From 20 Years Experience, 10 Lessons on Financial Forecasting
What twenty years of financial modeling taught me.
Early on, I built credibility through complexity. Multi-tab spreadsheets with hundreds of inputs, sensitivity analyses following statistical distributions, thousands of outputs, and forecasts stretching twenty years out.
I could command a boardroom with the stories these models told. I believed the numbers.
Then I checked the receipts. Short-term? Decent. Long-term? Garbage. Accuracy varied wildly by industry maturity, but the pattern held: complexity didn’t equal precision.
80/20 Rule
Also known as pareto’s principle: roughly 80% of effects come from 20% of causes. Find the vital few, ignore the trivial many.
A decade later, I stripped it all down.
Fewer inputs. Fewer assumptions. Leaned hard into the 80/20 rule—find the variables that actually move the business, ignore the rest.
The 80/20 principle comes from Vilfredo Pareto, an Italian economist who noticed that 80% of Italy’s land belonged to 20% of the population. The pattern showed up everywhere he looked. Today we call it the Pareto Principle: roughly 80% of effects come from 20% of causes. In financial modeling, that means most of your forecast accuracy comes from a handful of key drivers. Nail those, and the rest is noise.
The simpler models weren’t perfect. But they performed better. And teams could actually use them without a decoder ring.
The lesson: less is more. Here are ten things I wish I’d known earlier.
- Understand the goals of the model. Whether revenue, cash flow, market share, P&L, headcount, or volume, design a simple model focused on the core output. And resist the tendency to build the all-in-one model.
- Think modularly, for example, inputs, calculation engine, and outputs.
- Resist complex embedded formulas at first. Further, make it easy for others to edit and understand the logic.
- Use writing as an analogy to modeling. Work quickly to write a first draft, or MVM, a minimal viable model. And then begin an editing process until you create a structured and logical result.
- Subtractive is better than additive. Sculptors start with a block and chisel away until they execute their vision. I attempt to do the same. Start with knowns and chisel away based on confident subtractions. The additive approach, or bottoms-up, can provide a nice check and balance but tends to overshoot reality.
- Start with a shorter duration with realistic assumptions, then expand. Working backward from a target is valuable, but it often leads to disconnected results — so do both. Working the model from the goal also helps to assess input sensitivity.
- Show your results as scenarios; base case, moderate, and aggressive. Don’t fall in love with a single number or scenario.
- Focus on the inputs rather than argue about the outcomes — make it easy for reviewers to challenge inputs and see the results in real time.
- As confidence builds around the model, add new modules or consolidate formulas to reduce the size. Redo the editing step.
- Start from scratch with your models, or at the very least, start with other’s models that you trust and wholly understand.
Modeling anything takes a bit of art and science and most importantly – reps.
Book a Call
If you like what you see, we think you’re gonna love what you hear. Book a first consultation with us, and together we’ll figure out how to make your life a little better.
