Mirror month

This is the screen to open when you have five minutes and want to know how the company is really doing. Four quadrants answer four different questions: what happened, what lies ahead, what was promised and how much your forecasts tend to miss.

The internal credit score uses the same criteria a bank analyst would use: debt coverage, liquidity and cash generation. It lets you know the answer before asking for the money, not after.

The lowest balance over the next 13 weeks. If at any point cash crosses zero, the date appears along with how much it would cost to cover the gap through advance funding — because the real decision is not "I will go negative", it is "I will pay this much not to".

Commitments appear here when a scenario is approved. As long as nobody approves anything, the quadrant stays empty on purpose: promising is a deliberate act.

Selic, IPCA and the dollar are fetched from the Central Bank once a week, and the screen always says where each number came from. If your company works with its own assumption — because it negotiates locked-in exchange rates, for example — it overrides the official one, and the screen shows that. An environment assumption without a declared source is a guess dressed up as data.

The list of approved scenarios with how much of each promise was actually delivered. It is the agenda for the quarterly review: a company that approves optimistic scenarios and never meets them finds that out here, looking at its own history — which is much harder to ignore than a speech.

A system that only shows projections is impossible to challenge: any number seems reasonable. Displaying the model's own error next to the projection is what separates a tool from an opinion with decimal places. If the error is high, you will know — we do not hide it.

Before any projection, the module assesses the input: contracts with validity periods, receivables and payables with due dates, posted payroll, history of closings. The score appears at the top when it is below 60. It is not bureaucracy: with weak records the projection is born weak, and the error you would see on the scoreboard months later would be largely due to the records, not the model. Each item points to the screen where it can be fixed.

When the group view is turned on, the companies of the same tenant add up. What one group company charges another is eliminated before adding up: without this, the same service invoice between headquarters and branch appears twice and the group's revenue inflates without anything visibly going wrong.

This is the normal state for someone who has just turned on the module. Open Predictability and run the X-Ray: it reconstructs the months already closed and the scoreboard starts to exist right away, without waiting six months.