You're on the 35-minute train into London Bridge. You've got a laptop in your bag, emails to answer, an article to finish. But the seats are arranged three-abreast, there's no table, nowhere to rest the laptop, nowhere for a coffee. So you sit there, useless, for the whole ride. Meanwhile, somewhere in a rail company boardroom, engineers have pitched the Treasury on a very different problem: how to shave 4.7 minutes off the journey time. The Treasury says yes immediately, hands over a stack of money, because time-saving is a number they know how to fund.1 Nobody pitched a table. A table doesn't show up in the model.
This is the trap: an organization builds a dashboard, the dashboard measures something real but partial, and over time the organization starts optimizing the dashboard instead of the thing the dashboard was supposed to represent. Journey time is measurable, fundable, and legible to a finance department. "Whether the 35 minutes felt productive or wasted" is none of those things — so it never gets a budget line, even though it's the actual experience the customer is having.2 The rail company didn't fail because it lacked good engineers. It failed because it had a metric that could absorb capital and a customer experience that couldn't.
Rory Sutherland's diagnosis is specific: finance people fetishize cost-saving and automation because both appear to go straight to the bottom line or can be used to justify a lower price.3 Nobody in that structure is accountable for the customers who quietly disappear because they wanted a table, a bit of comfort, a human being who was pleasant to deal with. The 4-minute journey-time saving is visible in a spreadsheet. The 25 minutes of wasted productivity from bad seating is invisible in the same spreadsheet, even though it's larger, because nobody built a column for it.
Sutherland tells a companion story that sharpens the same point from the other direction.4 Royal Mail once spent a fortune improving operational efficiency and reliability — and saw no measurable change in customer sentiment. A marketing director named Alex Bachelor suspected why and tested it: the single biggest determinant of how people felt about Royal Mail as a brand was whether they liked their postal worker. A grumpy, punctual postie scored worse than a friendly one who occasionally missed a Wednesday. The metric the company had been optimizing — delivery reliability — barely moved the actual experience customers were forming an opinion from.
You're in a quarterly planning meeting. Someone presents a slide: average delivery time down 12%, cost-per-unit down 8%. The room nods — the numbers went the right direction, so the quarter reads as a win. You ask the question the room isn't asking: what did customers actually feel differently this quarter? Nobody has that number. It isn't that nobody cares — it's that nobody built the column.
Later that day you're reviewing a customer-service transcript, the kind that usually gets skimmed for compliance and archived. A customer is furious about a delayed order. The rep who handles the call doesn't rush them off script — she stays with the person, absorbs the anger, and by the end the customer says something like "well, at least you were straight with me." That interaction will never show up as a line item. It happened between the metrics, in the part of the business nobody was told to measure. You start asking, in every review from now on: what's the equivalent of the table on this train? What's the visible number we're chasing, and what's the invisible experience it's quietly costing us?
Here's the part that isn't really about trains. Any time a budget gets allocated, someone in the room has to make a case for it, and the case has to be legible to whoever's holding the checkbook. "This will save 4.7 minutes" is legible — you can put a number on it, compare it against the cost, run a return-on-investment calculation, and get a yes. "This will make the ride feel less wasted" has no unit. You can't put it in the same spreadsheet column as the timesaving, so it never even enters the competition for the money. It isn't that the invisible thing loses the argument — it's that the invisible thing never gets to make an argument at all, because arguments in a capital-allocation process have to be denominated in something measurable, and "felt less wasted" doesn't have a denomination.13 This is quietly how organizations end up with metrics that outlive their usefulness: not because anyone chose the metric over the customer, but because the metric was the only thing capable of competing for the budget line in the first place.
You can catch this pattern before it costs you 25 minutes on every train, every day, forever. The tell isn't a bad decision — it's a decision that nobody can articulate a customer-experience case against, because the customer-experience case was never built. Watch for a project getting funded where the pitch deck has exactly one slide with a number on it, and the number is an efficiency or cost metric. Watch for a debrief where "did the metric move" gets asked and "did anyone notice, and how did it feel" doesn't. Watch, especially, for the word "obviously" attached to a decision — "obviously faster is better," "obviously cheaper is better" — because "obviously" is usually standing in for "we didn't build the alternative case, so this one won by default." The Eurostar engineers weren't wrong that 4.7 minutes matters. They were operating in a room where nothing else got to matter, because nothing else had been made visible enough to compete.
A single wasted 25-minute train ride is a mild annoyance. The reason this pattern is dangerous isn't the single instance — it's that the metric that won gets reused as the standard for every future decision on that line, while the cost that lost never gets tallied anywhere, so it never accumulates into a number large enough to force a reversal. Multiply one commuter's 25 minutes by every commuter, every weekday, for the years the seating configuration stays in service, and you get a genuinely enormous aggregate loss of productive time — larger, almost certainly, than the value of the 4.7 minutes saved. But it will never appear as a line item anywhere, because it was never measured in the first place, and a cost that isn't measured can't be added up, escalated, or used to justify reversing the original decision. The asymmetry compounds silently: the winning metric gets tracked and defended every year; the losing cost gets rediscovered fresh, as a mild personal annoyance, by every individual rider, none of whom has the standing or the data to make it an organizational problem.
Sutherland doesn't offer a documented case of a rail company reversing course on this — the honest picture is that this trap, once set, tends to stay set. What would have had to be different is not more data (the invisible cost is invisible precisely because nobody collects it) but a different question in the room at decision time: not "what does this save," but "what does this cost that we're not measuring." Asking that second question doesn't require new data collection — it requires someone in the room whose job is to notice what the dashboard doesn't show, and to say so before the decision, not after the complaints start arriving. That's a governance change, not a metrics change, and it's the reason this page's Implementation Workflow above ends on a habit (asking the missing-column question in every review) rather than a proposed new metric — the fix isn't a better number, it's a standing objection nobody currently has the role to raise.
The strongest evidence for this pattern is that it recurs across completely different categories — trains and postal delivery — with the identical shape: a legible, fundable metric crowding out an illegible, unfunded experience.14 The open tension is that not every unmeasured thing deserves a budget line; an organization that tried to fund every soft, unquantifiable customer feeling would paralyze itself. The honest unresolved question is how you tell the difference between "this metric gap is costing us the business" (the train, the postie) and "this is just noise" — Sutherland doesn't offer a general test, only retrospective diagnosis after the damage is visible.
Sutherland's framing sits close to Goodhart's-Law-style critiques of metric-driven management generally, but he adds something those critiques often lack: a mechanism for why the blind spot forms specifically at the finance function. It isn't that finance people are careless — it's that cost-saving and automation are the two levers that visibly and immediately affect the number they're accountable for, so the organization's attention structurally drifts there regardless of individual intent.
Behavioral-mechanics — Goal Dilution Effect. Goal dilution describes how adding weakly-related sub-goals to a pursuit reduces motivation toward the actual target — attention gets spread thin across proxies that all feel like "progress" without moving the real outcome. The Eurostar case is goal dilution at the organizational scale: journey-time-saved becomes a proxy goal that absorbs capital and attention that should have gone to the actual target (a productive, pleasant 35 minutes), and because the proxy is easier to measure, it wins the resource fight even though it moves the true target less. The insight neither page reaches alone: proxy-goal capture isn't just an individual willpower failure, it's what happens automatically whenever an institution builds a dashboard — the dashboard becomes the goal.
Business — Brand Equity Metrics Beyond Purchase. That page (built in this same batch) documents the opposite move: a premium-brand operator deliberately building a wider metric set — hook rate, retention, cohort return — specifically so the org doesn't fall into the Eurostar trap of chasing the one visible number (purchases) while the brand quietly erodes. Read together, the two pages form a before-and-after: this page shows what happens when an organization has only the narrow metric; the other shows the deliberate discipline of building broader ones before the narrow metric can capture all the attention.
Sharpest implication: every dashboard is also a blind-spot generator — the moment you build a metric, you've created a gravity well that will pull organizational attention toward itself and away from everything it doesn't measure, whether or not that's the right allocation.
Generative questions: