Youth mentoring is one of the most widely implemented social interventions and one of the most extensively evaluated. The evidence is more equivocal than its popularity implies, and the pattern within it is instructive.
The average effect
Meta-analyses of youth mentoring programmes have consistently found small positive average effects across a range of outcomes.
Small effects across many outcomes at low cost is a defensible result, and mentoring is not an ineffective intervention. But the average understates something important, because the variation around it is very large.
Some programmes produce meaningful effects. Some produce none. And some relationships produce harm.
The duration finding
The clearest moderator identified in this literature is how long the relationship lasted.
Studies examining relationship duration have found that longer relationships are associated with better outcomes, that the benefits accumulate over time rather than appearing early, and that relationships ending within a few months are associated with outcomes no better than no mentoring at all.
The most significant finding, and the one most often omitted from programme design, is that relationships terminating early have in several analyses been associated with worse outcomes than not being mentored.
The proposed mechanism is straightforward. The young people who benefit most from mentoring are frequently those who have experienced unreliable adults. A relationship that is established and then ends prematurely confirms rather than corrects that expectation.
What this means for programme design
It implies that recruiting mentors who cannot commit for an extended period is not merely less effective. It may be actively harmful.
Programmes that screen for commitment capacity, that are explicit about the duration required, and that manage endings carefully perform better in evaluations than those that maximise the number of matches made.
This is uncomfortable for programmes measured on volume. A programme reporting a large number of matches, a substantial proportion of which lasted under six months, may have produced a worse aggregate outcome than a programme reporting a quarter as many matches that lasted three years.
Instrumental versus relational
A second useful distinction concerns what the mentoring is for.
Relational mentoring — building a supportive relationship, with activity and content secondary — was the dominant model historically.
Instrumental mentoring — organised around a specific goal, skill or transition — has been found in several analyses to produce larger effects on the targeted outcome.
Programmes combining both appear to perform best, which is unsurprising: the relationship provides the conditions and the goal provides the content.
Purely relational programmes with no shared activity struggle, partly because they place the entire burden of sustaining the relationship on conversation, which is difficult with an adolescent for the reasons discussed in this site's article on disclosure.
The matching question
Programmes invest heavily in matching, frequently on demographic characteristics.
The evidence that demographic matching improves outcomes is weaker than the practice implies. Several analyses have found little effect of same-gender or same-background matching on outcomes once other factors are controlled.
What does appear to matter is shared interest, which supplies the activity that sustains the relationship.
This does not mean representation is unimportant — the vicarious experience literature suggests perceived similarity matters for self-efficacy specifically. But it does suggest that a well-matched interest with a demographically different mentor may outperform the reverse.
Natural mentoring
The most interesting finding in this literature concerns relationships that were never programmed.
Studies of natural mentoring — an adult from the young person's existing network who takes on the role without any programme — have generally found effects comparable to or larger than formal programmes.
The likely explanation is selection into relationships that already work, plus duration, since these relationships tend to be long-standing.
The practical implication is that identifying and supporting existing relationships may be a better use of resources than creating new ones, and several programmes have shifted towards this model.
For a young person, it suggests that the teacher, coach or relative who already takes an interest is a more valuable asset than a formally assigned mentor, and that maintaining that relationship deliberately is worth the effort.
What a young woman should actually do with this
Ask directly. The help-seeking research consistently finds that people underestimate others' willingness to help, and a request for occasional advice from someone who already knows you is very frequently accepted.
Ask for something specific. Open-ended mentoring requests are hard to accept because the commitment is undefined; a specific question or a defined piece of help is easy.
Maintain it over time. The duration finding is the strongest in this literature, and duration is largely determined by whether contact continues after the initial reason for it ends.
And distinguish advice from advocacy. A mentor who tells you things and a sponsor who puts your name forward are different roles, and the second is the one that changes outcomes. Asking for it explicitly is the only reliable way to get it.