Mid-career attrition

The point at which women leave tech in the greatest numbers is not at the door — it is roughly a decade in.

Hiring campaigns and pipeline initiatives rest on a shared assumption: that the shortage of women in senior technical roles is primarily a supply problem, traceable to entry. The attrition data complicate that story considerably. Survey after survey of the British computing workforce finds that women who enter technical roles in their twenties leave them — or leave the sector entirely — at elevated rates between roughly the ages of 30 and 45. By the time the gap at senior level becomes visible and politically embarrassing, it has been accumulating quietly for years, driven not by what happened at recruitment but by what happened in the decade that followed.

The mechanism matters because it determines the remedy. A supply problem calls for more entrants; a retention problem calls for changes to the conditions that make staying difficult. Conflating the two produces organisations that recruit with one hand and lose experienced staff with the other — net pipeline narrow, seniority gap unaltered.

The structure of the drop

Mid-career attrition in tech is not a single event but a cluster of decisions shaped by converging pressures. Women who entered technical roles in the late 1990s or early 2000s — the cohort now most visible in the attrition data — came in under a particular set of assumptions about progression: that technical depth would translate into seniority, that performance would compound into influence. What many found instead was that the routes upward required moving away from technical work and toward management, and that management grades were staffed and evaluated in ways that disadvantaged part-time or reduced-hours arrangements.

The timing is not coincidental. The decade from 30 to 40 coincides, for many women, with the period of greatest caregiving demand — not only for children but, increasingly, for older relatives. British employment law provides statutory protections around parental leave, but protections against formal discrimination do not resolve informal penalties: the projects not offered, the travel requirements that become suddenly non-negotiable, the performance review that rates presence over output. These are the conditions that exit data omits by design, because they are rarely stated as leaving reasons to the employer being left.

There is also a grade structure effect worth separating from the caregiving story. Technical roles in British computing have, since at least the reclassification periods of the 1960s and 1970s, been organised into ladders where the upper rungs require either managerial responsibility or specialist credentialling that was, for some decades, harder for women to accumulate. Women who entered during periods when those ladders were formally or informally cut and frozen found themselves in roles with no internal upward path. Leaving was not a failure of ambition — it was a rational response to a blocked structure.

What the numbers actually measure

Published attrition figures require careful reading. Headcount loss is the most common measure, but it understates the real movement because it misses lateral exits — the women who remain employed but move from technical into adjacent functions such as project management, business analysis or operations, where they are no longer counted in the technical workforce. That quiet sideways step is structurally significant: it preserves employment while removing technical experience from the pipeline that feeds senior engineering roles. What the surveys count as 'in tech' varies enough between studies to move the attrition headline by several percentage points in either direction.

Leaving was not a failure of ambition — it was a rational response to a blocked structure.

Nor do headcount figures capture the concentration of exits by seniority band. Aggregate turnover rates can look unremarkable while masking a specific depletion at mid-grade — precisely where future senior leaders should be forming. Mean tenure figures are similarly blunt. The distribution matters more than the average, and the distribution in British tech, when disaggregated by gender and grade, shows a pronounced thinning at the mid-point that is not mirrored in entry-level or junior-senior transition data.

The implication is operational as well as structural. Organisations that rely on graduate recruitment to maintain diversity metrics are, in effect, topping up a leaking system. The cost — in recruitment, in lost institutional knowledge, in the compressed seniority pipeline — is substantially higher than the cost of addressing the conditions that drive attrition in the first place. Recruitment campaigns are a visible, reportable intervention; retention work is harder to quantify and easier to defer. That asymmetry has consequences that compound over careers.