Applications-to-Offer Calculator

A job search has two decisive conversions: applications that become interviews, and interview processes that become offers. Multiply them and you have the share of applications that end in an offer. Invert that and you have the number of applications an offer costs, on average, at your current rates.

A 6% interview rate and a 20% offer rate give a combined rate of 1.2%, or about 83 applications per offer. The sensitivity is worth noticing: lifting the interview rate from 6% to 9% — a change at the very first step — cuts the expected volume by a third, without touching anything about how you interview.

The expected value is a long-run mean. It is not the typical case, and it is not a number of applications after which an offer is due. If each application has a 1.2% chance, then reaching the average of 83 leaves a substantial probability of still having nothing.

That is why the calculator also shows the volume required for a given chance of at least one offer:

The even-odds row is the median: the point at which an offer is more likely than not. The 95% row is the point at which going without one would be genuinely unusual, and it is normally far larger than people expect. Seeing the gap between those two numbers is usually more useful than the average itself.

The volume table treats every application as an independent attempt with the same probability. This is the standard simplification, and it is the weakest thing on this page.

Your applications are correlated. They share a resume, a market segment, a level of seniority, and a moment in the hiring cycle. If the resume is the problem, it is the problem in all of them at once. Correlated attempts need more volume than the formula suggests to reach the same confidence, not less — so read these figures as a lower bound.

The assumption also holds your rates fixed. In a real search they move: interview practice tends to raise the offer rate, and fatigue tends to lower application quality. Neither effect is modelled here, because modelling it would mean inventing coefficients nobody has measured.