Most people who consider helping with something never complete a first task. They read about the group, form an intention, run into an obstacle that would take ten minutes to explain to them, and quietly stop. Nobody records this, because from the group’s side nothing happened at all.
The gap between interest and a first finished contribution is where nearly all volunteer recruitment is lost, and it is almost entirely made of small practical problems rather than large motivational ones. The useful thing about that is it makes the problem addressable by anyone willing to look at it honestly.
Labelling easy tasks is the standard fix and it is weakening
The most widely adopted solution is to mark certain tasks as suitable for beginners. Open source software projects have done this at scale for years, tagging issues as good first ones so that newcomers can find something tractable without needing to understand the whole system.
A study published recently followed this practice across thirty-seven active projects over four years, covering more than four hundred thousand tasks and eleven hundred first-time contributions. It found the approach working substantially less well than it used to.
The acceptance rate for first contributions fell by a third
In the first two years covered, just under sixty-two per cent of first contributions to beginner-labelled tasks were accepted. That figure held steady across both years. In the third year it dropped to forty-nine per cent, and in the fourth to forty-two per cent.
That is a fall of roughly twenty percentage points, concentrated entirely in the final two years after a long period of stability. A statistical test placed the point of change at the start of 2024. Whatever happened, it happened fairly suddenly.
The supply of suitable beginner tasks fell at the same time
The proportion of all tasks marked as beginner-friendly had been consistent at around nine in a thousand for three years running. In the fourth year it fell to under six in a thousand, a drop of roughly a third.
So there were fewer starter tasks available, and the contributions people made to the ones that remained were accepted less often. Both trends moved in the same direction at the same moment, which is unusual enough to be worth explaining.
Newcomers were trying just as hard and writing more
Two further figures from the same study complicate the obvious reading. The share of beginner-labelled tasks actually attempted by newcomers stayed flat throughout, between a quarter and roughly three in ten. Interest did not fall.
And the explanations people wrote alongside their contributions got substantially longer, rising from an average of around three hundred characters to nearly five hundred. People were putting more effort into describing their work, not less, while being turned down more often.
Three explanations fit the timing
The timing invites a single conclusion, which is that widely available writing assistance made it easier to produce work that looks right without understanding it. The date fits, and this is the strongest version of the argument.
Two other explanations deserve more weight than they usually get. The first is that the people reviewing contributions have not grown in number. If submissions rise while reviewer hours stay flat, the acceptance rate falls arithmetically with no change in quality at all. Surveys of maintainers consistently find around six in ten are unpaid and over four in ten cite exhaustion as a reason they have considered stopping.
The second is that genuinely easy tasks are a finite resource. In any long-running effort, the well-bounded work that needs no background knowledge gets done. What remains labelled may increasingly be work that was never really suitable, marked optimistically because something had to be.
A capacity problem and a quality problem look identical from outside
These two explanations produce the same visible outcome. A contribution sits, nothing happens, it is eventually closed. The person who made it cannot tell which occurred, and neither can anybody reading the statistics.
They call for opposite responses. A quality problem calls for clearer guidance about what is expected. A capacity problem calls for more reviewer time, and clearer guidance makes it slightly worse by adding another thing to read. A group can tell which it has by checking whether the time to first response rose during the same period the acceptance rate fell.
One label was never going to address the whole problem
The most systematic study of what stops newcomers catalogued fifty-eight separate barriers, drawn from interviews with thirty-six contributors across fourteen projects. Thirteen of those were social. The rest were practical.
Marking suitable tasks addresses exactly one of the fifty-eight, which is not knowing where to start. It does nothing about instructions that do not work, missing background, uncertainty about whether help is wanted, or not knowing whether three weeks of silence means refusal or a busy week. Designing one intervention for one barrier and then evaluating it as though it covered the category is the recurring error in this area.
The route from interest to contribution has six stages
The first is orientation. Can somebody work out in two minutes what the group does, whether it is active, and whether help is wanted. A group not currently accepting new people should say so, because ambiguity wastes more goodwill than a clear no.
The second is setup, meaning whatever practical preparation is needed before any useful work can begin. This is where the largest silent drop-off happens, and it is the easiest stage to test, by following your own instructions from scratch and noting where they fail.
The third is choosing something to do. The fourth is doing it and submitting it in whatever form is expected, where every undocumented expectation becomes an extra round of correction. The fifth is review. The sixth is coming back a second time, which almost nobody measures.
Waiting is the stage that decides the outcome
Of everything in that sequence, the strongest single predictor of whether somebody returns is how long they waited for a first human response, and how predictable that wait was.
A person who hears back substantively within a few days and sees their work concluded within a fortnight generally has a good experience even if the answer was no. A person who waits five weeks in silence does not come back, whatever the eventual decision. Speed matters less than the absence of an unexplained void.
Getting people through a programme is the solved part
Large organised mentoring programmes have been running for two decades and publish their results. One well-known scheme grew from around four hundred participants across forty-two organisations in its first year to more than twelve hundred participants from seventy-five countries, supported by over two thousand mentors, by 2024. Its completion rate has sat above eighty-seven per cent for a decade and reached nearly ninety-three per cent.
So getting people through a structured programme is solved. What happens afterwards is another matter. Research on participants found most do not continue with the specific project they were assigned to, with a little over half reporting increased involvement. People are generally open that they joined to gain experience and something to put on a record of achievement.
The durable outcome is that some of them come back as organisers
One figure from that research is more useful than the rest. Roughly eighteen per cent of participants went on to become mentors themselves in later years.
That is a different kind of outcome from continued contribution. It is how a scheme compounds year on year, and it suggests the right measure of success is whether somebody becomes capable of bringing in the next person.
Six numbers any group can work out from what it already has
How long people wait for a first reply, expressed as a spread, since an average conceals the long waits that do the damage. How long until their first task is concluded either way. What proportion of first attempts are accepted. What proportion of first-time helpers come back. How many rounds of correction a first contribution typically takes, where a high number points to unwritten expectations. And how many beginner tasks have sat unclaimed for three months, which means either the labelling or the advertising is wrong.
Quarterly is the right frequency. Monthly is noise at most sizes.
Where to look first when recruitment is not working
Follow your own joining instructions from the beginning, on a clean slate, without using anything you already know. Check how long the last ten newcomers waited. Count how many of last year’s first-time helpers appeared a second time.
Those three exercises take an afternoon between them and reliably locate the problem. The answer is rarely that people are less willing than they used to be, and almost always that something specific and fixable is standing in the way. That pattern repeats across almost everything volunteer groups struggle with.