The image of a "leaky pipeline" implies a vessel that loses flow continuously and more or less uniformly — a drip here, a drip there, across every stage from school to senior role. The picture is vivid and widely used, but it is structurally wrong, and that wrongness has consequences. If the loss looks even, any intervention looks as useful as any other. Mapping where attrition actually concentrates shows something different and less comfortable: the pipeline narrows at a small number of specific junctions, and the biggest single step is not the one most campaigns address.
The entry junctions
Begin at the point of application. Women make up a minority of computer science graduates in the UK — the figure has hovered around a quarter for the better part of two decades — but even that starting point is not uniformly distributed. The minority is sharpest in the degree routes that feed directly into software engineering and infrastructure roles; it is considerably less sharp in disciplines that carry computing as a component alongside other subjects. The consequence is that a recruitment process selecting exclusively from single-subject CS graduates starts with a narrower pool than recruitment from the broader graduate labour market, and the narrowing is structural before any individual decision is made. Advert language and formal entry requirements tighten that funnel further.
- Largest single proportional step
- mid-career (roughly 8–15 years' experience), not entry
- Women's share of UK CS graduates
- approximately one quarter, stable for roughly two decades
- The pipeline image implies uniform loss; the actual data show concentrated loss at specific junctions
- Attrition at the mid-career/senior transition widens the gap beyond what entry numbers explain
- Entry-only interventions cannot close a gap whose dominant cause is mid-career departure
The second entry junction is the one with the longest historical roots: reclassification. When programming was reconceived in the 1960s as technical rather than clerical work, it moved from a job category that women routinely held to one that required formal credentials they had not been given access to on equal terms. That shift deposited a pattern — female-majority clerical computing, male-majority programming — that subsequent decades adjusted only slowly. What the shift produced structurally was a grade boundary at which women disproportionately stopped, not because the work changed but because the label on the work changed. The entry requirements that were attached to programming in that period were, in part, a codification of who was already doing it.
Where the gap actually opens
Comparative data on the British workforce consistently show that the largest proportional step — the single point where the female share drops most sharply — is not entry and not the transition from junior to mid-level. It is the transition from mid-level to senior individual contributor or first-line management: roughly the eight-to-fifteen-year experience range. Women enter the profession in a minority, but the minority at entry is not as extreme as the minority at the top. The differential widens substantially in the middle.
That is not a reason to abandon early intervention — it is a reason to treat it as necessary but far from sufficient.
This matters for diagnosis. Entry-focused campaigns — outreach, scholarships, degree partnerships — address a real deficit but a smaller one than the mid-career step. If one hundred women enter the profession and the attrition at the mid-career junction is the dominant loss, then doubling entry numbers without addressing that junction produces more women leaving in their mid-thirties, not more women at senior levels. The arithmetic is simple and the implication uncomfortable for organisations whose diversity work is concentrated on recruitment.
- 1960s: programming reclassified from clerical to technical work, moving it from a female-majority to male-majority job category
- The reclassification attached formal credential requirements to work that had previously been done without them
- Pattern deposited by that shift persisted through subsequent decades, adjusted but not reversed
Mid-career attrition is the structural mechanism here: the reasons women leave at that stage are not the same as the reasons men leave, and the gap in departure rates is not random. Flexible working arrangements, pay progression visibility, sponsorship into senior projects, and the distribution of non-promotable administrative labour have all been documented as unequal at this stage. None of these is an entry problem.
What this implies for measurement
A published representation figure — say, the percentage of women in technical roles — masks which junction produced it. A figure that is stable year-on-year could reflect a situation where entry and mid-career attrition are both unchanged and roughly cancelling, or one where entry is rising and mid-career loss is accelerating, or the reverse. Without stage-by-stage tracking, the headline statistic tells you the current stock level, not the rate of flow or where flow is being interrupted.
The pipeline narrows in specific places, at identifiable points in employment history and career trajectory. The gap that matters most to close is not evenly spread across the route and is not at the beginning. That is not a reason to abandon early intervention — it is a reason to treat it as necessary but far from sufficient.