What is the Mercator problem in business?

The Mercator problem is the idea that any metric, dashboard, or KPI is a projection of a complex reality onto a simpler picture—and like a flat map of a round Earth, that projection must distort something to be readable. Borrowed from cartography, where the famous Mercator map inflates Greenland and shrinks Africa, it describes how business metrics inflate whatever is easy to measure and crush whatever isn't toward zero. The distortion is structural, not a sign your data is wrong.

Is Greenland really bigger than Africa?

No—and that's the whole point. Greenland looks roughly Africa's size on a standard Mercator map, but Africa is about 14 times larger (30.3 million versus 2.2 million square kilometers). The Mercator projection inflates land the farther it sits from the equator. It's not an error; the map was built for navigation, not for comparing sizes. Your dashboard does the same thing with whatever it can measure most easily.

How is the Mercator problem different from Goodhart's law?

Goodhart's law describes what happens when people game a metric: once a measure becomes a target, they optimize the number instead of the goal. The Mercator problem is deeper and requires no gaming at all. Even a perfectly honest, un-gamed metric distorts on day one, because reducing a multi-dimensional reality to a single number mathematically forces some dimensions to stretch and others to shrink. Goodhart is a behavior problem; the Mercator problem is geometry.

Can you be too data-driven?

You can over-trust any single metric, which is the practical failure mode of data-driven decision making. Being data-driven is the right cure for gut-feel management, but every number is one projection with a built-in distortion. The fix isn't less data—it's reading data like a cartographer: name what each metric sacrifices, carry a second metric that distorts differently, and routinely check the number against the actual work it claims to represent.

What's the most accurate business metric?

There isn't one, for the same reason there's no undistorted flat map—no projection can preserve everything at once. The most useful approach is to choose a metric whose distortion you understand and can tolerate for the decision at hand, then pair it with a second view that bends in a different direction. Trust the places they agree; investigate the places they don't. That's how you avoid measuring the wrong thing while believing you've measured everything.

The Mercator Problem: Why Your Dashboard Has to Lie

On almost any world map, Greenland looks about the size of Africa. It isn't close. Africa's area is 14 times as large—Greenland is 2.2 million square kilometers, Africa is 30.3 million. The map isn't broken, and it isn't lying to trick you. It's a Mercator projection, drawn in 1569 for sailors, and it does one job superbly: it turns a line of constant compass bearing into a straight line so you can cross an ocean with a ruler. To manage that, it had to stretch the far north until Greenland swallowed a continent.

That trade-off deserves a name. Call it the Mercator problem: the moment you flatten something complicated into one clean, usable picture, you inflate one thing and shrink another toward nothing—and the cleaner the picture looks, the easier it is to forget you did it. Your dashboard is a Mercator projection. So is every KPI you track, every OKR, every "engagement score."

And we have never had more maps. At the 2026 Gartner Data & Analytics Summit, the headline was a gap: four out of five organizations increased their AI investments this year, and only one in five can demonstrate measurable ROI. More data, more dashboards, more AI—and decisions that aren't measurably better. This piece is about why. Not because your numbers are fake or gamed, but because every metric, by mathematical necessity, has to distort. You'll see what the Mercator problem really is, why it's sharper than "metrics can be gamed," which Greenland your dashboard is quietly inflating, and how to read a number like a cartographer instead of a tourist.

What the Mercator problem actually is

Start with the part almost everyone gets wrong: the Mercator map is not a mistake. It is a faithful, internally consistent, brilliantly useful representation of the Earth. It just can't be faithful to everything at once.

Here is the hard fact underneath it. A sphere and a flat sheet of paper have different geometry—a sphere has nonzero Gaussian curvature, and a plane has zero. Gauss proved in 1827 that you cannot convert one into the other without stretching, and the consequence for mapmakers is brutal and absolute: every cartographic projection necessarily distorts at least some distances. There is no clever workaround coming. No flat map of Earth can be perfect, "even for a portion of the Earth's surface." A map that gets area right gets shape wrong. One that gets shape right gets area wrong. You choose which truth to keep, and the rest bends.

The cartographer Mark Monmonier, in his classic How to Lie with Maps, put it more bluntly than any analyst would dare: "Not only is it easy to lie with maps, it is essential. To portray meaningful relationships for a complex, three-dimensional world on a flat sheet of paper or screen, a map must distort reality." Read that as a sentence about metrics and nothing changes. A metric takes a high-dimensional reality—a team, a product, a customer relationship with dozens of moving parts—and projects it onto a single number you can read at a glance. That projection is exactly what makes it useful, and exactly what makes it distort. The Mercator problem is simply this: the act of flattening is the act of distorting. You don't get one without the other.

Why this isn't Goodhart's law

It's tempting to file this under the usual warnings, but the Mercator problem is a different, more unsettling thing.

Goodhart's law—when a measure becomes a target, it stops being a good measure—is about gaming. It needs an incentive and a person who responds to it; the number goes bad because people start optimizing the number instead of the goal. Surrogation, the accounting term, is a cousin: a measure of a construct "evolves to replace that construct" in people's minds, like "a manager tasked with increasing customer satisfaction who begins to believe that the customer satisfaction survey score actually is customer satisfaction." And "the map is not the territory" is the philosophical grandparent of all of it—a useful reminder that the model and the world differ.

The Mercator problem sits underneath those. It says the distortion is there before anyone games anything, before anyone confuses the proxy for the goal, on day one, with a perfectly honest team and a perfectly clean dashboard. It's not a behavioral failure you can train away. It's geometry. Even your most trustworthy number is a projection, and projection mathematically requires that some dimension of the truth gets stretched and another gets crushed. Goodhart tells you what happens when people fight the map. The Mercator problem tells you the map was already bent before they showed up. Both are real. Only one of them survives a culture of saints.

Every dashboard inflates a Greenland and shrinks an Africa

Let me steelman the thing this argues against, because it isn't dumb. "Be data-driven." "You can't manage what you can't measure." This is the right correction for a team that runs on gut feel, loudest-voice-in-the-room, and vibes. Measuring forces honesty, exposes wishful thinking, and turns arguments into experiments. Nobody serious is telling you to throw out the dashboard.

But here is the trap. When you apply the Mercator projection to a world map, it "inflates the size of lands the farther they are from the equator." Metrics inflate the same way—not by latitude, but by measurability. Whatever is easy to count gets a clean axis and a big number, your Greenland. Whatever is hard to count gets compressed toward zero, your Africa—still 14 times more important, now nearly invisible on the chart. The distortion isn't random. It tracks exactly what your instruments can see.

You can watch this happen in the wild. One engineer described it precisely on Hacker News: "MTTR is easier to measure and manage as a manager... Rinse and repeat that few hundreds or thousands of times while mounting MTTR KPI and you will see that number improve." Meanwhile, "MTBF is much, much, tricker to measure... It's about 'excellence in engineering' which is not measurable nor controllable." Mean-time-to-recovery is Greenland: visible, trainable, dashboard-friendly. Deep engineering quality is Africa: enormous, decisive, and shrunk to a sliver because no instrument captures it cleanly. Another engineer watched a whole culture tilt this way under a manager who insisted only work "that can be explicitly measured" be performed: "they forbade library upgrades, refactors... Important but thankless work gets left behind." The end state he describes is the map eating the territory: "a bunch of features customers hate (eg dark patterns), and a rickety codebase that everyone is disincentivised fix or improve."

This is also where most vanity metrics come from. They aren't lies; they're the Greenlands of your business—page views, talk-time, hours logged, raw message counts—measurable surfaces that loom huge precisely because they're easy to project, while the thing you actually care about hides in the distortion. The danger scales with polish. A messy spreadsheet invites suspicion; a beautiful real-time dashboard invites belief. The cleaner the map, the better it hides what it had to crush—which is how metrics that mislead do the most damage exactly when leadership trusts them most. Wells Fargo is the cautionary monument: as the cross-sell metric got harder to hit, "the rate of misconduct rose," ending in roughly 3.5 million unauthorized accounts. "Accounts per customer" was a gorgeous Greenland. "Are we actually serving this person" was the Africa that went to zero.

How to read your dashboard like a cartographer

You can't escape the Mercator problem—Gauss closed that door. But cartographers have worked productively with distortion for four centuries, and their habits transfer. The goal isn't a perfect map. It's knowing your map's distortion well enough that it stops fooling you.

Name the projection you chose

Every map announces its trade-off if you ask. The Mercator keeps bearings true and sacrifices area; the Gall–Peters projection keeps area true and "distorts most shapes." Neither is "the accurate one." Do the same for each headline number: what does this metric preserve, and what does it sacrifice to do it? "Weekly active users" preserves reach and sacrifices depth. A CSAT score preserves a comparable number and sacrifices the texture of why. Write the sacrifice down next to the metric. A trade-off you've named can't ambush you.

Find your Greenland and your Africa

For any dashboard, ask two questions. Which dimension is inflated because it's easy to measure? And which dimension—the one that would actually decide success—is being crushed toward zero because nothing on the board captures it? Often the most important variable in your business has no clean axis: trust, morale, technical health, whether the work is any good. This is the same blindness behind the 100-year-flood problem: the rare, unmeasured risk gets priced at zero until it isn't. It's worth distinguishing from the sightline problem, too—there the failure is not seeing something at all, while here the thing is in plain sight on the chart, just stretched badly out of proportion. Naming the Africa is most of the work. You can't fix a distortion you refuse to look at.

Carry a second projection

No sailor navigates with one map. The fix for dashboard distortion isn't a better single metric—Gauss says there's no such thing—it's a deliberate second projection that distorts differently. Pair the quantitative with the qualitative: read the MTTR chart and the incident write-ups. Pair the leading indicator with the lagging one. When two maps drawn on different principles agree, trust the overlap; when they disagree, you've found exactly where your main map is bending. This is also the honest version of measuring activity versus outcomes—activity metrics are one projection, outcomes are another, and you need both on the table.

Don't confuse the map for the territory

A projection is a stand-in for a place you could, in principle, go see. The most expensive proxy metrics errors come from treating the number as the thing instead of as a pointer back to the thing. A meeting "engagement score" or a talk-time leaderboard is a textbook Mercator projection of collaboration: it inflates whoever spoke most and shrinks the quiet engineer who solved the problem to a rounding error. The fix is to go look at the territory—the actual work that got made. This is one reason a shared, persistent canvas matters: when a session leaves behind a real artifact instead of only a transcript score, the record is the work, not a proxy for it. Tools like Coommit lean on this directly—contextual AI that watches the canvas while it hears the conversation captures what the team actually built, so you're checking the metric against the territory instead of mistaking one for the other. The dashboard points; you still have to walk over and look.

Re-project on a schedule

A map of a coastline goes stale as the coast erodes. Metrics drift the same way: a number that honestly tracked value a year ago can quietly invert as the business changes, the way even a North Star metric can outlive the strategy it was meant to serve. Put a recurring date on the calendar to ask of every core metric: is this still pointing at something we care about, or are we steering by a star that moved? Re-projecting is cheap. Sailing confidently toward the wrong Greenland is not.

The most dangerous map is the one you trust completely

A century before dashboards, Boston Public Schools made news by switching its classroom maps from Mercator to Gall–Peters in March 2017—the first U.S. district to do it—not because the new map was "right," but because a generation had grown up believing one particular distortion was the world. That's the real risk of the Mercator problem at work: not that your metrics distort, but that you forget they do, and start managing the map.

So stop hunting for the undistorted number. It doesn't exist, any more than a flat map without stretching exists. The question was never whether your dashboard lies—it has to, the same way Greenland has to look too big. The question is whether you know which truth it's preserving and which one it's crushing to zero. As AI hands every team more maps faster, the edge won't go to whoever has the most dashboards. It'll go to whoever can still read the distortion—name the projection, find the shrunken continent, and remember there's a territory under all that ink. Pick one metric your team treats as gospel this week, and ask out loud: what's the Africa this is hiding?