My bids feel solid. But three out of my last five jobs came in over budget. I've been doing this for 12 years — am I missing something?

The situation

Ray runs a residential remodeling business. He's been estimating kitchen remodels for over a decade and considers himself an accurate bidder. His estimates are detailed — he breaks down labor, materials, subcontractor costs, permits — and he adds a 10% contingency. Yet on a $28,000 kitchen job, he's finding himself losing $1,500–$4,000 regularly.

The analysis

Uncertain line items compound. A flat 10% contingency covers single overruns but not simultaneous ones — which happen on 1 in 4 jobs.

The Spreadsheet Sim models his cost structure, but instead of using a single point estimate for each line item, it assigns each one a realistic range. Lumber and materials: +/- 15% (supply chain volatility, waste). Labor hours: +/- 20% (job complexity reveals itself during work). Permit timeline: 2–8 weeks (affects carrying cost and crew scheduling overhead). Subcontractor availability: +/- 10% on plumber and electrician bids.

The critical insight is compounding uncertainty. When you model each line item independently as a right-skewed range (small downside, longer upside — the shape overruns actually take) and add them up, the combined distribution of total project cost has a long right tail. A job with a $25,000 base estimate (before contingency) has a true expected cost near $27,400 — about 10% above the estimate before Ray opens a single wall. Roughly 1 in 4 jobs runs at least 15% over ($28,750+), and the 90th percentile lands near $29,600. Ray's flat 10% contingency covers single-item overruns; it doesn't cover the compounding case that hits on 1 in 4 jobs.

The fix is a risk-tiered contingency: simple jobs (clear scope, standard materials, familiar subcontractors) keep the 10% buffer. Complex jobs (custom finishes, older homes, new subcontractors, tight timelines) get a 20–22% buffer. The Spreadsheet Sim helps Ray identify which jobs fall into which category before he bids.

The decision

Ray implements a job-complexity scoring sheet (5 criteria, 1–3 points each) that maps directly to his contingency rate, and saves his material and labor variability distributions to his library with Monitoring turned on. His margin on the next six jobs improved from -2% to +9% on average. A year later, Monitoring flags that material cost variability has drifted well past his original ±15% fit — supply chain volatility has structurally worsened — and he raises the buffer on simple jobs before a string of them start eating margin again.