Mentoring as Network Infrastructure
Dare IT’s Mentoring program is best understood as an intervention in access, not as a softer version of technical training. The question here is how four months of one-to-one contact with an experienced woman in technology can turn existing ability into better information, stronger relationships, and more credible routes to a job. Following that sequence also clarifies what the study can identify: Mentoring changed employment and network measures for the people who received access, but a larger network count is evidence about a channel, not proof that every new connection produced a hire.
The distinction matters because “mentoring” can describe almost anything from occasional encouragement to intensive professional sponsorship. In the Dare IT design, the program’s stated emphasis was networking. Mentors helped participants expand their professional connections, make introductions to hiring managers, and learn how to use those relationships. Interviews with mentors also identified tacit recruitment knowledge, confidence reinforcement, and technical feedback. The program was relational, but its purpose was concrete: reduce the distance between a capable job seeker and the people, information, and decisions that govern entry into technology work.
What a mentor changes in the search process
A participant can arrive with technical skills and still lack a workable route through the hiring market. She may not know which roles fit her profile, which employers are genuinely hiring, how a recruiter interprets a career transition, or how to ask an unfamiliar professional for useful advice. A mentor can change that search environment in several connected ways.
The first is access to information. A technology professional knows more than the public description of a vacancy. They may understand which skills are actually used on a team, what a hiring manager will ask about, how a candidate should frame a nontraditional background, or which opportunities are realistic at a particular stage. This information does not guarantee a job, but it can reduce wasted applications and make search effort more targeted.
The second is access to people. A connection to a working professional can lead to another conversation, an introduction to a hiring manager, or a referral that places a résumé inside a process rather than at its outer edge. The value is not simply that the mentor knows many people. It is that the relationship can make the participant legible to someone who has a reason to pay attention. The connection is a bridge into a labor-market network that the participant may not be able to build quickly on their own.
The third is knowledge of how to use the bridge. A referral is not a magic credential. The participant still has to explain their experience, prepare for an interview, follow up appropriately, and decide which opportunities are worth pursuing. Mentors can teach these tacit practices: how to ask for an introduction, how to describe a career change without apologizing for it, and how to turn a conversation into a next step. This is different from supplying a script that works everywhere. Its value lies in making the rules of a particular professional setting less opaque.
The fourth is confidence grounded in action. Confidence is often treated as an internal trait, but in this program it can be reinforced by repeated contact with a successful woman who has navigated the same sector. A mentor can help a participant interpret rejection, persist through a longer search, and recognize that a technical question in an interview is not evidence that she does not belong. That effect may alter not only whether she applies, but which opportunities she considers and how long she stays in the search.
Technical feedback can reinforce these channels. Code review or discussion of a project may improve the participant’s work, but the more important point for the network mechanism is that feedback happens inside a relationship connected to the sector. The participant is not merely acquiring an additional lesson; she is learning how practitioners judge work and how to present it to people who may hire.
The intended pathway is therefore cumulative rather than linear. A mentor learns what the participant can do. The participant receives advice about roles and search. The mentor makes or facilitates a connection. The participant uses that connection to obtain better information, a referral, or an interview. The resulting experience can strengthen later applications and expand the participant’s own network. Any one step can fail, but the program creates more opportunities for the steps to reinforce one another.
The sequence should not be read as a guarantee that every mentor makes a direct introduction or that every introduction produces an interview. It is a model of the mechanism the program is designed to activate. The study’s evidence becomes persuasive when an intermediate outcome moves in the predicted direction.
The network channel in the evidence
Mentoring increased participants’ LinkedIn connections by about 52.7 relative to a control-group mean near 225, roughly a 23% increase. Challenges, which did not use mentor introductions, produced no detectable network expansion. That contrast is important because both programs increased technology employment. If connection growth were merely a by-product of getting a job, the two groups might have shown similar network changes as participants acquired colleagues. Instead, the network effect appeared specifically where the intervention supplied relationship-building.
The distribution of the change is also informative. Mentoring’s effect was larger at the lower and middle portions of the connection distribution and not detectable at the upper portion. In practical terms, people who began with sparse networks had more room to gain from the program. This is what the network hypothesis predicts: adding another contact to an already dense professional network may be useful, but adding a path into the sector to someone with few relevant contacts can alter the search environment much more substantially.
The study also observed different timing. Mentoring participants began finding technology jobs during the program, consistent with active referrals or direct help with placement. Challenges participants’ gains were more concentrated after the program, consistent with a participant finishing work that could be deployed in later applications. Timing does not prove the full causal chain, but it fits the distinction between an intervention that opens access during the relationship and one whose main asset is portable evidence after completion.
The employment effect provides the outcome-level test. Access to Mentoring increased technology employment by about 15 percentage points at twelve months. Because the program also changed connections in the predicted direction, the evidence is stronger than a simple before-and-after story: the program affected the outcome and an intermediate measure that its design specifically targeted. Still, the study does not isolate networking from every other component. Mentors also provided feedback, tacit knowledge, and confidence support. The defensible conclusion is that the results are consistent with network expansion being an important mechanism, not that the study has measured the exact share of the employment effect caused by introductions.
The network measure has limits. LinkedIn connection count is not the same as a useful relationship. A connection may be distant, inactive, or unrelated to hiring. The study also notes a 500-connection display cap, which can hide further expansion and push the measured difference toward zero. A larger count is therefore a useful indicator of changed access, not a complete account of network quality. To evaluate the mechanism in practice, a provider would want to know whether participants reach relevant employers, receive useful information, obtain referrals, or enter interviews—not only whether a profile displays more contacts.
Selection is part of the story, not the treatment effect
Mentoring was substantially oversubscribed. In the studied cohort, more than a thousand people applied for roughly 150 places, so access was not assigned from an undifferentiated public queue. Mentors first identified preferred candidates. The researchers then randomized access within pairs of applicants whom the same mentor had selected. This design preserved an important feature of the real program—mentor discretion—while creating a comparison that could separate selection from treatment.
That separation prevents a common error. Suppose participants who receive Mentoring find jobs at a higher rate than people who did not receive it. The difference could reflect the program, the fact that mentors chose promising applicants, or both. The study found that selected applicants assigned to the control group had stronger baseline employment prospects than applicants whom mentors did not select. Mentors were therefore identifying people who were already more likely to succeed. Comparing all participants with all nonparticipants would overstate the program’s effect.
Within the mentor-preferred pairs, however, one applicant received access and the other served as a control. The contrast asks a narrower causal question: among applicants whom mentors considered suitable to work with, what changed when one received Mentoring and the other did not? That is why the 15-percentage-point effect should be interpreted as an effect for the studied selected population, not as a guarantee for every person who might apply to a general mentoring program.
This design also reveals a practical tension. Selection is valuable because mentors may see information that a form does not capture. But selecting people who look promising is not the same as selecting people who need or benefit most from the intervention. A participant with strong baseline prospects may be a sensible mentee, yet the greatest marginal value may lie with someone whose network barrier is severe and whose chances without assistance are lower. Later allocation decisions must keep those objectives separate.
Why the program is difficult to scale
The same feature that gives Mentoring its power creates its capacity limit. The intervention depends on working professionals who are willing and able to spend four months in weekly one-to-one contact. Their time, attention, sector knowledge, and access to hiring contacts are scarce. The program cannot be expanded simply by adding more seats to an online platform. Recruiting more mentors may change the quality and density of the networks available to participants; asking existing mentors to serve more people may dilute the relationship that carries the mechanism.
This does not mean that Mentoring is inherently small or that online delivery is automatically superior. It means that the resource being scaled is not content. It is credible, situated access to a professional network. A provider expanding the program would need to track whether participants still receive meaningful contact, whether introductions are relevant, and whether mentors can give advice grounded in current hiring conditions. The number of enrolled participants is not enough to show that the mechanism survived expansion.
The trade-off is visible in the comparison with Challenges. Challenges can serve large online cohorts and has a reported participant cost of about $15 in the studied design, because a reviewed project can be delivered without assigning every participant a scarce senior professional. But its mechanism has its own bottlenecks: assignments must reflect real work, feedback must be credible, participants must complete enough to produce evidence, and employers must understand what the evidence means. Mentoring concentrates scarce human access; Challenges concentrates the work of producing and validating a portable signal. Neither eliminates the need for quality control.
Mentoring is therefore best viewed as network infrastructure for a transition. It can make opportunities, expectations, and people more reachable for participants whose existing connections are thin. The study’s employment and network results support that interpretation, while the randomization and the LinkedIn measure set important boundaries around it. The next comparison asks what happens when the central problem is not reaching an employer, but giving that employer a credible way to see practical ability without relying on a mentor’s network.