Legal tables
The INSS, IRRF, family-allowance brackets, and the ceiling that payroll uses in its calculation. They're valid per pay period: running payroll for one month with another month's table gets the deduction wrong without flagging an error — check the validity period before closing.
The screen
The real screen, in the theme you picked above.
What you find on this screen
A portrait of the screen, read from the screen itself — it changes when the screen changes.
Available actions
Compute outcomes nowGenerate baseline from the pastReassessTurn offTurn back on
The list shows
TemplateVersionBrier (model)BaselineLast calibrationStatusHashActions
Form fields
ModeloComo o modelo é chamado. É por este nome que ele aparece no parecer e na auditoria.
MétodoTécnica usada para chegar ao número (regressão, sobrevivência, tábua).
VersãoVersão publicada do modelo. Número diferente é modelo diferente, mesmo com o mesmo nome.
BaselineResultado do palpite ingênuo usado como piso de comparação.
BrierErro médio das previsões. Quanto menor, mais o modelo acerta; acima do baseline, o modelo não serve.
Última calibraçãoQuando o modelo foi reajustado por dados novos. Modelo sem calibração recente decide com o passado.
SituaçãoSe o modelo está aprovado para uso, em observação ou suspenso.
Motivo da situaçãoPor que o modelo está nessa situação. É o que o comitê lê antes de liberar de novo.
Base legalNorma ou parecer que autoriza usar este modelo na decisão.
Start here
What this screen answers
Which statistical models are in use, which method each one uses, how accurate they are compared with a simple guess, when they were last calibrated and under which legal basis they operate.
What this screen is
The auditable sheets (model cards) of the actuarial core models: documented method, baseline always in view, Brier of the latest calibration and status. A model that does not beat the baseline is turned off; a model that loses calibration (drifts) is turned off automatically — and the product says so. Here governance reassesses calibration from forecasts with outcomes and turns each model on/off.
How to
Brier vs. baseline: the model’s grade
It measures the forecast error against the error of a simple reference. A model that does not beat the reference is not helping — it is adding complexity and risk with no gain. The highlighted champion is the one with the lowest error among the active models.
Settle outcomes without waiting for the job
A forecast only closes after its deadline passes and the actual result is observed. The daily job does this assessment, but you don't need to wait: the "Assess outcomes now" button observes the overdue forecasts and records the outcome immediately. "Generate past baseline", in turn, rebuilds from the closed periods what the naive forecast would have said — the yardstick every model has to beat. Useful when the calibration looks empty: almost always what is missing is an assessed outcome, not a model.
Rules
Deriva: quando o mundo muda e o modelo não
A model calibrated last year may have stopped being valid without anyone noticing — it does not throw an error, it just starts getting more wrong. Reassessment recalculates the calibration against what actually happened and adjusts the status automatically.
Turning a model off has a chain effect
Credit score, provisioning, outcome-based pricing and reserves draw from here. Turning it off is the right move when the model has drifted — but be aware that the screens that depend on it start operating without that input, and they say so plainly instead of faking a number.
What this screen answers
If the actuarial layer is turned off by parameter, this screen says so clearly instead of showing an empty panel that would look like "no models".
Campeão e desafiantes: quem manda no número
When there is more than one active model, the screen highlights the champion — the best-calibrated active model (lowest Brier against the baseline). The other active models are challengers: they keep being calibrated and compared, and can take the spot when reassessment shows a better Brier. "Reassess" recalculates the calibration from forecasts with outcomes and applies the automatic status; "Turn off" and "Turn back on" record the reason in the status column.
Frequently asked questions
Which question this screen answers
Which actuarial numbers can be trusted right now — and whether any model lost calibration and was turned off. Each row is a model card: documented method, version, Brier of the latest calibration next to the baseline, and the status with the reason written out. A model that does not beat the baseline or that drifts is turned off, and the screen says why on the card itself — governance reassesses calibration from forecasts with outcomes and turns each model on/off.
Was this article helpful?
In the same module: Simulate before deciding16
Investment AnalysisAssesses whether a project/contract/purchase pays for the cost of capital and creates value. Discounts the pr…Team journeyWho's most engaged: adoption by role, the week's scoreboard, and who needs a nudge.CFO CalculatorsThree everyday financial decisions solved on the spot: how to amortize a loan, compare rates on different bas…CenáriosTeste a decisão antes de tomá-la e veja o efeito no resultado e no caixa.Scenarios (what-if)Compare up to 3 futures side by side: hiring, losing a client, adjusting price — and see the effect on result…Comparar cenáriosCaminhos lado a lado, com o pior caso à mostra.Dossiê de investimentoO parecer que um comitê exigiria, em uma página.Margin ForecastingShows where your margin is heading: projects revenue, cost, and margin for the next 6 months based on the rea…
Tenka Academy
Explore the AcademyStep-by-step learning paths for your industry, inside the platform.