[{"data":1,"prerenderedAt":4},["ShallowReactive",2],{"article-body-half-the-forecast-is-a-person":3},"\nEvery enterprise forecast that matters has two authors. The first is a\nmodel, whose error is measured to two decimal places, tracked in reviews,\nand litigated in software selections. The second is a person, who adjusts\nthe model's number before anyone uses it, and whose contribution is\nmeasured, in most organisations, never.\n\nThe evidence on that unmeasured half is not flattering. Fildes, Goodwin\nand De Baets, analysing roughly 147,000 forecasts across six studies,\nfound judgemental adjustments improved accuracy for only just over half\nthe items examined, with upward adjustments notably more likely to make\nthe forecast worse ([International Journal of Forecasting,\n2024](https://www.sciencedirect.com/science/article/pii/S0169207024000736)).\nA coin toss, purchased at the price of your most experienced people's\ntime.\n\n*We measure the model obsessively and the person not at all, and the\nperson is half the forecast.*\n\n## Why the overrides exist, and why they should\n\nThe wrong conclusion is that adjustments should be banned. The planner\nadjusts because she knows things the model cannot: the promotion was\ncancelled yesterday, the customer is switching, the port strike will bite\nin week three. None of it is in the master data, and all of it belongs\nin the forecast. Some overrides are the most valuable information in the\nwhole planning process.\n\nThe problem is that the valuable overrides and the destructive ones are\ncurrently indistinguishable, because nothing records which was which.\nThe optimism nudge that pads the number to match the target travels\nthrough the same unmeasured channel as the genuine intelligence about\nthe cancelled promotion. Both are labelled \"experience\".\n\n## Scoring the human half\n\nThe remedy is neither trust nor prohibition. It is measurement, the same\ncourtesy extended to the models:\n\n- **Record every adjustment as a decision**: who, when, direction, size,\n  and the stated reason, captured at the moment of the override rather\n  than reconstructed later.\n- **Score it against what the model alone would have done.** Forecast\n  value added is an old, simple idea: did the human step improve on the\n  input it received? Run item by item, it separates the planner whose\n  market knowledge consistently beats the model from the adjustment\n  ritual that consistently subtracts value.\n- **Feed the verdicts back.** The planner with a strong record gets her\n  overrides weighted up and her reasons mined for signals the model is\n  missing. The adjustment class that reliably fails, typically the\n  small, frequent, upward nudge, gets retired without ceremony.\n\n> Nobody's judgement is confiscated. Its results simply become\n> visible, which is exactly what happened to the models.\n\nOne cultural objection deserves an answer: scoring people feels\npunitive. In\npractice the effect runs the other way. Unmeasured judgement is cheap to\ndismiss, which is why planners spend review meetings defending their\nnumbers. Measured judgement with a track record is authority. The\nstrongest argument a planner can bring to a forecast review is not\nseniority but a scored history of beating the model.\n\n*Judgement is not the enemy of the forecast. Unmeasured judgement is.*\n\nProphesee records overrides as decisions and scores the human half of\nevery forecast alongside the machine half. See what your overrides have\nbeen worth. [Start here](/contact).\n",1786786820103]