Gestion d’entreprise · Budget, prévisions et scénarios
Financial planning with versioned budgets, a rolling forecast that reprojects at every close and what-if analysis to test decisions without touching the official plan. Projected income statement (DRE), cash flow and balance sheet stay tied to the same model, with scenarios, sensitivity, Monte Carlo, a multi-year plan, valuation and an objective measure of how far off the last forecast was.
O problema e a saída
Se alguma dessas situações parece familiar, é porque elas se repetem em quase toda empresa de formatura, até o sistema certo entrar em cena.
The company knows how much cash it has and how much it has approved to spend, but nobody can say how the year ends.
A rolling view that adds year-to-date actuals to the projection of what is left, redone at every close. The question "if nothing changes, where does December land" finally gets an answer with a date and an owner.
The forecast lives in a spreadsheet someone edited yesterday, so comparing it with actuals measures nothing.
Frozen, immutable versions. Reprojecting creates a new version, the previous one becomes history and is still measured. Unfreezing requires a reason and is logged with author, before and after.
The projected DRE says there will be a profit and nobody knows if there will be cash in the bank next month.
DRE, cash flow and balance sheet are a single model. Projected cash is derived from opening balance plus inflows minus outflows, the opening balance comes from the actual treasury position and the balance sheet check shows up in every accrual period.
Answering "what if dropouts rise 3 points and the CDI falls to 9%" takes two days of spreadsheets.
A scenario stores the adjustment, not a copy of the plan: fixing the base fixes it for all. A tornado chart ranks variables by impact, Monte Carlo runs with a seed so it is reproducible and the comparison view puts the base and up to four scenarios side by side.
Projected revenue is a number someone typed in and nobody can defend.
Drivers before values: classes times graduates per class times average ticket times one minus dropout rate. History comes from the system itself, and when a series is too short the method falls back to a simpler one and says so, instead of inventing a trend.
Nobody knows whether last quarter's forecast was off, in which direction, or who owned the line.
Accuracy measured at close: error in currency, relative error and signed bias, by version, by line and by owner. Chronic optimism in forecasts stops being a hunch and becomes a number.
The variance against the plan shows up at close and the explanation turns into a debate of opinions.
A profit bridge that breaks the gap down into the lines that explain it most, with a categorized justification, action plan, owner and deadline. Timing variances are kept apart from real losses: revenue that only moved to another month is not a cost cut.
Na prática
Situações do dia a dia de empresas de formatura, do problema ao resultado, usando os módulos que você vê nesta página.
Reforecasting the year depended on each department sending its own spreadsheet, and the consolidated view was ready only after the month had already turned.
The monthly cycle was opened with a window, an owner per department and a deadline. Each manager submits only their own part on the "my submission" screen, and the consolidation shows who is still pending.
The month's version is frozen on time, with a trail of who submitted and who approved, and each department's accuracy is now measured against what it projected itself.
The board was considering buying equipment early, and the discussion kept circling around "I think we can afford it".
The capex went into the plan with its cash outlay and depreciation, and a scenario adjusted dropouts and average collection period. Projected cash flow was recalculated without cluttering the working version.
The model pinpointed the exact month cash would turn negative in that scenario. The purchase was rescheduled for the following quarter, with the comparison view showing both paths side by side.
Every quarter projected revenue came in above actuals, and nobody could tell whether it was one specific line or the whole plan.
The accuracy measure ran at the close of each accrual period, with bias by line and by owner.
The error was concentrated in two add-on lines, always in the same direction. Those lines' drivers were switched to a seasonal method and the relative error fell by half the next quarter.
The expansion to two branches was drawn up in slides, with an estimated annual growth rate and no cash check at all.
The multi-year plan projected the following years by method from the annualized base year, with discounted cash flow valuation and covenants tested against the model.
The plan showed the year the debt to EBITDA ratio would breach the covenant. Financing was resized before the conversation with the bank, not after.
April closed below target and the company needed to revise May through December without rewriting months already closed or losing the forecast presented in March.
The close pulled actuals from the mapped sources, froze April and opened a new rolling version for future periods only. Each manager updated their submission and the consolidated view kept the previous version for comparison.
The board saw the original budget, the March forecast, actuals through April and the new year-end estimate side by side, with the variance explained by volume, price, cost and timing.
Before raising the package price, the company needed to know whether the increase would offset a possible rise in dropouts and in supplier costs.
One scenario changed ticket, dropouts and cost inflation without touching the base plan. The tornado ranked assumptions by EBITDA impact and the probabilistic simulation showed the distribution of ending cash and the chance of missing the target.
The decision no longer relied on a single optimistic case: the company chose the price increase range that protected margin and cash even at the most conservative percentiles.
Margin closed below forecast, but revenue, event costs and sales expenses had moved in different directions.
The profit bridge isolated the effects behind the gap, separated timing from real loss and required a justification, owner, action and deadline on the material lines.
The meeting stopped arguing over which spreadsheet was right and started deciding on three quantified causes, tracked with a target and an alert until corrected.
Como funciona
Cada etapa é um módulo de verdade, e o que sai de uma entra na outra sem ninguém digitar de novo.
Módulo por módulo
11 módulos e 96 recursos nesta página, o mesmo cadastro de turma, formando e evento sustentando todos eles.
Le plan de l’exercice en versions, chaque chiffre gardant la trace de son origine.
Le réalisé jusqu’ici, la projection à partir de là, et la frontière qui avance chaque mois.
Les trois états dans un seul modèle, avec le contrôle de bouclage sous les yeux.
« Et si l’abandon monte de 3 points et que le CDI tombe à 9 % ? » en quelques secondes, sans refaire le plan.
Le chiffre d’affaires n’est pas un nombre saisi : c’est promotions × diplômés × panier × (1 − abandon).
Investissements, effectifs et financements avec des règles que personne ne calcule juste à la main.
Ce qui s’est passé à côté de ce qui était prévu, et l’explication qui tombe juste.
De combien la dernière projection s’est trompée, et dans quel sens.
L’écran unique, et le chiffre qui va vers la personne au lieu de l’attendre.
Le plan à trois ou cinq ans, la valeur de l’entreprise et la consolidation des établissements.
Le chiffre sort du système, et le module s’entretient tout seul.
Compare os módulos e escolha a configuração que acompanha sua operação.