Recipe 42 - The Forecast Accuracy Post-Mortem: Know Exactly How Wrong You Were and Why

Know Exactly How Wrong You Were, and Why, Before the Next Cycle Starts

From The Spiral Sales Office Recipe Vault | By Paul Woodley

Forecast accuracy is the only measure in hotel sales management that grades your judgment rather than your effort. You can work hard and forecast badly. You can be busy every day and still miss by twenty percent in the same direction for six consecutive months.

Most properties track accuracy. Almost nobody reviews it systematically. And so the same error repeats, because nobody has named it clearly enough to fix it.

Recipe 42 breaks that pattern. It takes the forecast you actually submitted alongside the actuals you actually produced and returns an honest read on how wrong you were, which direction you consistently lean, and the one habit change that improves the next cycle.

What This Recipe Makes

A structured forecast accuracy analysis that identifies variance size and direction by revenue stream, surfaces directional bias if it exists across multiple months, and produces a two-sentence accuracy summary and one specific habit change before the next cycle begins.

• Head Chef: DOSM with Director of Catering and Events
• When to Cook It: Monthly after close; directional bias reviewed quarterly
• Ingredients: Submitted forecasts by month at a consistent lead time; actuals for the same months; both split by revenue stream where possible

The Rule That Protects the Whole Exercise

The prompt is explicit about one non-negotiable requirement: you need the forecast you submitted, not the one you would have submitted knowing what you know now.

A forecast quietly revised on the twenty-eighth of the month and then compared to actuals is not a measurement. It is decoration. The only number that grades judgment is the one you committed to before the month closed.

If your platform does not preserve prior submissions, the prompt recommends starting a dated copy file immediately. That habit alone is worth more than the recipe over time.

The Four Outputs This Prompt Returns

1. Variance by Month and Stream - How far off the submitted forecast was from actuals, by revenue stream. Not just a total miss. A line-by-line read showing where the accuracy was good and where it was not
2. Directional Bias Analysis - Whether the misses have a consistent direction. A single miss is noise. Six months of misses in the same direction is a personality trait showing up in a spreadsheet. The bias read tells you whether you consistently forecast high or low, and in which streams
3. Root-Cause Questions - For each significant variance: did the event mix change, did historical pickup fail to materialise because the market shifted, or did personal bias pull the number in a direction the evidence did not support? These questions point at the assumption to fix rather than the outcome to apologise for
4. The Two-Sentence Accuracy Summary and One Rule Change - Written to be read out loud in a month-end review, honest about the gap and specific about the fix. Not a general commitment to do better. One rule, one report, one field, or one meeting behaviour that changes before the next cycle

Why the By-Stream Split Matters

A total accuracy figure of twelve percent off tells you something. A finding that you were three percent off on food and thirty-one percent off on beverage, every month, in the same direction, tells you something actionable.

The by-stream split is where the diagnosis lives. It points at a specific assumption in a specific category that can be investigated and adjusted. Without it, the accuracy conversation produces general discomfort. With it, it produces a specific fix.

Six Months Is the Minimum

The directional bias read requires enough history to distinguish a pattern from a single bad month. Six months is the minimum for the bias analysis to be meaningful. Reviewing accuracy for one month in isolation tells you what happened. Reviewing six months tells you how you think, and what you habitually get wrong.

The prompt notes this explicitly: if the history is shorter than six months, the bias read is illustrative rather than conclusive, and it labels it accordingly.

Where the Output Goes

The monthly critique, where the accuracy read becomes agenda context rather than a defensive explanation of what went wrong. Your own one-on-one with the GM, where honest forecasting builds more trust than optimistic forecasting that consistently misses. The accuracy log that feeds the next cycle's Recipe 39, where the bias read informs how conservatively or aggressively to position the near-in assumptions.

The Mirror

In which direction do you consistently miss your catering forecast? If you have to think about it, the answer is probably that you do not have a systematic read on it. Recipe 42 produces the read. What you do with it is the leadership question.

A Note on Data Handling

Confirm what your employer, owner, brand, and management company permit before pasting forecast and actual data into any AI platform. Monthly revenue actuals and submitted forecast figures may be commercially sensitive. When in doubt, use indexed values rather than absolute revenue. Every output is a starting position. The accuracy conversation and the habit change still belong to you.

Your Next Step

After next month closes, before you write the narrative for the critique, run Recipe 42. Pull the forecast you submitted, the actuals you produced, and the six months of history behind them. Read the directional bias output. Then make the one specific change it recommends before the next cycle starts rather than after the next miss.

Forecast accuracy compounds. The team that gets better at it each month produces fewer surprises, builds more credibility, and makes better decisions with the time they recover from explaining variances. Recipe 42 is where that improvement starts.

Paul Woodley | The Spiral Sales Office