The period between roughly 2012 and 2018 saw an enormous number of initiatives aimed at getting girls into computing. Corporate funding was abundant, media coverage was favourable, and programmes launched in large numbers.
Most no longer exist. This is not a scandal — it is the normal life cycle of grant-funded programmes — but the ones that survived are worth examining, because their common features are not what the founding pitches emphasised.
Feature one: they were not primarily about coding
The programmes that persisted generally organised themselves around producing something — a game, an app, a website for a local organisation, a piece of hardware — rather than around teaching a language.
This matters for retention. A curriculum organised around syntax has a completion point and then nothing. A programme organised around building things has an indefinite supply of next projects.
It also matters for the participants who were not going to become programmers, who are the large majority. A girl who spent two years building things and then studied medicine did not waste her time; a girl who spent two years learning a language she then stopped using largely did.
Feature two: they were run by someone local, not by a brand
The programmes that closed were disproportionately those delivered as a franchise — a curriculum and a brand deployed into a venue, with delivery by rotating volunteers.
When funding ended, there was nobody with a personal stake in continuing.
The ones that survived generally had a specific individual — a teacher, a librarian, a parent, a local engineer — whose own commitment predated the funding and outlasted it.
This is an uncomfortable finding for funders, because it means the durable programmes are the ones that would probably have existed anyway, and the fundable ones are the ones that will not persist.
Feature three: they were cheap
Programmes with high per-participant costs — specialised equipment, paid instructors, dedicated premises — closed when the funding cycle ended, because their cost base could not adjust.
Programmes running in a library or a school room, using whatever machines were available, with volunteer delivery, could continue at reduced scale indefinitely.
The uncomfortable version of this is that the impressive programmes were the fragile ones.
Feature four: they had an older cohort
Programmes that retained participants across several years developed something valuable and unplanned: participants who had been there for three or four years and could teach the newer ones.
This solved the volunteer supply problem, which is the binding constraint on most community programmes. It also produced the near-peer role model effect that the research suggests is more powerful than distant eminent examples.
Programmes designed as a fixed-length course could not generate this, because everyone left at the same point.
Feature five: they solved transport
The most banal finding and one of the most consistent.
Programmes running at times and places that participants could actually reach retained them. Programmes requiring a parental lift at an inconvenient time selected for participants with available parents, which narrowed the intake and reduced numbers below viability.
This recurs across every domain covered on this site — sport, education, extracurricular participation. Access is logistics before it is anything else.
What the evaluation problem hides
It is worth being honest that the evidence here is weak.
Very few of these programmes were evaluated in any rigorous way. Almost none tracked participants into further study or employment. The claims made about impact were generally based on participation numbers and satisfaction surveys.
Selection is the deeper problem. Programmes that recruit voluntarily attract participants already inclined towards the subject, and distinguishing the programme's effect from that inclination requires designs that community programmes cannot afford.
So the observations above are about survival rather than about effectiveness, and those are different questions. A programme can persist for a decade without changing anyone's trajectory.
What this suggests for anyone starting one
Build around output rather than curriculum, so there is always a next thing.
Keep the cost base low enough to survive without funding, because the funding will end.
Plan for participants to stay for years and to teach, rather than for cohorts to complete and leave.
Solve the transport and timing question before anything else, because it determines the intake more than the content does.
And find the person whose commitment is not contingent on the grant, because that person is the programme. Everything else is infrastructure around them.