_Most business debates happen at the wrong level of precision._ --- ## The precision trap Someone estimates a feature will take seven weeks. Someone else says eight. The discussion burns twenty minutes on a difference that doesn't matter — both estimates are probably wrong anyway, and wrong in ways that dwarf the gap between them. This happens constantly. Traditional ROI calculations look rigorous. Score impact from 1-10, estimate effort in story points, divide one by the other, rank the results. Clean. Quantified. Wrong. The problem is noise. A feature scored at 60 impact and 4 effort looks great — until it actually delivers 48 impact and takes 6 effort. The ROI halves. And that's a modest error. Real estimates routinely miss by 2x or 3x. When your inputs have 50% error bars, debating the second decimal place is a waste of time. The precision is fake. Worse, it feels real — spreadsheets full of numbers create false confidence in conclusions that don't survive contact with reality. --- ## Fermi thinking Enrico Fermi was famous for estimation problems. How many piano tuners in Chicago? How many gas stations in Los Angeles? He didn't guess. He decomposed: population × cars per person × fill-ups per month ÷ capacity per station. Each component is easy to estimate within an order of magnitude. Multiply them and you're usually within 3x of reality. The business version: stop estimating in fine increments. Use powers of ten. Is this worth £1K/month, £10K/month, or £100K/month? Not £47K — that's false precision. Pick a bucket. Is this two days, two weeks, or two months? Not "seven to eight weeks." If you can't tell whether something is closer to two weeks or two months, that uncertainty is the information. The buckets don't need to be exact. What matters is that adjacent options become absurd. The debate shifts from false precision to genuine disagreement. --- ## What changes When you think in orders of magnitude, three things happen. Real disagreements surface. When one person says £10K and another says £100K, that's worth discussing. When one says £47K and another says £52K, it isn't. Fermi buckets reveal the debates that matter. Decisions become robust to error. If you estimate £10K and it's actually £15K, your decision probably doesn't change. The thinking survives normal estimation noise. Fine-grained estimates don't — they flip conclusions on small errors. Speed increases. Most numbers are trivial to estimate at order-of-magnitude. You don't need a spreadsheet. You need a whiteboard and five minutes. Spurious precision slows you down for no benefit. --- ## The underlying skill This is number sense applied to decisions. Not arithmetic — intuition for scale. Knowing that a 10x error matters and a 1.5x error usually doesn't. Knowing when to calculate and when to estimate. Knowing that the goal isn't the right number — it's the right decision. Most decisions don't need precision. They need to be robust to the uncertainty that's already there. Order-of-magnitude thinking accepts the uncertainty instead of hiding it behind false precision. If the answer changes when your estimate is off by 30%, you don't have an answer. You have a guess dressed up as analysis. --- **See also:** [[Ideas/Base Rates|Base Rates]] · [[Ideas/Bayesian Probability|Bayesian Probability]]