Forecasting 101

Bessemer Atlas article, bvp.com, May 26, 2023. Practitioner interviews with Elena Gomez (CFO, Toast), Meena Srinivasan (fractional CFO, Tia; former CFO, Ginger), Rami Essaid (co-founder/CEO, Finmark), David Appel (Sage Intacct), and Jeff Epstein (Operating Partner, Bessemer).

Key Takeaways

• Forecasting is not budgeting. “Forecasts are regularly updated, whereas budgets are not.” Epstein: a detailed revenue forecast weekly or monthly; expenses at a higher level, as variances from budget.
• The best metaphor in the piece is Srinivasan’s: “Forecasting is like climbing a mountain—even when you have done everything to prepare and plan, conditions keep changing and avalanches can threaten to knock you off your course.”
• A model decays fast. Essaid: “I think a model has a half-life of about one month, meaning that every month that goes by, its accuracy decays by 50%.” Cadence should track how dynamic the business is — weekly/monthly for volatile, quarterly for steady.
• Accountant vs. FP&A, put memorably. Gomez: “It’s a lot like sports, where there’s a play-by-play commentator and a color commentator. The person who tells you what happened play-by-play is the accountant, and the color commentator who explains the forecast and future trends is FP&A. I think you need both.”
• The first forecast is not about accuracy. Essaid: “Building your first forecast isn’t about getting it right—it’s more about getting your ideas down on paper.” Its job is to make a nascent business model concrete — price per unit, ACV, headcount, runway.
• Gomez’s three framing questions: What is the goal? What are the underlying drivers of actual results? What possible scenarios do you have in mind (base, upside, downside)?
• Essaid’s five mechanical steps: collect and visualise 6–12 months of history; roll flat line items forward; interrogate why variable ones move; identify the main revenue drivers and how hard you can lean on them; understand the inputs behind each driver.
• Epstein’s stagger chart, from Andy Grove’s High Output Management: revise the remaining forecast with each new month of actuals, so the drift itself becomes visible.
• Board plan vs. management plan. Gomez: run a management forecast 5–10% ahead of the board plan, so you have room to always hit the board number while the team still chases an aggressive one.
• The qualitative layer is not optional. “The texture underneath the forecast is as important as the forecast itself.” And Srinivasan’s diagnostic: if everyone is hitting 30–50% of quota, the quota is wrong, not the team.
• Gomez’s five parting lessons: don’t focus on too many levers (three or four are the real needle-movers); learn from the past by understanding the why, not the what; get input from different stakeholders; build different models for different parts of the business; build in contingencies. The piece closes on optimism bias — humans overestimate good outcomes, so forecasts need buffers.

Connections

• Saved by Kyle under a note titled simply “Forecasting,” which fed Predicting The Future. The bridge to that draft is the gap between forecasting as a discipline (this piece: drivers, cadence, scenarios) and prediction as an epistemic problem — the practitioner literature almost never touches the second.
• The optimism-bias close connects to Behavioral Economics and the planning-fallacy material in the corpus.
• SaaS Metrics and Company Building for the operating context; the stagger chart is a High Output Management idea.