The organisation that collects leaving reasons is the one being left — and that shapes every answer it receives.
Exit interviews and leaver surveys are the primary instrument for recording why women leave tech roles, and both share the same structural flaw: they are administered by, or on behalf of, the employer. A woman who is leaving because of a manager, a pay decision, a blocked promotion, or a pattern of exclusion has little incentive to say so clearly to the institution responsible for her reference. Responses migrate toward the neutral — personal circumstances, career change, a new opportunity — and the real precipitant quietly disappears from the data.
- Exit interview / leaver survey
- records the stated proximate reason at the moment of a managed departure, not the structural cause
- Lateral move / internal reclassification
- a woman who shifts into a non-technical role within the same employer generates no exit record; the technical workforce loses her silently
- Reference asymmetry
- the employer holds the reference and the final appraisal at the point data is collected; this shapes what a leaver will say
- Inflection point
- the moment a woman stopped expecting to progress, which may precede the resignation by months or years and is not captured by exit instruments
This is not a failure of individual honesty. It is a rational response to asymmetric power at the moment of departure. The employer holds the final appraisal, the reference, and sometimes a notice-period negotiation. The leaver holds almost nothing. Under those conditions, answers to "why are you leaving?" are not evidence about causes; they are evidence about what feels safe to say.
The result is a systematic bias in the direction organisations least want corrected. Mid-career attrition among women in computing is substantially higher than among men at the same career stage, yet exit records at many organisations show no corresponding spike in reported dissatisfaction. The gap between what the numbers show and what is actually happening is precisely the gap that the collection method creates.
Under those conditions, answers to "why are you leaving?" are not evidence about causes; they are evidence about what feels safe to say.
A second distortion comes from who is counted at all. Women who leave a specialism but stay in the same organisation — moving sideways into project management, operations, or business analysis — typically generate no exit record whatsoever. They have not resigned; they have reclassified. The technical workforce loses them, the attrition statistics do not catch them, and the causal chain that produced the move stays entirely invisible. Understanding how that internal drift works requires a different kind of data: role-change histories, promotion-flow analysis, and longitudinal tracking that most organisations do not routinely build.
There is also a timing problem. Exit data is collected at the point of departure, when the decision is already made and often long settled. The actual inflection — the moment a woman stopped expecting to progress, stopped feeling the role was tenable — may have occurred months or years earlier. Collecting reasons at the end of a process that began much earlier recovers only the final prompt, not the accumulation that made it decisive.
None of this means exit data is worthless. It measures something real: the stated proximate reason, in the circumstances of a managed departure. The error is treating that as a measure of structural cause. What it omits is usually more diagnostic than what it contains.