Better PMs Build Better PMs

Mentoring other product managers — especially ones you don't manage — is one of the highest-leverage moves available to a senior PM. Not because it's generous, but because it pays the mentor back in career growth while also paying the organization back through stronger retention, faster promotions, and knowledge that would otherwise stay trapped inside one team. Almost everything written about mentorship stops at advice for finding one; there's comparatively little on why an experienced PM should want to mentor other Product Managers. The good news is recent research supports my 2020 claims about the value PMs realize by mentoring others.

Before It Had a Name

I was mentoring people before I had a job that used the word for it, and the model didn't come from a management seminar.

I made the Delaware Olympic Development Program team as a teenager and stayed on it for several years — the tier of youth soccer where a state's strongest players get pulled into a shared training pool feeding the regional and national pipeline. Somewhere along the way, I started coaching younger players and I didn't stop when the rest of my life got harder. Despite being a full-time architecture student at Drexel — a program whose studio weeks routinely demanded three or four nights in a row with no real sleep — I kept coaching and mentoring youth teams and rising stars on the side.

Showing a fifteen-year-old how to hold a back line under pressure solved a different problem than anything else I was doing that week — worth protecting even when I had no spare hours left to give it.

I didn't have a framework for why at the time. I only knew it mattered.

That model had two sides: structure and logistics on one, presence and judgment-transfer on the other. Kram puts an academic name to that same split. It's one I carried into every job and leadership role I've had since 2006. I used it as vice president of Drexel Smart House, the university's student-run smart-home research program. I refined the technique when receiving direct mentorship from almost a dozen architects when I joined John Milner Architects in Chadds Ford, Pennsylvania, a small firm where apprenticeship wasn't a program — it was simply how the practice of architecture got transmitted from a principal to a young associate, the way it always has been. I used it teaching at Singularity University and again at two micro-mobility startups. I expanded my focus beyond just PMs once I got to Udacity and then Google, where the people I ended up mentoring most were engineers and designers, not product managers.

Those experiences proved to me the model works well beyond any one person's specific domain or role and that the mentorship model was never about title to begin with.

Session Twelve, Six Years Later

The argument I made in 2020 has held up better than the platform I made it on — which is itself a small case study in why you shouldn't wait for a program to organize this for you. In May of that year, I gave a livestream talk to the twelfth cohort of The Product Mentor, a mentorship program Jeremy Horn built out of a New York meetup he'd been running since the 2000s. I was asked how to mentor effectively, how to find PMs worth mentoring, and — more interestingly — why an experienced PM should bother in the first place.

Volunteer-run mentorship infrastructure eventually collides with the volunteer's actual life — and The Product Mentor is a good illustration of that, not an exception to it. The program still pairs mentors and mentees for six-month sessions, but its site hasn't listed a new cohort since it flagged one for Q1 2024. That's not a knock on the program, or on Jeremy — running something like this well, for over a decade, on top of an actual job and an actual life, is a bigger commitment than almost anyone attempts. Almost any mentorship infrastructure built around one generous person's bandwidth, rather than a durable ecosystem that outlives any single steward, eventually runs into that person's actual life. Most informal mentorship infrastructure works this way, which is a good reason not to wait for a program — his or anyone else's — before you start.

Generativity Is Not Altruism

Most PMs who mentor describe it as generosity — giving something away, or paying something forward. Erik Erikson would have called that backwards. The psychologist spent a career arguing that adult development doesn't stop once you've secured a stable identity and a partner. Mid-adulthood brings a further choice, between what he called generativity and stagnation: either you turn some of your attention toward establishing and guiding the next generation — not necessarily your own children; Erikson was explicit that this included teaching, mentorship, and any productive work that outlives the person doing it — or you don't. The alternative isn't neutral. Stagnation, in his model, is a kind of psychological calcification: rigidity, self-absorption, and a shrinking capacity to learn anything you didn't already know.

Generativity is a developmental task with its own psychological payoff, not a favor extended downward, and skipping it has its own cost. Senior product people who stop teaching rarely plateau gently. They calcify. Their mental models stop updating. They get defensive about markets and tools they haven't touched in five years, mistake seniority for correctness, and become the last person in the room to hear about a shift everyone else already adapted to. Mentoring is one of the best defenses against becoming that person, because it forces you to keep re-deriving your own judgment out loud, for an audience with no obligation to nod along.

The Mentor's Dividend

For decades, mentoring research focused almost entirely on protégés. That changed when researchers started asking the opposite question: what does the mentor get? A 2004 meta-analysis by Tammy Allen and colleagues pooled the existing research and found that, compared to unmentored peers, protégés see a moderate bump in promotions (r = .31) and smaller but real gains in compensation (r = .12) and career satisfaction (r = .21). Modest effect sizes by the conventions of the field, but real and replicated. Companies spent a generation selling mentoring as something seniors do for juniors.

The question got more interesting once researchers turned the same instrument around. In 2013, Rajashi Ghosh and Thomas Reio published a meta-analysis asking what mentors themselves get out of the relationship. Career-focused mentoring correlated with the mentor's own career success at r = .44 — a strong effect, larger than almost anything in the protégé-side literature. Psychosocial mentoring correlated with the mentor's organizational commitment at r = .216. Both correlations describe a return — satisfaction, commitment, career trajectory — that the protégé-side research had been measuring in mentees for twenty years, just measured in the other direction, not mentors being flattering about themselves in a survey.

I think of this as the mentor's dividend — my own term for it, since I haven't found one in the literature — measurable, directional, and almost entirely absent from how PMs are taught to think about their own career development. Nobody puts "mentored three PMs this half" on a leveling rubric, so nobody budgets time for it, so the return goes uncollected by exactly the people best positioned to earn it.

None of this research is specific to product management, or even to tech. Kram's original interview subjects, discussed below, were eighteen pairs of corporate managers, not product people; Allen and colleagues' meta-analysis and Ghosh and Reio's both pool studies across industries and professions. I'm writing about PMs because that's my domain and this is a product-focused essay, but the mechanism doesn't care what you do for work. Swap in engineer, designer, researcher, or line cook, and the argument holds: teaching someone else in your field pays the teacher back too, measurably, and that's not a quirk of product management.

It's also, I'll be honest, not a finding that comes from a very wide slice of the world. Kram, Allen, Ghosh and Reio, Payne and Huffman — all of that is Western, mostly American, management research, and I don't think that's a small caveat. So it's worth noting that the same basic pattern shows up outside that frame too. Jeevan Jyoti and Poonam Sharma's study of call center employees in India found that mentoring functions significantly predicted career development on their own, and that the organization's broader mentoring culture and mentoring structure further shaped how much career benefit those functions produced — a different labor market, a different industry, a different continent, and the core relationship still holds.

Two Functions, Not One Instinct

Kram's forty-year-old framework explains why the dividend has a shape to it: mentoring is actually two separate jobs, and doing only one of them half-works. The talk I gave in 2020 included a two-column list I'd built from a decade of doing this informally — one column of logistics (check your biases, establish expectations, run a needs assessment, set goals, keep a schedule), one column of behavior (listen deeply, ask probing questions, guide rather than tell, be accountable, be accessible). At the time, I didn't know that Kathy Kram had already given both columns names in 1983.

Kram's original study, drawn from interviews with eighteen pairs of senior and junior managers, split mentoring behavior into exactly the two functions my two columns had rediscovered by trial and error. Career functions are the logistics-shaped half: sponsorship, coaching, protection, exposure and visibility, challenging assignments — the things that move a protégé's actual position. Psychosocial functions are the relationship-shaped half: role modeling, acceptance and confirmation, counseling, friendship — the things that shape a protégé's sense of competence and identity, independent of title. Kram's finding, which has held up across four decades of replication since, is that the two functions aren't substitutes for each other. A mentor who only does logistics — introductions, sponsorship, a quarterly check-in — produces a protégé who advances without developing judgment. A mentor who only does the relationship half — a sympathetic, always-available ear, no scaffolding — produces a protégé who feels supported and doesn't move.

That's a useful thing to notice about most "best practices" PMs pass informally between each other: they're frequently somebody's independent rediscovery of a finding that already has forty years of citations behind it. Knowing that doesn't change the advice. It changes your confidence that skipping either half — running purely sponsorship, or purely supportive, feelings-only mentorship — isn't a stylistic choice. It's an incomplete implementation.

What the Org Gets

The organizational case is separate from the individual one, and it's worth stating plainly rather than leaving it implied: a company with more senior people mentoring outside their own team gets lower regrettable attrition, a promotion pipeline that fills faster, and knowledge that moves between teams instead of staying trapped in one of them. None of that requires taking my word for it.

On retention: mentoring measurably reduces the odds someone quits, and it does so across very different economies and incentive structures. Stephanie Payne and Ann Huffman tracked more than a thousand U.S. Army officers over multiple years and found that mentoring reduced actual turnover behavior, not just stated intent to stay, and that the effect ran through affective commitment — people who felt invested in stayed longer, an effect still measurable up to ten years out. The mentoring in that study wasn't limited to a person's own chain of command. That matters for the argument here specifically: the retention benefit doesn't require the mentor to be someone's manager. It requires someone senior to have taken a real interest. Peter Hom and Zhixing Xiao found something adjacent studying 417 employees across four Beijing technology firms: the strength of an employee's embedding in guanxi — the dense, reciprocal-obligation networks that structure a great deal of Chinese professional life, of which mentoring-type relationships are one strand among several — predicted retention so strongly that tighter network closure lowered quit risk by 89.25 percent, and reduced network constraint lowered the odds of quitting by 95.67 percent. Mentoring wasn't the whole mechanism in that study, but it was explicitly one of the relationship types the researchers asked employees to name when mapping who kept them embedded. Different country, a broader mechanism than mentoring alone, but the same underlying shape: being woven into somebody senior's network, mentoring included, is what keeps people from leaving.

On the promotion pipeline: sponsoring other people is correlated with your own promotions, and it compounds across levels. Sylvia Ann Hewlett's research, drawn from a nationally representative survey conducted for The Sponsor Effect (supplemented by in-depth interviews with leaders), found that senior people who sponsor others are 53 percent more likely to have received a recent promotion themselves. Entry-level managers who have a protégé are 60 percent more likely to have received a stretch assignment (Hewlett, HBR, June 2019). That's the mentor's dividend again, from a different angle, plus a second-order effect: sponsorship compounds. People who were sponsored become sponsors, and the pipeline gets thicker at every level, rather than staying dependent on however many people happen to report directly to whoever owns talent development this quarter.

On cross-team knowledge transfer: a mentee outside your own org is, structurally, exactly the kind of connection that moves information a closed team never sees. This is the oldest and most durable finding in the set. Mark Granovetter's 1973 paper on the "strength of weak ties" found that novel information — a job lead, a solution someone in a different group already solved, a piece of context that changes a decision — travels through loose, infrequent connections between people in different parts of a network far more often than through close, frequent ones. Your own team already knows what you know; that's what makes them your team. A mentee in a different org is, by definition, a weak tie.

It's worth taking the complication seriously rather than gliding past it, though. The sociologist Alena Ledeneva's research on Soviet and post-Soviet blat — the informal favor economy that filled the gaps a planned economy left open — found closer to the opposite of Granovetter in a genuine scarcity environment: it was strong, trusted ties, not weak ones, that reliably moved scarce goods and hard-to-verify information, because in a system where lying about what you could deliver carried real cost, trust was the whole mechanism. That's not a contradiction so much as a boundary condition — weak ties win when information is abundant and the risk is missing it; strong ties win when information is scarce and the risk is being misled. A mentoring relationship that crosses a team boundary is, in effect, a weak tie you've deliberately upgraded toward the trusted end of that spectrum — which may be exactly why it works as an information channel in both kinds of environment. Multiply that relationship across a company and you've built an informal information-routing layer that no reorg or documentation mandate reliably replicates.

On ramp time: the two functions of mentoring speed up two different things, and you need both to get a new hire to full contribution fast. A 2022 study of newcomers at two manufacturing companies in China found that career-focused mentoring built new hires' task mastery and performance, while relationship-focused mentoring built their social integration — the same career-function-versus-psychosocial-function split Kram named in 1983, above, applied to an entirely different question, in a different country, and still holding up four decades later. Two separate pathways, doing separate work, both of them speeding up how fast a new person becomes a full contributor rather than just how they feel about the job.

I've watched a version of this firsthand rather than just read about it, twice, at very different scales. I helped scale a startup from four people to forty with no formal mentoring program in place, and the informal mentoring relationships that formed across functions during that stretch were, in hindsight, doing a meaningful share of the retention and knowledge-transfer work a program would have been built to do, years before anyone thought to build one. I saw a version of the same thing play out much larger at Udacity, where I joined an eight-person team when the company was around 500 people. By the time I left in 2019, that team was over 72 people — and it had become, by any informal measure, one of the core teams left standing at the company, even as Udacity as a whole scaled up past 1,200 employees and then contracted back down to roughly 300. That's not a controlled experiment; it's one data point from the inside. But the pattern across both experiences was consistent: wherever the mentoring relationships were densest — across functions, across levels, not just top-down — that's where the retention and the institutional memory concentrated too, program or no program. None of this shows up on next quarter's dashboard. All of it shows up eventually.

Mentoring on the Jagged Frontier

My read on this is a prediction, not a finding — nobody's run the study — but here it is: agentic AI is compressing the skill gap mentorship used to close, and only the technical half of it. I've argued elsewhere that agentic AI is collapsing the technical distance between a junior PM and a senior one: a first-year PM with a competent agent stack can now draft a defensible PRD, write a passable SQL query, and produce a first-pass competitive teardown at a fidelity that used to take a couple of years to earn. If mentorship's value were mostly the transfer of that kind of technical scaffolding, agentic AI should be quietly making mentors redundant.

Instead, I'd predict close to the opposite, for a reason Kram's framework anticipates if you take it seriously: AI is compressing the career function faster than the psychosocial one. An agent can help someone produce a good-looking document. It's much less able to tell them which of the three good arguments in that document is the one their VP will actually fight them on, because that isn't a document-quality problem — it's a pattern a mentor has watched fail in a room before.

This isn't really a junior-PM phenomenon, and I don't think seniority is the operative variable at all — I've watched it happen to IC8s just as often as first-year PMs. My actual claim, stated so it can be proven wrong: fluency with AI tooling and quality of judgment are two separate variables, correlated with experience today mostly by coincidence, because experienced people also tend to be early adopters. As AI fluency spreads faster than judgment does — fluency being the easier of the two to acquire — confident, well-structured, wrong output should start showing up at every level, not just among juniors, and specifically around a stakeholder or a piece of institutional history the model has no access to. If that holds, agentic tooling is making the artifact cheap without making judgment any cheaper to acquire — which was never really about the artifact in the first place.

The Objection: You Already Manage People

The strongest objection to everything above is a fair one, and worth stating plainly rather than ignoring: if you already manage a team of PMs, you're already doing the generative work Erikson describes — coaching, sponsoring, developing people — as part of your actual job. Why go find more people to mentor outside it?

I'd agree with a version of that objection. Mentoring outside your own remit matters most, and comes closest to necessary, before you have a team of your own to develop — when you're a senior IC or a first-time manager with no direct reports yet, and the generativity outlet Erikson describes has nowhere else to go inside your day job. That's exactly the population The Product Mentor and programs like it are built for, and it's not a coincidence that most of the mentees I've worked with over the years were in that position.

Once you do manage a team, the case changes shape but doesn't disappear. Mentoring someone outside your reporting line, in a market or a function you don't already touch, isn't redundant with managing your own team — it's a different kind of input into the same job, and it pulls you into domains you'd otherwise have no reason to learn. My own deep experience is concentrated in a few markets — B2B2C, B2B, and D2C products in micromobility, education, and online commerce, and, in the years since, media and ad tech at Google and commerce and its adjacent verticals at Meta. Everything else is exposure, not expertise, and mentoring is where I've gotten it: furniture, oil and gas, logistics SaaS, real estate tech, and air-gapped enterprise SaaS are all domains I've never practiced product management in myself, and all domains a mentee has pulled me into. Marketing tech sits in a different, in-between category — I've never been a PM inside a platform like Adobe Marketing Cloud, Salesforce, or HubSpot, but I've been the one deciding which of those stacks to run and how to wire them into a company's actual operations, at Social Bicycles, Superpedestrian, and Udacity alike. At Superpedestrian I led marketing directly for a short stretch, reporting to the COO, while we built out enough of a team that it made sense to bring in a dedicated marketing leader to take it further. Mentoring outside your remit also forces you to re-derive your judgment for someone with a completely different context than the people you already manage, which is a harder and more useful exercise than repeating advice to people who already share your assumptions. Both of those transfer directly back to your own team. You end up a better manager of the people who report to you specifically because you've had to explain yourself, from scratch, to someone who didn't already agree with you.

Good Mentors, Bad Pairings

None of this is a guarantee mentoring will go well, and the research on what goes wrong doesn't assign blame in the direction most people assume. A great mentor can end up with a mentee who's a poor fit. A person who would make an excellent mentee can end up with a mentor who's wrong for them. That's not evidence mentoring doesn't work — it's evidence this is a relationship like any other, subject to fit, not a service you can standardize and guarantee.

Lillian Eby and Stacy McManus's 2004 study, built from mentors' own accounts of relationships that went badly, found that protégés meaningfully contribute to relationship dysfunction too, not only mentors — through disengagement, using the relationship for unrelated favors, or simply an unwillingness to be coached — challenging the assumption, common in the earlier literature, that a bad mentoring relationship is by default the mentor's failure. Most of what they found wasn't malicious on either side, just mismatched: reasonable behavior from one person that didn't fit what the other person actually needed. A companion study by Eby and colleagues in 2008, surveying mentors directly about their own negative experiences, found distinct categories of difficulty — protégé performance problems, interpersonal friction, and in the worst cases destructive patterns — that predict mentor burnout and a declining willingness to keep mentoring at all.

A good mentor and a good mentee can still be a bad pairing, and that's worth taking seriously rather than pushing through out of obligation. Do a real needs assessment before committing to anyone. Be willing to end a pairing that isn't working. That advice was on my 2020 list under "do a needs assessment," without much justification behind it at the time. It has more behind it now.

Finding Someone One Step Behind You

You don't need a formal program for any of this, and I'd caution against waiting for one. The 2020 version of this talk listed a handful of Slack communities and invite links for finding mentees; I'm not reproducing that list here — invite links rot within a couple of years, and a stale link is worse than no link. What doesn't rot is the more boring advice: look inside your own company first. Most organizations, whatever their formal mentorship program does or doesn't offer, have more informal capacity for mentoring than anyone advertises. If you're the first PM at your company, or one of a hundred, someone one level behind you — or one level behind you in a different function entirely — is trying to work out the same things you were trying to work out a few years ago. Programs like The Product Mentor, and communities built around conferences and newsletters, are funnels into that same relationship, if you want a more structured starting point than your own hallway. Pick whichever funnel is closest to you and use it.

Mentoring other PMs — or engineers, or designers, or whoever shares your actual domain — will not appear on your performance review, and it will not be a line on your promotion packet unless you build the artifact yourself. Do it anyway. The dividend doesn't show up where you're looking for it. It shows up eighteen months later, in the market you now understand because a mentee dragged you into it, in the argument you win with your own VP because you'd already made it out loud to a stranger a dozen times first, and in the version of yourself that never calcified into stagnation, because someone one step behind you kept asking why. Go find that person. You don't need a title, a program, or anyone's permission to start.