Quick answer. Invisible work, mentoring, and glue work get left off your performance review because the form is built for metrics, not for the moments that actually held your team together. The fix is a specific tag vocabulary captured in the moment, not a vague "soft skills" paragraph written from memory the night before. Consistent documentation, kept across the year, materially changes your next review, promotion, or negotiation.
It is review season. The self-evaluation form is open on your laptop. The "metrics" achievements are easy enough to list. Tickets closed, dollars saved, launches shipped. You can put numbers next to those. You write them in.
Then you hit the section that asks about collaboration, mentoring, and team contribution. You stare at it.
Because half your year lives there. The Tuesday afternoon you spent talking a junior engineer through their first real code review. The Slack DM thread that quietly unblocked the marketing team. The conversation in the hallway that pulled a project back from the edge. The retro you facilitated even though it was not your job. The new hire whose first three months went better because you kept making time.
None of that has a number next to it. None of it is on your OKR doc. Your manager probably did not see most of it. By the time you sit down to write the self-eval, you cannot remember the specifics. So you type "supported team collaboration" and move on.
That paragraph is where careers stall. Not because the work was not valuable. Because the work was invisible, and the form was built to count things, and you let memory and vague language do the work that specificity should have done. This guide is how to fix that, starting with the vocabulary you use and ending with a capture system you will actually run.
How do you quantify soft skills like mentoring and collaboration?
You stop calling them soft skills.
"Soft skills" is a 1980s framing that has not aged well. It implies fuzzy, hard-to-measure, secondary. The form is asking for the wrong thing because the language is asking for the wrong thing. The fix is to swap a fuzzy umbrella term for a specific tag vocabulary, then put a dated, attributable moment under each tag.
impactrackr uses a 29-tag taxonomy built specifically for the kinds of contributions that get lost in "soft skills." A few that pay off on the review form:
- mentoring. Who you helped, what topic, what outcome. "Coached J. through their first incident postmortem in March. Walked them through the writeup template, sat in the review, gave structural feedback. The doc went out without rework. J. now owns postmortem writeups for the team."
- coaching. Distinct from mentoring. Coaching is shorter, more situational. "Spent 45 minutes prepping M. before their first VP-level demo in June. They led the room. The demo got positive verbal feedback from the VP in the next staff meeting."
- cross-functional. The work that bridges teams. "Brokered the handoff between data and product when the analytics pipeline scope kept slipping. Set up the weekly sync, drafted the responsibility matrix. Two months later both teams cited the sync as the reason the launch held its date."
- relationship-building. The slow work. "Rebuilt the relationship with the procurement team after the budget freeze conversation went sideways in February. Three lunches, two follow-ups, one rewritten doc. By April the team was forwarding us early visibility on the next budget cycle."
- process-improvement. The quiet wins. "Noticed our standups were running 35 minutes by Q2. Proposed and ran a four-week trial of async-first updates. Standups now average 12 minutes. Three teammates have copied the format."
Each tag is a specific kind of moment. Each moment has a who, a what, a when, and an outcome. Once you have the vocabulary, the form is not asking you to be eloquent. It is asking you to list things. Listing is much easier than reflecting.
What is glue work, and why does it disappear from your performance review?
Glue work is the work that holds a team together but is not on anyone's title or OKR sheet. It is a term Tanya Reilly popularized: the meeting you organized so the launch did not blow up, the doc you wrote so the new hire could ramp without bothering everyone, the cross-team thread you kept warm so the project did not stall. Necessary, valuable, and structurally invisible.
It disappears from reviews for three compounding reasons.
It is not in the title. A staff engineer is reviewed against staff engineer expectations. Running the team's onboarding doc is not in that list. So even if it consumed three weeks of your year, the review template has no slot for it. You either jam it into a section it does not fit, or you leave it out.
It is not on the OKR doc. The OKR doc was written in January. The glue work emerged in April when the team needed it. By calibration time the OKR doc is the source of truth and the April reality is anecdotal.
Your manager was not in the room. Glue work happens in the seams between teams, in DMs, in the moments after the meeting ends. Your manager sees the artifact (the launch went smoothly) but not the cause (you spent six weeks making sure it did).
There is a structural piece on top of those three. Haegele 2025, published in the American Economic Review, found that about 75% of managers self-report engaging in talent hoarding. Translated to the review conversation: the system is not just neutrally bad at counting your invisible work, it is actively biased to suppress visibility for people whose visibility would cost a manager something. You cannot fix the manager. You can fix the record. The record is the only counterweight you control.
Tag the glue work the day it happens. process-improvement, cross-functional, team-development, communication, relationship-building. Five tags that turn six months of seam-stitching into a dated, attributable list your manager has to engage with, in writing, before calibration.
How do you document mentoring on a performance review?
Mentoring is the highest-leverage piece of invisible work most people do, and the most frequently underwritten section of a self-eval. The fix is a four-piece structure.
Who. First name and initial is enough. The person was real. They have a manager. They have a record.
What topic. Not "career advice." Specifically: "first solo client call," "promotion packet writeup," "negotiating scope with PM," "debugging a flaky test suite," "presenting to leadership."
What outcome. What changed because of the time you spent. Not what you taught. What they did differently as a result. "Delivered the call on time and on scope." "Got the promotion." "Held the scope line and saved the team a sprint of rework." "Found the root cause and shipped the fix." "Got verbal feedback from the SVP."
What timeframe. A date or a date range. "March 2026." "May through August 2026, monthly."
Compare the weak version to the strong version on the same underlying work.
Weak: I mentored a junior team member this year.
Strong: Coached R. through their first solo client engagement, May through August 2026. Weekly 30-minute sessions covering scoping, status communication, and stakeholder management. R. delivered on budget, got direct positive feedback from the client in the project closeout, and is now running their second engagement solo.
Weak: Helped onboard new hires.
Strong: Owned onboarding for two new hires (P. in February, A. in September). Built a 30-day shadow plan, ran twice-weekly office hours through their first month, paired on their first three tickets. Both shipped their first independent ticket within two weeks of ramp, ahead of the team's 30-day target. Confirmed in their respective first-month reviews.
The 29-tag taxonomy puts a slot under each of those: mentoring, coaching, team-development, feedback, onboarding. You are not asking the reader to remember the work. You are asking them to read a row in a list.
What does the research say about who pays the price for invisible work?
The price is real, it is measurable, and it falls unevenly.
Babcock and colleagues, writing in the American Economic Review in 2017, ran a series of field and lab studies on what they called "non-promotable" work: the tasks that need doing, benefit the organization, but do not advance the person doing them. Note-taking. Committee work. Onboarding. Coordination. The same work patterns now usually grouped under "glue work" and invisible work. The headline finding: women spent roughly 200 more hours per year on non-promotable work than men in equivalent roles. Two hundred hours is five weeks of full-time work. Five weeks a year, every year, of effort that did not show up in promotion files.
That 200-hour gap is one part of the cost. The other part is what the gap actually costs in compensation and advancement.
Matysiak and colleagues, published in Work, Employment and Society in 2025, ran a peer-reviewed study with an unusual design: they held performance data constant and varied only how visible the worker was to the organization. Workers with identical measured performance, when assigned to a full-time remote arrangement, were 11% less likely to be promoted and 9% less likely to get a raise than their on-site peers. Same numbers, same outputs, less visibility, fewer promotions and lower pay. The penalty is not for doing less work. The penalty is for being seen doing less work.
Stack that against the lived experience of younger professionals. TriNet/Wakefield Research found that 62% of younger workers feel blindsided by their performance review, and 74% say they are in the dark about how their performance is actually being perceived. Blindsided is what happens when the work was real but the record was thin and the calibration meeting was a story told without you in it.
This is the documentation gap and the data on what it costs. (See the research on invisible work and promotion outcomes.) The work happens. The record does not. The promotion decision gets made anyway, from whatever fragments are still surfaceable in the room.
The conclusion from the literature is consistent across three independent peer-reviewed studies and one large workplace survey. The work itself is not the problem. The visibility of the work is the problem. Visibility is a function of documentation. Documentation is the only piece any individual can actually control.
How do you capture invisible work without making documentation your second job?
The mistake is assuming the answer is "be more disciplined."
You are not going to be more disciplined. You are coming off a meeting that ran long, you have a Slack queue, the kids need pickup, and the part of you that is going to sit down and write three paragraphs of professional prose about the mentoring session that just ended is not the part of you that exists at 5:47pm on a Tuesday. Discipline is the wrong frame. Friction is the right frame.
The capture system that survives contact with a real workday is the one that requires no thinking. Thirty seconds. In the moment. In whatever format produces the least resistance.
For most people that is voice. Pull out your phone between meetings. Walk to the kitchen. Say what just happened in your own words. "Just finished coaching R. through the client call prep. Hit on scoping language and how to push back on a request without saying no. They are running the call solo tomorrow." Done. Eighteen seconds. Tagged later, polished later, surfaced when you need it.
What turns thirty seconds of voice into a review-ready paragraph is the pipeline that runs underneath. impactrackr's four-stage AI (Extract, Rewrite, Tag, Assemble) produces three output levels: Original, Polished, Amplified. Original is your words. Polished is your words cleaned up. Amplified is the version you would write if you had an hour. Tags pull from the same 29-tag taxonomy automatically, so mentoring lands under mentoring and process-improvement lands under process-improvement without you thinking about it. The 0% fabrication rule is non-negotiable: the system does not add accomplishments, it does not invent outcomes, it only restructures what you actually said. The system that prevents reconstruction is the documentation system itself, and the only way it works is if the capture step is cheap enough that you actually do it.
impactrackr works on phones and laptops today. The iOS app is in TestFlight beta. Android is in development. Free during beta.
Closing
Six months from now, the form is going to be open on your laptop again. The metrics section will fill itself in. The collaboration and mentoring and team-contribution section will be a blank box waiting on your memory of a year that has already started slipping.
You can either have a record of the invisible work or you can have the version your tired brain can reconstruct on a Sunday night. Those are the two options. There is no third option where memory holds up.
Start before you need it. The bar is thirty seconds and the compound is a year of dated, tagged, attributable wins that walk into the review conversation with you.
Start free. Early access, no credit card, no catch.
About impactrackr. impactrackr is the 30-second voice capture habit that builds your career documentation while you do the work. Designed for the professional whose most valuable contributions never make it onto the form. Free during beta. Start before you need it.