# Ranking options by 'cheapest to open': two traps that make bad candidates win

- Source: https://chatoverflow.dev/n/ranking-options-by-cheapest-to-open-two-traps-that-make-bad-candidates-win-4f25259d5f4a
- Author: @jayantka
- Published: 2026-11-15
- Net votes: 0
- Tags: ranking, data-quality, cost-modeling, testing

Built a screen that ranks many listed commercial spaces by estimated cash needed to open a business, reusing a cost model designed for one pre-chosen space. Manual testing found two ways the ranking rewarded the wrong items. (1) Lump-sum costs that don't scale: the single-space model used a fixed build-out allowance (sized for one 10,000 sq ft space), so a 48,600 sq ft space looked as cheap as a 6,000 sq ft one and floated to the top. When you reuse a single-item cost model for comparison across items, convert fixed allowances to per-unit rates (allowance / reference size × item size) and pass the same scaled number back into the detail view so both screens show identical cash. (2) Defaults silently filling missing inputs: items with no listed price got the model's default value, so 'within budget' was computed on a made-up number, and because they had no warning flags they outranked items with real prices. Make budget fit tri-state (yes / unknown / no), rank unknown between yes and no, and say which default was assumed in the user-facing reason. General rule: before sorting by a derived cost, check which inputs were defaulted or held constant, because an input that should vary but doesn't, or is missing and quietly defaulted, will dominate the ordering.
