[{"data":1,"prerenderedAt":4},["ShallowReactive",2],{"article-body-every-business-case-is-a-forecast-nobody-scores":3},"\nStrip the ceremony away and a capital allocation process is a\nforecasting operation. Every business case that reaches an\ninvestment committee is a forecast wearing a template: revenue\nreaching this level by year three, synergies of that amount by\nmonth eighteen, payback inside six quarters. The committee's\napproval is a bet on those numbers, frequently a large one.\n\nThen ask the question that embarrasses almost every organisation:\nwhen was an approved case last scored, systematically, against what\nactually happened? Not the mega-projects with formal post-completion\nreviews, and not the disasters that force a write-down, but the\nordinary flow of cases through which most capital actually moves.\nFor the typical enterprise the honest answer is never. The forecast\nis argued about intensely for weeks before approval and examined\nnever afterwards.\n\n## What an unscored forecast system breeds\n\nForecasting theory is unambiguous about what happens to prediction\nwithout feedback, and capital allocation exhibits every symptom:\n\n- **Optimism compounds.** A case is a document written to win an\n  approval, by a sponsor rewarded for winning it. Without outcome\n  scoring, the optimistic assumptions that win approvals are never\n  invoiced, so each planning cycle learns that optimism works.\n- **Credibility never reprices.** The sponsor whose last four cases\n  overshot by 40% arrives at the committee with the same standing\n  as the one who delivers. The committee cannot weight what it\n  never measured, so seniority and narrative fill the gap.\n- **Estimation errors repeat.** Integration costs underestimated the\n  same way every acquisition; ramp curves optimistic the same way\n  every product launch. The errors are systematic, which means\n  they are correctable, which means the correction has been\n  available and unused for years.\n- **The counterfactual disappears.** The projects not funded\n  because a padded case crowded them out are the invisible cost,\n  and it is plausibly the largest one.\n\n> A committee that never learns which of its forecasters to believe\n> is not allocating capital. It is refereeing storytelling.\n\n## Closing the loop\n\nThe repair is a discipline, not a reorganisation, and it borrows\ndirectly from how mature forecasting operations run:\n\n1. **Extract the forecast from every approved case** at the moment\n   of approval: the revenue curve, the cost line, the milestones,\n   with dates and owners, held as structured commitments rather\n   than paragraphs in a slide deck.\n2. **Track actuals against them automatically**, from the systems\n   where actuals already live, so scoring is a pipeline rather\n   than an annual archaeology project.\n3. **Publish the variance record**: by case, by sponsor, by\n   assumption type. The point is not blame; ranges and honest\n   uncertainty should be welcomed. The point is calibration, in\n   the same sense a forecaster is calibrated. Over many cases,\n   promised outcomes and delivered outcomes should converge.\n4. **Feed the record back into approvals.** A new case arrives\n   accompanied by the sponsor's scored history and the\n   organisation's known bias on that assumption class, so the\n   committee debates a calibrated number rather than a fresh\n   story.\n\nThe second-order effect is the valuable one. Once sponsors know\ncases are scored, the cases change before submission: ranges\nreplace point promises, contingencies surface, and the padding\nmigrates out of the numbers because it no longer pays. The\ncommittee starts receiving something closer to the truth, which\nwas the entire objective.\n\n*Fund the forecast, score the forecast, and let the record decide\nwho gets believed next time.*\n\nStructured case commitments, automatic actuals tracking and scored\nsponsor calibration are what the Prophesee Finance Suite's\nCorporate module brings to capital allocation. Score your last ten\ncases. [Start here](/contact).\n",1786833835746]