Which Forecasting Workflow Should I Use?
Forecasting is not one spreadsheet job. The right workbook depends on whether you are creating a forecast, monitoring forecast quality, handling sparse demand, translating volume into staffing, or planning liquidity.
Boardflare’s published forecasting templates deliberately cover those different jobs rather than presenting one automatic model as universally appropriate.
Choose by the decision you need to make
| If you need to… | Start with | Why |
|---|---|---|
| Build a regular demand forecast and compare understandable baselines honestly | Demand Forecast & Backtest Explorer | Uses rolling-origin evaluation to compare seasonal-naive, recent-average, and trend-plus-seasonality models before selecting a forward forecast. |
| Forecast an item with many zero-demand periods | Intermittent Demand Forecast for Spare Parts | Preserves zero-demand periods and compares Croston/SBA methods with a recent-mean baseline before translating the result into reorder implications. |
| Diagnose whether existing forecasts are accurate or systematically biased | Forecast Accuracy & Bias Monitor | Aligns forecast vintages to realized target periods, then reviews WAPE, signed bias, MASE, horizons, series, and exceptions. |
| Turn a short-horizon workload forecast into a staffing estimate | Workforce Volume Forecast | Separates historical volume forecasting from visible productivity, shrinkage, and service-buffer assumptions used to calculate staffed shifts. |
| Project cash headroom under planning scenarios | Cash Flow Forecast & Liquidity Outlook | Fits transparent receipt/payment trends, applies worksheet scenarios, and reports cash balances, downside bands, and liquidity-threshold breaches. |
The important distinction: forecast creation versus forecast governance
The Demand Forecast & Backtest Explorer, Intermittent Demand Forecast, Workforce Volume Forecast, and Cash Flow Forecast all create forward-looking values, but they solve different business problems and use different assumptions.
The Forecast Accuracy & Bias Monitor does something else: it evaluates forecasts that already exist. Use it when the operating question is not “what should the forecast be?” but “where are our forecasts wrong, biased, or degrading by horizon?”
That distinction matters because a forecasting process needs both model evaluation before a model is selected and performance monitoring after forecasts are used operationally.
Start with the simplest method that matches the data
These templates intentionally favor transparent baselines and visible diagnostics over automatic complexity:
- For regular monthly demand, compare simple baselines with time-aware backtests before introducing more complex models.
- For sparse demand, keep zero periods in the history and use methods designed for intermittent events rather than treating the series like ordinary continuous demand.
- For operational staffing, keep workforce-policy assumptions separate from the demand model so planners can see which changes come from volume and which come from staffing policy.
- For cash planning, combine a transparent statistical baseline with explicit management scenarios instead of presenting an uncertainty band as a guarantee.
- For forecast monitoring, align vintage and horizon correctly before calculating portfolio or series-level accuracy metrics.
Keep the workbook as the operating surface
Across all five examples, Excel keeps the durable business inputs and review outputs while the notebook owns the iterative or stateful Python work. That division is intentional: a technical author can maintain the model while another workbook user operates visible assumptions and reviews results.
For a handoff workflow, validate the workbook in Excel, save the intended presentation, and have a second user complete the operating task without author intervention. App mode and sharing explains that workflow and its trust boundaries.
What these examples do not claim
The published templates are teaching and application-pattern examples, not a single production forecasting platform. They do not automatically account for every promotion, holiday, hierarchy, stockout, causal driver, financing event, scheduling constraint, or forecast-governance policy.
Use each template’s validation and limitations sections to decide what must be extended for a real operating model.