Most enterprises don't struggle to run a mentoring program. They struggle to prove it worked.
Ask a Head of L&D at a 5,000-person company whether mentoring is valuable, and they'll say yes without hesitation. That’s because they've watched it happen in one-on-one conversations, in retention saves, in promotions that trace back to a good match.
Ask the same person to put a single defensible number on that value in front of the executive committee, and the confidence drops fast. Not because mentoring doesn't work. Because the systems, timelines, and reporting habits most enterprises have in place were never built to capture what mentoring actually produces.
This isn't a "how to calculate mentoring ROI" article — we've written that one already (see How Do You Measure the Success of a Mentoring Program? and Mentoring Program Evaluation for the KPIs, frameworks, and step-by-step methodology). This one answers a different question: why does measurement break down in the first place, specifically at enterprise scale, even when the will to measure is there and the software budget has already been approved?
Enterprises don't struggle to prove mentoring works because the impact isn't real. They struggle because the systems around it were never built to capture it. Six causes drive this: mentoring data sits in disconnected HR systems, the outcomes mentoring produces are qualitative while reporting templates want numbers, measurement typically starts after launch (so there's no baseline), success criteria get agreed too late, software gets bought to solve logistics rather than to prove impact, and attribution gets difficult the moment other talent programs (leadership development, coaching, ERGs) overlap with mentoring's target cohort.
The fix isn't a better spreadsheet. It's sequencing: agree on two or three business metrics and capture a baseline before the program launches, choose software that reports on outcomes rather than just activity, treat qualitative check-in data as evidence rather than decoration, and be upfront about partial attribution rather than overclaiming credit for a shared result.
Small mentoring programs (fifteen pairs, one team, one enthusiastic HRBP running it on the side) can often get away with a spreadsheet and a gut feeling. Enterprises can't. At scale, the same forces that make mentoring valuable (breadth across business units, geographies, and job families) are the forces that make it nearly impossible to measure with the tools most organisations already have.
Six causes show up again and again in enterprise mentoring programs, and they compound each other rather than acting alone.
Analyst Josh Bersin's research on the HR technology market puts the average large company at more than 80 separate HR tools, with many global organisations running close to double that.
Mentoring data (session logs, goal tracking, satisfaction scores) typically sits in one of those 80+ systems, while the data you'd need to prove mentoring's business impact (turnover, promotion rates, performance ratings, engagement survey results) sits in three or four entirely different ones: the HRIS, the engagement platform, the performance system, sometimes a separate LMS.
Connecting a mentee's participation record to their retention outcome eighteen months later means joining data across systems that were never designed to talk to each other. Most enterprises don't have that pipeline built, so the analysis either doesn't happen or happens once a year as a manual, error-prone export-and-match exercise in a spreadsheet.
Mentoring's clearest short-term signals (trust, confidence, clarity about a career path, a mentee feeling less alone in a new role) are real, but they aren't naturally numeric. Executive reporting templates, on the other hand, are built around numbers: revenue, headcount, attrition percentage, cost per hire.
This forces a translation problem that most programs never fully solve. Teams either force qualitative outcomes into shallow quantitative proxies (a single "satisfaction score" that hides everything interesting happening underneath it), or they present rich qualitative findings that a board or exec team dismisses as "nice stories, but is it working?" Neither approach earns the program long-term credibility with the people who control its budget.
A recurring pattern in enterprise programs: mentoring launches, runs for six or twelve months, and only then does someone ask "so how do we know if this worked?"
By that point, there's no pre-program baseline for engagement, confidence, or intent-to-stay to compare against — the comparison group has already been contaminated by the program itself, or was never captured at all.
Without a baseline, every result is a single data point in isolation rather than a measured change. "78% of mentees report increased confidence" sounds strong until someone in the room asks what the number was before the program started, and there's no answer.
Ask five stakeholders in a large mentoring program what it's supposed to achieve, and enterprises frequently get five different answers: retention, leadership pipeline, DEI representation, onboarding speed, engagement scores. Each one implies a different metric, a different data source, and a different timeline for seeing results.
When success criteria are set after the program has been running (usually when someone finally asks for a report), the measurement gets built around whatever data happens to be easiest to pull, rather than the outcome the program actually exists to drive. That's a design flaw, not a data flaw, but it shows up as a measurement problem every time.
Most mentoring software purchases at enterprise scale are triggered by a logistics pain point: manual matching doesn't scale past a few hundred pairs, or nobody can track who's actually meeting whom. Analytics and reporting are rarely the first thing evaluated in the buying process — they're assumed to be "in there somewhere."
The result is that many enterprises are running mentoring software that automates scheduling and matching well, but was never asked to prove its measurement capability during procurement.
When the CHRO eventually asks for a business-impact report, the admin discovers the platform tracks session counts and satisfaction surveys, but has no way to connect that activity to retention, promotion, or performance data without exporting everything into a spreadsheet by hand.
Enterprises rarely run mentoring in isolation. The same cohort of high-potential employees is often also in a leadership development program, receiving manager coaching, and participating in an ERG — all in the same year mentoring is supposed to move the retention or promotion needle.
Even with perfect data, isolating mentoring's specific contribution from everything else happening to that population is genuinely difficult. Most enterprises don't have the statistical infrastructure (control groups, matched cohorts) to do this rigorously, so they either overclaim credit for a shared result or, more often, quietly stop trying to make the case at all.
These six causes don't just make reporting awkward — they put the program itself at risk. HR.com's Future Demands in Coaching and Mentoring research found that while 60% of organisations run mentoring initiatives and 70% run coaching programs, only 45% of leaders say they see a substantial impact on business success from them. That's not necessarily a sign mentoring doesn't work. It's consistent with a measurement gap: leaders can't confidently say a program is delivering because the evidence trail was never built.
That gap has a direct budget consequence. Programs that can't produce a credible impact report are the first line item questioned when budgets tighten, regardless of how much informal, anecdotal enthusiasm exists for them. A program administrator who can't answer "what did we get for this" in a language finance and the exec team trust is fighting for renewal every single cycle.
None of the six causes above get solved by trying harder to build better spreadsheets. They get solved by changing what's captured, when, and where — before the reporting deadline arrives, not after.
For the detailed KPI list, the Kirkpatrick framework, and a step-by-step evaluation timeline, How Do You Measure the Success of a Mentoring Program? and Mentoring Program Evaluation walk through exactly what to track and when. Our Mentoring ROI Calculator is a starting point for putting a first, defensible number on retention and engagement impact using your own program's figures.
Enterprises don't struggle to measure mentoring's impact because mentoring is unmeasurable. They struggle because the data lives in the wrong places, the success criteria get defined too late, and the software doing the day-to-day running of the program was rarely asked to also produce evidence.
Fix the sequencing (decide what matters and capture a baseline before launch, then choose tools that report on outcomes rather than just activity) and the "we can't prove this is working" conversation stops repeating every budget cycle.
If you're evaluating whether your current mentoring software can actually answer the impact question your leadership team is asking, book a demo with Brancher to see how dashboard reporting, goal tracking, and Ava AI's between-session data collection are built to close exactly this gap.
It's possible, but only with the right sequencing: agreed success metrics and a baseline set before launch, and software that connects participation to outcome data rather than just activity data. Programs that try to retrofit measurement onto a program that's already running, using whatever data happens to be lying around, are the ones that conclude it "can't be measured."
Scale multiplies every cause above. More business units mean more disconnected data sources; more overlapping talent programs mean harder attribution; more stakeholders mean more disagreement about what success even means. A 30-person program can survive on a shared spreadsheet and a program owner who knows every pair personally. A 3,000-person program cannot.
For a single report, sometimes. As a sustainable system across regions, business units, and program cycles, rarely — spreadsheets don't scale with headcount, don't integrate with other HR systems, and depend entirely on one person's manual upkeep, which is a single point of failure the moment that person changes roles.
Set the success metrics and capture a baseline before the program launches, not after someone asks for a report. Every other measurement problem is easier to solve with a baseline in place; none of them are solvable retroactively without one.
The strongest platforms let you segment reporting by cohort and export participation data in a format that lines up with your HRIS or people analytics tool, so retention or promotion comparisons between participants and non-participants can be built without a manual data-matching exercise every time someone asks for a report.
Holly Brailsford is Co-Founder and CEO of Brancher, and a registered organisational psychologist specialising in workplace mentoring design. Holly works directly with enterprise People and L&D teams on program design, measurement frameworks, and stakeholder reporting.