Who Owns That Number?

Why Organizations Can Add a Measure and Never Subtract One

Issue 280, September 3, 2026

Which of these measures would you remove?

I’ve started asking that in every room where a dashboard is on screen, and the answer is almost always the same silence. Not defensiveness. Not annoyance. But puzzled looks, as if I’d asked which of their body parts they’d like to give up. Which for me is always a fun visualization as I wonder what they might say.

Then, as one might expect, the reasons start to arrive like clockwork. This one feeds the board report, the CEO says, so it has to stay. That one’s wired into the compensation formula, the Chief People Officer adds. This one’s pulled by an integration somebody in finance built two years ago for a CFO who has since left, and nobody in the room can fully explain it. And when I ask what they actually and truly inform, or what anyone in the building does differently because of them, nobody seems to know. In one session, a finance director finally said the quiet part out loud. “I assumed somebody needed it.”

But they have always existed, or they exist because…fill in the blank here.

That is the reason for this issue, because what I keep watching is not a failure of analytics, and it isn’t a lack of discipline. It’s something stranger and far more common. These organizations have built a measurement system they know how to add to and have no way to subtract from, and the pile has started making decisions nobody signed off on. I have come to think of it as a measurement ratchet, and once you can see it, you will see it in every organization you walk into, including your own.

The Initiative Dies. The Reporting Lives On.

A few weeks ago, in After the Theater (Issue 277), I laid out a ninety-day sequence for organizations that have recognized Transformation Theater in themselves and want to do something other than launch another working group at it. That issue was about stopping things and being honest about how much change a group of humans can absorb at one time. It ended with a question. Which initiative would your organization kill first, and what has stopped you from ending it already?

This issue points that question somewhere harder.

Here’s what I’ve watched happen over and over. An organization does the difficult work. It runs a portfolio audit, it names the initiatives that have quietly died, it stops three of them, and it earns back some trust from the workforce. Real and substantive progress. No question.

And then, a year later, dashboards in use across the organization still carry the measures those initiatives installed. Nobody noticed where they came from, only that they were there and should remain there.

The initiative ended. The reporting obviously didn’t.

That’s the engine underneath a great deal of what we call change and transformation fatigue. Initiatives have sponsors, which means ending one is a conversation with a human being or group of human beings who will be disappointed. Those conversations are hard but at least they are possible. A measure frequently has no sponsor at all. Which sounds like it should make removal easier but does exactly the opposite, because there’s nobody willing to take a stand to end it as it is so ingrained in so many different ways across the organization, the board, the market, and yes, even the public.

We Reach for Numbers Where Trust Is Thin

To understand why, it helps to go back a bit over thirty years.

In 1995, John Kotter published Leading Change: Why Transformation Efforts Fail in Harvard Business Review, and the diagnosis he offered has been quoted in nearly every change management deck produced since. That same year, the historian Theodore Porter published Trust in Numbers: The Pursuit of Objectivity in Science and Public Life. Porter set out to explain why modern institutions are so committed to counting things. The conventional answer is that numbers spread because they work. Porter argues that this is close to exactly backwards, and he builds the case through the history of actuaries, engineers, and the rise of the accounting profession.

His conclusion is that quantification is what he calls a technology of distance. Reliance on numbers, in his words, minimizes the need for intimate knowledge and personal trust, which is what suits it to communication that travels beyond a community that knows one another.

Numbers do their heaviest work where authority and trust are thinner than they should be. Humans love numbers; we love counting, we generally like positive numbers over negative numbers, we rely upon numbers to generate the feeling of accomplishment or completion. We love numbers to measure performance across nearly everything in life and business. (Ask anyone wearing a step counter how they feel at 9,400, and here I admit I am very much one of those people.)

Writing about the engineers who built cost-benefit analysis, Porter concluded that the regime of calculation was imposed not by all-powerful experts, but by relatively weak and divided ones. The appeal of numbers, he wrote, is especially compelling to officials who lack the mandate of a popular election, or divine right.

Which inverts the framing most of us carry. Organizations don’t adopt elaborate measurement because their judgment is strong. They adopt it because their judgment is exposed.

Two diagnoses published the same year, thirty-one years ago. One of them we quote constantly and haven’t absorbed. The other we never read at all. I’m not sure which is the more reflective of the challenges our very humanness creates in today’s environment.

There is, however, a distinction underneath this that I use constantly with clients. Picture a veteran surgeon asked why she chose a procedure, then a new hire asked to produce the protocol she followed. Same question, two very different relationships to trust. Some groups get to make a call and explain it afterward. Others have to show their work up front, in a form anyone can check, because nobody will take their word for it. Lorraine Daston and Peter Galison named that second thing in 1992, mechanical objectivity, and Porter picks it up and runs with it. Rules pile up where the standing is thin.

Put that into a conference room you’ve actually sat in. A leadership team that trusts its own judgment, and is trusted by its board, argues and then decides. A team that isn’t trusted builds a scorecard. The scorecard is a way of asking permission without having to ask anyone for it. None of which is a criticism of the people involved. Building it is a rational response to the conditions they believe they are in.

When Numbers Got Cheap

Here’s where Porter’s world and ours part company.

The institutions he studied operated under very different conditions. Numbers were scarce, slow, expensive to compile, and hard to argue with, and that scarcity was itself the source of their authority. We went the other way entirely. Every system in your organization now produces data whether anyone asked it to or not, and a dashboard that would have taken a department a quarter to assemble in 1975 assembles itself before you finish your morning coffee.

Think about what happened to photographs. When the film was expensive, and a roll gave you maybe twenty-four shots, you thought before you took the shot, and you didn’t know what you shot until it got developed. Now we take four hundred shots on a Saturday and look at none of them. The pictures got cheaper, and they got less interesting to us at the same time. Numbers went the same way, and we never really adjusted. The conditions have changed completely, but our behavior at its core didn’t change at all.

One more difference is worth putting out there. Porter’s engineers and actuaries built their own instruments. Ours increasingly arrive from vendors, with a maturity model attached and a certification available. Trust comes bundled with them. The data is assumed good, and whatever it produces is assumed authoritative.

Let me give you an example that should rattle that default. The Blackbaud Institute’s June 2026 report Bridging the AI Effectiveness Gap, based on March 2026 surveys run with Edge Research, reports that the average organization saves roughly $503 per employee per week using AI. The number looks authoritative. It came from a named institute, it has a research partner attached, and it is precise enough to benchmark against. And to be fair to the report, it is transparent about how the figure was built in a way most vendor research is not. Respondents were asked what an hour of their time was worth and how many hours a week they saved, and the two were multiplied.

Now look at what the arithmetic does. Two guesses, one of them a guess about your own hourly worth, become a dollar figure carried out to the single digit. The calculation is real. It is also entirely a property of the multiplication and has nothing to do with the world. That’s mechanical objectivity showing up in 2026 exactly as Porter described it in 1995. You’ll find the same construction in every industry’s productivity claim, including the ones your organization has made and has continued to take actions upon or leverage for decision-making.

I fell into the same trap myself. For years I cited a figure on the cost of unproductive meetings that turned out to have no locatable primary source at all. I’d seen it everywhere; it lodged in my mind, and I fell back on it again and again. When I went looking for more on the topic, I found it traced to a magazine item from 1989. It circulated for decades because it was specific, and specificity reads as verification. I pulled it, which was uncomfortable, and which is the smallest possible version of the thing this issue is about.

The Ratchet, Not the Pendulum

When I started drafting this issue, I saw the situation as a pendulum. Organizations swing from too little measurement to too much, never finding the middle. It was a comfortable framing, and the longer I sat with it, the more clearly I think it’s wrong.

A pendulum comes back. Measurement infrastructure doesn’t. In more than twenty years of this work, I’ve watched hundreds of measures get installed and can count on one hand the ones I’ve watched get deliberately retired, despite advice, counsel, very direct statements, and even lengthy assessments that should have changed minds. What we contend with is a ratchet. It turns one way, holds where it stops, and every turn adds to the load.

And the weight isn’t the whole problem. This is the part leaders consistently underestimate. Every measure that survives its first year gets wired into something else. A monthly report. A compensation formula. A cell on a board slide the board has come to expect. An integration built to feed it. Somewhere along the way it turns up in a grant commitment, a client agreement, or a regulatory filing, and at that moment it stops being a measure and becomes an obligation.

Each of those attachments turns removal into a negotiation with people who aren’t in the room. So by the time anyone notices a measure has outlived its usefulness, removing it is no longer an analytical act. It’s a political one. And so it stays, steering decisions from inside the plumbing, long after anyone could tell you what it was for.

Which raises a question worth sitting with. Who in your organization is allowed to remove a number? Not question it in a meeting. Not footnote it. End it. Can you name the person? Would they know it was their job?

A measurement system without a mechanism for retiring measures is not a measurement system. It is sediment.

What the Sediment Costs Now

For most of the time I’ve been doing this work, the cost of the ratchet was attention. Ninety minutes of a review covering eight of forty measures. Executives reading dashboards the way the rest of us read terms of service. That cost is real, and it’s the one most leaders would name if you asked them.

Two others matter more, and the second is newer than people think.

The first is integrity. A pile that grows without anyone tending and curating it eventually degrades, and it degrades quietly. In the Blackbaud study I mentioned earlier, fewer than twenty percent of professionals rated their organization’s data health as excellent, and the rate was roughly double that at the organizations the study called AI-Adaptive. I wouldn’t read that gap as a story about AI maturity so much as a story about data and measurement housekeeping. Organizations that only ever add have worse data, for the same reason a house where nobody takes out the trash has a horrible smell.

The second changes the stakes, and it’s why this argument opens a season of The Human Factor Podcast rather than sitting in the middle of one.

A neglected pile of measures used to just sit there. Wasteful, but inert. It isn’t inert anymore. Your AI systems read it, summarize it, and act on it. Every abandoned metric, every stand-in that stopped tracking what it was built to track, every field somebody quit maintaining in 2021, now feeds a system that hands it back in fluent, confident prose with no hint of how old it is or where it came from. And we act on it.

I wrote in The Confidence Transfer (Issue 271) about how quickly people defer to a confident answer. Put that on top of a sedimented set of measures, and you get something worse than either problem alone. The machine doesn’t know which of your numbers are load-bearing and which are fossils. It treats them identically. And your people, reading the output, are two steps away from the judgment that produced the measure underneath, if any judgment ever did.

Two issues ago, in The Proxy Collapse (Issue 278), I made the case that when the real thing is hard to measure, we settle for a stand-in, and then forget it was ever a stand-in. A fossil metric is that forgetting, left to age. The pile we can count keeps growing. The judgment that would tell us which part of it matters hasn’t grown at all.

The Retirement Practice

Let me be clear about where I stand. I’m a measurement believer. I write and talk a great deal about measuring what matters. Measurement is how we assess progress, function, and impact, and an organization that can’t tell whether it’s getting better isn’t humble; it’s blind. The problem was never that organizations have too many numbers. It’s that they have numbers nobody owns. That is a solvable problem, and the rest of this issue is how.

Tom Serena, closing out last season of the podcast, gave a new leader the best advice on this I’ve heard from anyone. Don’t come in to transform. Come in to find out what needs to be transformed.

I want to apply that literally, because it’s the whole practice in one sentence. A leader who arrives and redesigns the measurement system is coming in to transform. A leader who inventories what the organization actually produces, and asks what each item is for, is coming in to find out. The second takes longer and it’s the only one that works. You don’t have the standing to retire a measure until you understand what it does. Standing isn’t conferred by title. It’s earned by having looked.

Every measurement conversation I’ve been part of eventually arrives at whether a measure is still relevant. It’s the fourth of the four questions in the Measuring What Matters work, and by far the hardest, because answering it honestly requires a human being to stand behind a removal and say why. Adding a measure spreads the responsibility around. Removing one puts it on a single desk. And nobody has ever been fired for adding a metric.

So the practice has to make removal survivable rather than heroic. Here’s how I run it.

Inventory what is actually produced, not what is on the dashboard. The dashboard is the visible layer. Underneath it sit the standing reports, the scheduled exports, the board attachments, and the spreadsheet somebody keeps by hand every Thursday because a director asked for it in 2019. Expect the real list to run two or three times the size everyone believed. That surprise is itself the finding.

Put each item through the four questions. What is the purpose of this measure? What decision or action does it inform? How does it connect to other measures? Is it still relevant? If the answer to the second one is that it informs nothing, you’ve found sediment, and you should say so out loud in the room.

Trace the attachments before you touch anything. For every candidate, establish who uses it, what formula depends on it, which outside commitment references it, and what breaks if it disappears. This is what separates a retirement practice from a spring cleaning, and skipping it is how good intentions blow up a compensation cycle.

Name the person who owns the removal, by name, in the minutes. Not the committee. Not the function. A person who says on the record that this measure has served its purpose and here is why it ends. Organizations will try hardest to skip this one, because it’s the only one that costs anybody anything.

Put it on a cadence and let the cadence carry the weight. An annual review that is expected, scheduled, and normal is a completely different social event from a purge somebody has to champion. The cadence is what makes the previous step survivable, because the person doing it is following a standing practice rather than attacking a colleague’s work.

Measure the first cycle by what stopped. Nothing else. Longtime readers will recognize that line from Issue 277. The repetition is deliberate. It’s the same test applied one layer down.

Two cautions before you run this, because both have bitten people I work with.

The first is continuity. A metric with eleven years behind it can answer questions a new one can’t touch for another decade, and the person proposing its removal is usually the person least able to see who depends on it downstream. Retire carelessly and you destroy the comparability that made the measurement worth having in the first place.

The second is what comes back. In Noise, Daniel Kahneman, Olivier Sibony and Cass Sunstein show how much human judgment varies on identical cases, and every measure you retire is a place where that variation returns. Some of it will be wisdom. Some of it will be mood, fatigue, favoritism, and the last conversation someone had in the hallway. That is an argument for retiring on a cadence, with the attachments traced and a name in the minutes. It is not an argument for leaving the pile alone.

And one more, because I’ve watched this get misused. Coming in to find out has no natural end point. Discovery is comfortable in exactly the way a pilot is comfortable. It threatens no one, it produces something to report, and it can be stretched out forever by anyone who would rather not decide. If finding out isn’t time-boxed, and doesn’t end in a decision a named person owns, it isn’t patience. It’s theater at a slower tempo, which is harder to spot and worse.

The Cost of the Middle

Both ends of this are attractive for the same reason. Scarce measurement lets authority rest on status and intuition. Abundant measurement lets it rest on the dashboard. The balanced position, the one we all say we want, is the only one where a person has to stand behind a call and explain it. The swing isn’t a failure to find the middle. It’s a long flight from what the middle costs.

So the real test of a measurement system isn’t what it captures. It’s whether anyone inside the organization could survive taking something out of it.

Season 3 Begins in September

The Human Factor Podcast returns next Thursday, September 10, and this issue is the argument that opens it. Season 2 ended on a sentence I keep coming back to. Transformation isn’t an event; it’s a practice, and the shorthand for it is patience, repeated. Season 3 takes that as a theme, and the ratchet is that problem wearing different clothes. Adding happens at machine speed. Subtracting happens at human speed, when it happens at all.

Subscribe wherever you listen, on Apple Podcasts, Spotify, or YouTube.

And as the season gets underway, I’d like to hear from you. Tell me about a measure your organization actually retired. Who owned the removal, what broke, and what it cost the person who put their name on it. I’m collecting these across the season, and I’ll bring back what I learn.

Related Reading

After the Theater (Issue 277). The ninety-day sequence for organizations that want to stop things rather than launch something at the problem. This issue is what happens to the instruments those stopped initiatives leave behind.

The Proxy Collapse (Issue 278). A fossil metric is what this one looks like after it has aged.

The Confidence Transfer (Issue 271). Why deference to a confident machine answer happens by default rather than by discernment. Read it alongside this issue and a sedimented data estate stops looking untidy and starts looking dangerous.

Transformation Theater (Issue 275). How the performance of change substitutes for its substance. Elaborate apparatus is the tell.

The Permission Paradox (Issue 269). An organization’s instruments show a workforce waiting to be transformed. The workforce had already transformed itself in private.

Measuring What Matters. The four questions used in the retirement practice above, in their original form.

The Transformation Psychology Series. The research underneath the identity threat that shows up whenever someone defends a measure they built.

The Transformation Readiness Assessment. Worth a baseline before you pull on any of this.

On The Human Factor Podcast

From Diagnosis to Practice, Season 2 Episode 031. The Season 2 finale, tracing the arc of eighteen episodes and previewing where Season 3 goes.

The Long Game, Season 2 Episode 030. Tom Serena on pace and patience as the denominator of change, which is the argument standing behind this issue.

The Measurement Ratchet: Why Nothing You Measure Ever Leaves. Season 3, Episode 1. The companion episode, out September 10, going back seventy years to the research that called all of this in 1956, and further into what the retirement conversation actually sounds like in the room.

Research Sources

1. Porter, Theodore M. (1995). Trust in Numbers: The Pursuit of Objectivity in Science and Public Life. Princeton, NJ: Princeton University Press.

2. Daston, Lorraine and Galison, Peter (1992). The Image of Objectivity. Representations, No. 40, Autumn 1992, pages 81 to 128.

3. Blackbaud Institute and Edge Research (2026). Bridging the AI Effectiveness Gap: New Research on What Drives AI Impact and Trust in the Social Sector. Published June 24, 2026. Parallel surveys of 1,389 social impact professionals and 1,034 donors, United States, fielded March 2026.

4. Kahneman, Daniel, Sibony, Olivier, and Sunstein, Cass R. (2021). Noise: A Flaw in Human Judgment. New York: Little, Brown Spark.

5. Kotter, John P. (1995). Leading Change: Why Transformation Efforts Fail. Harvard Business Review, March/April 1995.

Join the Conversation

LinkedIn: https://www.linkedin.com/company/2040-digital

Substack: https://20forty.substack.com

The Truth About Transformation

The Truth About Transformation: Leading in the Age of AI, Uncertainty and Human Complexity. The book behind the method, on why most transformations fail and what changes when psychology comes first.

Speaking and Keynotes. If your leadership team needs to have the retirement conversation and cannot find a way into it, this is the kind of session I run.