[{"data":1,"prerenderedAt":4},["ShallowReactive",2],{"article-body-the-naive-baseline-is-beating-you":3},"\nThere is a forecasting audit that costs nothing, takes a week, and\nis more informative than most vendor evaluations: compute the naive\nforecast, same as last year adjusted for trend, for every item you\nplan, and measure your actual forecasting process against it, over\nthe last two years of history you already possess.\n\nFew organisations ever run this test, and the published evidence\nsuggests why the silence is comfortable. The forecast value added\nliterature, and the field's long-running competitions, keep\nproducing the same finding: simple methods are embarrassingly hard\nto beat, and a large share of enterprise forecasts, after the\nstatistical model, the planner's adjustments and the consensus\nmeeting have each had their turn, land worse than the free\nbaseline. The human half of that chain has its own uncomfortable\nevidence base ([half the forecast is a person, and nobody scores\nthat half](/insights/half-the-forecast-is-a-person)).\n\n*Half the forecasting effort in industry is spent losing to a\nnumber that costs nothing to produce.*\n\n## What the naive test actually measures\n\nThe naive baseline is not a straw man. It is the operational\ndefinition of \"no forecasting process at all\", what you would know\nwith zero software, zero analysts and zero meetings. Every step of\nyour actual process, model, override, consensus, exists to add\naccuracy on top of it, and forecast value added is simply the audit\nof whether each step does:\n\n- **The model versus naive**: does the statistical layer beat last\n  year plus trend, item by item? For stable, fast-moving items it\n  usually does. For intermittent, lumpy and short-history items,\n  frequently not, and knowing which is which redirects both the\n  modelling effort and the software spend.\n- **Each human step versus its input**: did the planner's\n  adjustment improve on the model? Did the consensus meeting\n  improve on the planner? Measured per step, the process reveals\n  where it adds value and where it ritually subtracts it.\n- **The whole chain versus naive**: the final, uncomfortable\n  number. If the end-to-end process loses to the free baseline for\n  a third of the portfolio, that third is being planned at a\n  premium price for negative value.\n\n> Accuracy that does not beat the naive baseline is not accuracy.\n> It is expensive noise with a review cadence.\n\n## Why this precedes any technology decision\n\nThe current market moment makes the test urgent rather than\nmerely hygienic. AI forecasting claims are everywhere, vendor\nbenchmarks are self-administered, and accuracy improvements of\n8 to 20% are routinely cited, usually without stating the\nbaseline they are measured against. A buyer who has never\ncomputed their own naive benchmark has no way to price any of\nthose claims: an \"85% accurate\" system may be brilliant or may be\nlosing to last-year-plus-trend, and the sales material will not\nvolunteer which.\n\nRun the audit first and every subsequent conversation changes.\nYou know which segments of your portfolio have forecastable\nsignal and which do not, so you can demand segment-level proof\nrather than portfolio-level averages. You know what your current\nprocess genuinely adds, so a vendor must beat your reality, not\nyour anxiety. And you have the discipline that should govern the\nsystem after purchase: every model, every adjustment and every\nmeeting continuously scored against the baseline, with the steps\nthat fail retired without sentiment, because the test that\njustified the purchase keeps running in production.\n\nOne more outcome deserves preparation: some demand is not\nforecastable beyond naive, by anyone, with any technology, because\nthe signal is not in the data. The right response to\nthose items is not a better model but a different operating\nposture, faster response, more flexible supply, honest buffers,\nand only the naive test tells you which items those are.\n\n*Before you buy a forecast, find out what forecasting nothing\nwould achieve. Everything you pay for is measured from there.*\n\nContinuous forecast value added, per item and per process step,\nwith the naive baseline enforced as the floor, is standard in the\nProphesee Supply Chain Suite's demand module. Run the audit on\nyour own history. [Start here](/contact).\n",1786833838479]