FP&A Today: The Chief Forecaster Will See You Now
FP&A Today Podcast · 52 min
In this episode of FP&A Today, host Paul Barnhurst sits down with Nate Kaemingk — who carries the distinctive title of Chief Forecaster at Better Forecasting — for one of the most substantive conversations available on the theory and practice of financial forecasting. The discussion covers rolling forecasts vs. annual budgets, probabilistic forecasting frameworks, driver-based modelling architecture, and what most FP&A teams are still getting wrong.
Nate Kaemingk's title — Chief Forecaster — is unusual enough to be worth explaining: it signals an organisation that has elevated forecasting to a dedicated strategic function rather than treating it as a side responsibility of the broader FP&A team. His conversation with Paul Barnhurst is one of the most rigorous and contrarian discussions of financial forecasting methodology available in podcast form — particularly his critique of how most FP&A teams measure success.
The Rolling Forecast vs. Annual Budget Debate
Nate opens with his perspective on why the annual budget, despite its widely acknowledged limitations, remains dominant in most organisations. His argument is nuanced: the problem isn't the annual budget itself — it's treating the annual budget as a forecast. When companies use the annual budget as both a commitment tool and a prediction tool simultaneously, both functions suffer. Rolling forecasts solve the prediction problem; they don't replace the need for annual commitments.
- Annual budgets work well as commitment and accountability tools
- Annual budgets fail as predictions because they can't be updated
- Rolling forecasts should be maintained separately from budget commitments
- Variance analysis: compare actuals to rolling forecast (prediction quality) AND to budget (performance accountability)
Probabilistic Forecasting: The Missing Layer
The most distinctive element of Nate's approach is his advocacy for probabilistic forecasting — expressing forecasts as distributions rather than point estimates. Where most FP&A teams produce a single revenue forecast number, Nate advocates for a revenue distribution: 'we estimate 60% probability of revenue between $X and $Y, with meaningful probability of both $Z upside and $W downside.' This approach communicates real uncertainty rather than false precision.
Driver-Based Modelling Architecture
Nate provides practical guidance on how to structure driver-based models — arguably the most actionable segment of the podcast. His framework starts with identifying the 3-5 operational drivers that explain 80%+ of revenue variance in your specific business, building a model that translates changes in those drivers directly into financial outcomes, and then focusing forecasting effort on those drivers rather than on financial line items.
- Step 1: Identify which 3-5 operational metrics explain 80%+ of revenue variance
- Step 2: Build direct driver-to-financial-outcome linkages in the model
- Step 3: Forecast the drivers, not the financial outputs
- Step 4: Use historical data to validate driver-to-outcome relationships before going live
What Most FP&A Teams Are Still Getting Wrong
Nate's critique of mainstream FP&A practice is direct and specific. Most teams, he argues, are forecasting the wrong thing (financial line items rather than operational drivers), in the wrong time increments (annual rather than rolling), with the wrong level of certainty communication (point estimates rather than distributions), and with the wrong success metrics (budget accuracy rather than decision quality).
We measure FP&A teams on budget accuracy. Then we're surprised when their forecasts are biased toward the budget. Change the success metric and you change the behaviour.— Nate Kaemingk, Chief Forecaster, Better Forecasting (FP&A Today Podcast, Nov 2024)
Practical Implementation Checklist
- Identify your 3–5 operational drivers that explain 80%+ of your revenue variance using retrospective analysis — don't guess; run a statistical analysis of the last 8–12 quarters to find the genuine drivers
- Build your rolling forecast and annual budget as formally separate processes with separate ownership — the rolling forecast is owned by the Chief Forecaster (or equivalent); the budget is owned by finance business partners and the CFO
- Experiment with probabilistic forecasting in a low-stakes context first: present your next monthly revenue estimate as a distribution ('60% probability of revenue between $X and $Y, 20% probability above, 20% below') to build stakeholder comfort before making it standard
- Challenge your success metrics: measure 'decisions influenced by finance analysis' and 'forecast accuracy vs. actuals' alongside 'budget vs. actuals' — the first two measure what actually matters for FP&A value
- After listening to this episode, read Kaemingk's articles on the FP&A Trends website for the written companion to his podcast philosophy — the articles provide the detail that a 52-minute conversation format can't fully cover
- Apply the driver-based model architecture to one revenue stream this quarter: identify the 3 drivers, build the driver model, and compare its accuracy over the next 3 months against your current method
The Chief Forecaster podcast is required listening for any FP&A professional who wants to challenge their assumptions about how forecasting should work. Kaemingk's critique of budget-accuracy as a success metric, his advocacy for probabilistic distributions, and his driver-based model architecture together constitute a coherent alternative philosophy of forecasting that is more honest, more useful, and more accurate than traditional approaches.
Key Takeaways
Rolling forecasts and annual budgets serve different purposes — maintain both rather than choosing one
Probabilistic forecasting (expressing outcomes as distributions) communicates real uncertainty vs. false precision
Forecast the operational drivers, not the financial line items — the financial outputs follow from the drivers
Most FP&A teams use the wrong success metric: budget accuracy instead of decision quality
Driver-based model architecture: identify 3–5 drivers that explain 80%+ of variance, then model those
Changing the success metric (from budget accuracy to forecast accuracy and decisions influenced) changes the behaviour — incentives determine forecasting culture
Probabilistic format: express forecasts as '60% probability between $X and $Y' — this communicates real uncertainty rather than false precision of single-number estimates
