
Measuring Invisible Work - MAC156
here is a specific moment this episode is built around, and you have almost certainly lived it. You are writing up your contribution — for a review, a self-assessment, a conversation you are trying to prepare for. You know what happened. You know it mattered. You get to the word saved — "which saved us about" — and you stop. Because the next word is supposed to be a figure, and you do not have one. So you write "a lot of time" instead, and the sentence quietly dies. That is the failure point. And it is not a laziness problem. It is a permission problem. You believe that to put a dollar figure on your own work, you would need the real number — HR's actual turnover model, finance's actual cost-per-incident, something audited and defensible that lives in a system you do not have access to. Since you cannot get it, you conclude you are not entitled to a number at all. And so you arrive at every performance conversation armed with adjectives while the organization around you runs entirely on estimates. The revenue forecast in that room is an estimate. The headcount plan is an estimate. The projected savings from the reorg somebody presented last quarter was, structurally, a guess with a confidence level and a name attached to it. Nobody in that room is working from audited truth. They are working from numbers somebody was willing to own. You are not missing the data. You are missing permission to estimate. There is a second thing working against you: you will aim low. Most people place their own contribution below where outside evidence puts it, and the error runs in that direction far more often than the other. The career self-diagnostic episode (MAC-152 at managingacareer.com/152) covered this in detail — the calibration gap is real, it is directional, and knowing about it does not automatically fix it. So when you finally do reach for a number, your instinct will be to shave it. Know that going in. Where the numbers actually live. There are three sources, and you probably have access to all three right now. The first is public. There is a whole industry of aggregated employer data sitting in the open, and it covers exactly the metrics leadership already watches. On turnover, for instance, the pooled employer numbers put the cost of replacing someone at roughly 40 percent of their salary for a frontline role, around 80 percent for a mid-level professional, and north of 200 percent for a leader. That is not your company's model. It is a range your company's model almost certainly falls inside. The second place is your own organization's ordinary paperwork — the one people never think to check. The job posting for the role you helped fill has a salary band printed on it. Your recruiting team's open requisitions tell you how long a seat stays empty. Your own calendar tells you how many hours a month go into the meeting you eliminated. None of that is confidential. It is just sitting there, un-mined, because you have never thought of a job posting as a pricing document. The third place is a person. Somebody in finance, HR, or ops owns the cost you are trying to estimate, and they will usually answer one specific question if you ask it as a specific question. Not "what does turnover cost us." That gets you nothing. Try: "When we backfill an analyst on my team, roughly how long does the seat sit empty?" That is answerable in one line, and it is the only piece you were missing. The anchor number. The thing you build out of those three sources has a name, and naming it is what makes it usable. Call it an anchor number: a public or borrowed figure you pick up on purpose, label out loud as borrowed, and use as the starting point of an estimate rather than the end of one. Forty percent of salary is not what your company pays to replace somebody. It is the number you multiply against until somebody hands you a better one. An anchor number is not your company's number. It is a number you can defend while you wait for a better one. The one rule that matters: say where it came from. The whole thing works if you say "using the industry benchmark of roughly 40 percent of salary" and it collapses the instant you say "our turnover cost is." One of those is an estimate with its sources showing. The other is a claim you cannot back, and the moment somebody checks, every other number you have ever given them gets re-examined too. The repricing pass — four moves. Take one entry and walk it all the way from a note to a sentence. Move one: name the metric. Which of the four does this touch — turnover, ramp time, rework, or escalation? Pick one. Most entries touch two, and you will be tempted to claim both. Do not. Two half-credible numbers read worse than one solid one. Move two: find the unit. Every metric has a thing you count. Turnover counts people. Ramp counts weeks. Rework counts hours. Escalation counts incidents. Figure out what you actually moved and how many of them. Move three: attach the anchor. Price the unit using one of your three sources. Move four: put the assumption inside the sentence. Not in a footnote. In the sentence, out loud, where the person hearing it can see exactly what they would have to disagree with. Here is what that looks like, fully worked. The entry: I cut the new analyst's ramp from six weeks to three. Metric: ramp time. Unit: weeks of one person's capacity. Anchor: the salary band on that role's own job posting — say it topped out around ninety-five thousand. Fully loaded, with benefits and payroll tax, that is conventionally somewhere around a quarter to a third above base, so call it one hundred and twenty-three thousand a year, or roughly twenty-four hundred dollars a week. A person still ramping is not producing nothing — call it a third of full output early on. So each week you removed is worth about two-thirds of twenty-four hundred. Sixteen hundred dollars. Three weeks, call it five thousand. And the sentence: "I cut the new analyst's ramp from six weeks to three. Using the posted band for that role and assuming a new hire runs at about a third of output in month one, that's somewhere around five thousand dollars of recovered capacity — and I'd expect the same on the next two hires, because the onboarding doc is reusable." Five thousand dollars is not an impressive number. That is the point. It is a number that survives a follow-up question, and every step of it can be walked backward in front of a skeptic. A number you can walk backward beats a bigger number you cannot. The discipline the whole episode rests on: show the arithmetic instead of presenting the total. Label how confident you are in each input — you can be highly confident about a posted salary band and only medium confident about the output assumption, and saying so makes you more credible, not less. And when you are choosing between the modest figure you can defend and the impressive figure you would have to defend with adjectives, take the modest one. Always. Two more examples across different work: Rework. Entry: I rewrote the recurring report so people stopped guessing at it. Metric: rework. Unit: hours. Four people were spending about ninety minutes a month each re-deriving what the report meant — that is six hours a month, seventy-two hours a year, call it two working weeks the department stops burning. High confidence on the hours, because you can ask the four people. Medium confidence on whether every one of them actually reinvests it. Escalation. Entry: I caught the client discrepancy on Friday before it reached their finance team. Metric: escalation. Unit: incidents. You do not have to model this one — go find the last time it was not caught. "The last time this got through, it took four people two days and a written apology." That is your anchor. It happened at your own company, and it costs you one search to find. One number vs. a track record. This is where the episode turns on itself, and on purpose. Because the natural response to a framework this clean is to use it once — the week before your review, on three entries — and then never again. That version mostly does not work. Not because the numbers are wrong, but because of what one number sounds like when it arrives alone. One number that appears the week before performance conversations reads as advocacy. It reads as somebody who went looking for ammunition, and the person hearing it discounts it accordingly. Ten numbers, produced steadily over two quarters, read as something else entirely. They read as a person who tracks their own work, which is a character claim you cannot make about yourself and can only demonstrate. Nobody has to believe you about it. They just look at the record. One number is a claim. A quarter of them is a track record. This also matters because of where those numbers have to travel. Your manager is not your career sponsor (MAC-139) — the decisions that move you happen in rooms you are not in, made by people who have never watched you work. What crosses into those rooms is whatever your advocate can repeat from memory. "She's great with the new people" does not survive that trip. "She cut analyst ramp by half and the onboarding doc is reusable" does,...















