Restaurant survival rate

Survival here is the share of restaurant attempts still licensed at each horizon. An attempt only enters a horizon’s denominator once it has had that long in which to fail, so the ten-year figure is computed over a smaller, older cohort than the two-year one. That is why the denominators fall as the horizons lengthen.

Read the short horizons with care, and we would rather say so than not. A closure only enters this record once the licence has lapsed past its renewal window, which takes about 3 years. Anything measured before then counts businesses as alive that have already stopped trading. The federal series, which is built from payroll records and has no such lag, puts two-year survival at 73.6% against our 85.7%. That gap is our measurement, not Chicago. From four years on the two agree closely.

The complement is not a closure rate. A 60.9% four-year survival does not mean 39.1% closed. Part of the remainder is attempts whose outcome is not yet observable. Closures take roughly three years to surface in the licence record, so any recent-period closure figure would be an undercount, which is why none is published.

That is also why these figures stop short of the present rather than running to today. A restaurant that closed last month has not yet lapsed past its renewal window, so the city record still shows it open, and counting it as a survivor would be as wrong as counting it as a closure.

It is also most of the reason the published failure rates disagree with each other so widely. Two companies put the first-year rate at 0.9% and 17%, and the gap is mostly definitional rather than factual.

Chicago restaurant attempts from the city licence record, excluding the grocery, pharmacy, hotel and institutional licences the same 'Retail Food Establishment' category covers.

That exclusion is larger than it sounds. Chicago’s “Retail Food Establishment” licence covers groceries, corner shops, pharmacies, hotels and school cafeterias as well as restaurants: 19,110 of 40,695 licence attempts are not restaurants and are not counted here. A survival rate computed over the whole licence record would be a different number about a different thing.

Kaplan-Meier survival via maturedSurvivalRate() in src/lib/survival-core.ts — the same code the model trainer and the product use. An attempt enters a horizon's denominator only once it has had the full horizon in which to fail.

No figure is published off a cohort smaller than 200 matured attempts; a horizon that cannot clear that bar is shown as no figure rather than as a number with too little behind it.