these are class reminders, renewals and birthdays: some reminders are still held at the entry point, and dropout risk does not trigger a message yet.
The loop closes at the front desk
The front desk records each entry by name, and attendance returns as an event. Without that return the system speaks and does not listen: what gets measured is what actually happened, not what the message promised.
Impact
Where this stands today, and what it cost to get here.
Events have been continuous since 25 September 2026. The first messages went out on 26 September, and the daily routine has run since 30 September. The effect on the churn curve has no measurement window yet, and saying so is part of the method. What cost the most was not the model. It was finding out that the system claimed to be working: a connector emitted the invoice-created event where the lake accepted invoice-paid, a participle of difference, and revenue arrived at zero with no error anywhere. The failure mode that matters is not the one that shouts. It is the one that says it worked.
Domain
training, class and subscription modelled as structure
Automatic
reminders, renewals and birthdays, from event to send, with no human step
SYNQi³
risk in Odoo, with the reason; profile and messaging rules in the intelligence layer
Closed loop
attendance recorded at the front desk returns as an event
Connections
Where this case connects within the ATUO group.
Is your operation waiting for churn to show up on the report?
If your product depends on retention, it is worth a conversation about how engineering + intelligence can act before, not after.
Engineering is not decoration. It is direction.