Leaver Reason Analyzer

The biggest slice is often the least informative one

A reason for leaving is recorded from what somebody chose to tell HR on their way out. A person who wants a good reference has every reason to say “personal reasons”, “a better opportunity”, or nothing at all. So the usual pie chart ranks a polite non-answer alongside “pay” and “my manager”, and it is frequently the largest slice.

Here the non-answers are counted on their own and never ranked as a reason. The headline gives the most common real reason beside the share of people who gave no usable reason, so it is always clear how much of your turnover the ranking actually describes.

Not everybody chose to go

Dismissals, redundancies and fixed-term contracts ending are reasons somebody left, not reasons they chose to, and they answer a different question. They are counted, and kept out of the ranking. If your file has its own voluntary column it is used, and anybody it calls voluntary whose reason reads as a dismissal or redundancy is pointed out, because one of the two columns is wrong.

The same reason typed a dozen ways

“Better pay”, “Salary”, “More money” and “Compensation” are one reason recorded four ways, each looking small. Every recorded reason is placed in a theme by the words in it, and every placement is listed with a menu to change it. Anything the tool cannot place is treated as unspecified rather than guessed.

Early leavers and long servers leave for different things

Somebody who goes in their first ninety days and somebody who goes after five years are rarely describing the same problem, and those problems usually have different owners. With hire dates in the file, reasons are broken down by length of service, so one chart does not average two different stories into one neither group would recognize.

Most year-on-year changes are noise

Pay rising from 12% to 18% of reasons sounds like a trend and may be eight people. Each theme’s share this year is compared with last year’s using Fisher’s exact test, which is exact at the small sizes exit data comes in, with a correction for testing every theme at once. A change that does not clear that bar is labeled as within chance, and the tool says so when a big-looking move is really a handful of people.

Small groups are not broken down

A department with three leavers and a breakdown of their reasons identifies the three people who gave them. So a reason is only compared across departments where at least five voluntary leavers gave one, the same threshold the Survey Cut Analyzer uses for the same reason. Where a reason turns up far more in one department than elsewhere, it is tested before it is reported.

What this does not do

A recorded reason is evidence about why people leave, shaped by what they were willing to say. It is not a verdict, and this tool does not pretend otherwise. For how many people are leaving and at what rate, use the Turnover Analyzer.