More than most companies calculate. The salary is only the visible part. Add the hiring cost of the original search and the replacement, the manager time spent on performance conversations, the work that did not get done, and sometimes a good teammate who leaves. The cause is usually a weak hiring process rather than one person's mistake.
Is a bad hire a people problem or a process problem?
Usually a process problem. Blaming the recruiter is common, but a poorly written JD sets the wrong expectations, weak filtering lets the wrong candidates through, and interviewers who have not been briefed evaluate for the wrong things. Any one of these can produce a bad hire, and together they make one likely.
How long does a bad hire stay?
In smaller companies, a bad hire is often noticed within months. In larger organisations, it can last years, because more layers of process sit between the person and accountability. By the time it surfaces, the cost has built up.
What does a bad hire really cost?
Start with what you can count: the cost of the original hire and of the replacement search. The cost-per-hire calculator gives you that figure, and the attrition rate calculator shows how often it happens.
Then add what does not show up on a report: the work that did not happen, the manager's hours spent on performance conversations, and the colleague who left because of a difficult teammate. In companies still building HR processes, these losses are rarely tracked.
How can you reduce bad hires?
You cannot remove them entirely, but you can reduce them. Build the hiring process around a precise JD, targeted reach, structured screening and consistent interviews with a clear brief for every interviewer. Relevancy scoring helps decide who to interview first. AI phone interviews ask every candidate the same questions. Assessments give you evidence before a person decides.
Then set clear goals, milestones and early check-ins for new joiners. The sooner you read the signal, the lower the cost. Sometimes not making a hire is the right call.