When mentoring works for women founders: an output function of accelerator mentoring

Purpose Mentoring is the signature learning mechanism of accelerators, yet increasing mentoring volume does not reliably produce founder momentum. This paper resolves that dissociation by modeling mentoring as an organizational learning system that, in this sample, appeared most useful when it produced specific functional outputs, not on relational intensity alone. Design/methodology/approach Semi-structured interviews with 15 women founders in accelerator and innovation-center contexts were analyzed using inductive coding and cross-case comparison to construct a thematic map of how mentoring signals are interpreted and converted into action under uncertainty. Findings Effective mentoring produces three outputs: (i) navigation clarity (what to do next and who to call), (ii) credible access via sponsorship (introductions and advocacy that alter opportunity structure) and (iii) usable feedback that supports near-term experimentation while preserving founder confidence and action capacity. Two cross-cutting conditions constrained whether these outputs emerged: translation scaffolds that make guidance portable across context and identity-coded credibility dynamics, and boundary governance that protects legitimacy and psychological safety. When outputs were unreliable, founders did not simply disengage; they appeared to filter mentoring signals and tighten boundaries in ways that protected action capacity but slowed iteration and attenuated learning transfer. Practical implications Accelerators should manage mentoring as infrastructure: engineer early navigation sprints, specify sponsorship as a deliverable and track conversion, standardize feedback hygiene into short-cycle experiments, embed portability checks and enforce boundary rules with no-fault rematching to restore learning flow. Originality/value The study offers an output-specified mentoring architecture with preliminary indicators of where mentoring systems may break down, enabling accelerators to shift from counting encounters to designing routines that more consistently produce venture movement.