Forecasting and driver-based models

A forecasting model built in Excel, Google Sheets or a connected platform like Mosaic, structured around the two or three drivers, headcount, pricing, churn, conversion or your own equivalents, that actually explain your historical results. Scenario toggles let you compare a base case against an upside or downside case without rebuilding formulas each time. Cash and runway calculate directly from the same driver inputs as revenue and cost, and the model includes a rolling update process that compares each period's actuals against the prior forecast.

We start with your historical actuals, typically twelve to twenty-four months where available, to identify which drivers genuinely explain the pattern in your revenue, cost and cash rather than assuming the standard drivers for your industry apply unchanged to your specific business. The model is built around those drivers with explicit scenario toggles, tested against a recent historical period to confirm it reproduces what actually happened before it is trusted to project what happens next. Once live, updates run on the agreed schedule, comparing actuals to the prior forecast and adjusting assumptions where the variance reflects a real change rather than noise.

Venture-backed startups needing a credible runway number for board meetings and fundraising conversations are the clearest fit, since investors specifically probe whether a forecast is built from real drivers or backed into from a desired outcome. SaaS businesses with clear unit economics, subscription churn and expansion revenue, and construction or project-based businesses where backlog and crew utilization drive the numbers, also benefit strongly. A very early business with only a few months of actuals can still get value, with the caveat that less historical data means more reliance on stated assumptions until real trend data accumulates.

The most common problem is a spreadsheet that has grown by accretion, new rows copied from old ones, broken references nobody caught, until nobody can say with confidence which cells actually drive the bottom line. The second is a forecast built once at fundraising time and never revisited, drifting so far from actuals by the next round that it damages credibility with investors who remember the original numbers. The third is cash and runway calculated separately from the revenue and cost forecast, as a rough manual overlay, so the two never quite agree when someone checks the math. The fourth is a model built by an outside consultant using logic nobody in-house understands, so any question about how a number was derived requires calling that consultant back.

Models are most commonly built in Excel or Google Sheets for full transparency and ownership, since anyone on your team can open the file and trace every formula without a subscription or a login. For businesses wanting a connected, always-current view without manually pulling fresh actuals each cycle, we build on Mosaic, which pulls actuals directly from QuickBooks Online, Xero or your billing system. The choice depends on how often you want to touch the model yourself versus how much you want automated.

Model build and setup is scoped based on business complexity and the number of drivers and scenarios needed, and is priced and published on our pricing page rather than as a one-size number. An ongoing rolling-update engagement, where we run the monthly comparison and adjustment cycle with you rather than handing the file over and stepping away, is priced separately and often pairs with FP&A or a fractional CFO engagement.

The clearest measure is forecast accuracy over time, how close the model's prior-period prediction came to what actually happened, tracked explicitly rather than assumed. A model that consistently overshoots or undershoots on the same driver gets that driver's assumption revisited rather than the whole model rebuilt. We also check that the model reproduces known historical periods correctly before it is trusted with a forward projection, since a model that cannot explain the past has no real basis for projecting the future.

At the end of the build, the file and its underlying logic are handed over in full, with documentation explaining what each driver represents and how it flows through the model, so your own finance team, controller or fractional CFO can run and update it independently. This matters most for businesses that want the discipline of a driver-based model without permanent reliance on an outside firm to touch it every month; the rolling update engagement is available for businesses that want that support ongoing, but it is a choice, not a lock-in built into the model itself.