Check What a Company Thinks of Its Customers

Have you tried to cancel something recently? Almost everybody has, at some point. You signed up in about eleven seconds, on a phone, half watching something else. Cancelling took a different shape entirely. There was a settings page that did not have the option. There was a help centre article that sent you back to the settings page. There was a chat window with a bot that offered you two months at half price, then offered it again. Somewhere in there you were told to call a number that is open between ten and five on weekdays, which is when you are at work.

And at some point you thought, they built this on purpose. They did. Somebody sat in a meeting and decided how hard leaving should be. That decision has a name, and understanding it changes how you look at your own business.

0101Nudge

Start with something smaller. Think about the last form you filled in where a box was already ticked. You probably left it ticked. Most people do, most of the time, and the reason is not laziness. Richard Thaler and Cass Sunstein published Nudge in 2008 and gave the world a phrase for it. A nudge is a change to the way choices are presented that predictably shifts behaviour without removing any options and without changing the economics. The default is the purest example.¹1 Thaler & Sunstein · 2008

The study everyone cites is Eric Johnson and Daniel Goldstein’s, published in Science in 2003 under the title ‘Do Defaults Save Lives?’. Germany asks you to opt in to organ donation, and when the researchers looked, around 12 percent of Germans were registered. Austria presumes consent unless you object, and effective consent there ran at almost 100 percent. Same language, shared border. One line on a form.²2 Johnson & Goldstein · 2003

The default moves the paperwork enormously and the outcome by considerably less. Hold on to that gap. It comes back later.

0202Sludge

Sludge is friction. Specifically, in Sunstein’s framing, it is excessive or unjustified friction: the burden placed between a person and something they have already decided they want, which a reasonable observer would say does not earn its cost.³3 Sunstein · 2021, 2022

The costs come in three parts, in a taxonomy Sunstein takes from Donald Moynihan, Pamela Herd and Hope Harvey, who set it out in the Journal of Public Administration Research and Theory in 2015. There are learning costs, the effort of working out what you are even supposed to do. There are compliance costs, the actual time and documents and steps. And there are psychological costs: the stigma, the confusion, the low grade humiliation of being made to prove something obvious about yourself to a system that will not listen.⁴4 Moynihan, Herd & Harvey · 2015

The third is the one businesses never count. You know it even if you have never had a name for it. It is the specific fatigue of a KYC re-verification for an account you have held for nine years. It is a refund process that requires a photograph of a box you have already returned. It is being asked, again, for a document you uploaded in March.

0303Where the standard version goes wrong

Nudge and sludge get taught as a moral pair. Nudge is the good one, sludge is the bad one. Behavioural science conferences run on that framing and it is wrong in both directions.⁵5 Thaler · 2018

Good sludge is real, and badly underrated

Think about the last time a screen asked whether you were sure. Confirming before you delete something permanently. The warning before you reply to four hundred people. Amazon telling you that you already bought this book in 2019, which has saved me twice. Cooling off periods on large financial decisions. The pause between an impulse and an irreversible outcome. Every one of those is friction, deliberately added, and every one of them is on your side. Sunstein makes this point himself and it gets lost, because friction that protects people is less fun to write about than friction that traps them.

Bad nudge is real too, and everywhere

Any default that serves the seller rather than the person is still a nudge. Pre-ticked consent. Enrolment in a renewal you did not ask for. The largest option positioned as the normal one. Nothing about the nudge mechanism makes it virtuous.

The mechanism is neutral. What you attach it to is not.

So the two axes are separate. One question is whether you are adding friction or removing it. A completely different question is whose interest that friction serves. Put both on a pair of axes and you stop having a moral ranking and start having four boxes. (Figure 1)

0404The asymmetry

Look at almost any subscription business and map it. The way in has every possible friction removed. One tap. Face recognition instead of a password. Payment details already stored. No confirmation step, because a confirmation step is friction and friction on the way in costs conversions. The way out has friction added at every stage. Extra screens. Retention offers you did not ask for. A phone number. A window of hours that does not overlap with your working day. (Figure 2)

Both decisions were made by the same team, in the same building, using the same skill. That is what makes the asymmetry a confession rather than an accident. It takes real design effort to make leaving hard, and somebody signed off on the effort. Arunesh Mathur and colleagues, working mainly out of Princeton, crawled around 53,000 product pages across 11,000 shopping websites and catalogued 1,818 instances of what they classified as dark patterns, across 15 types.⁶6 Mathur et al. · 2019

0505What the regulator did and did not do

In October 2024 the US Federal Trade Commission finalised a revised Negative Option Rule, widely called click to cancel, whose central requirement was simple. Cancelling has to be at least as easy as signing up.

On the 8th of July 2025, six days before the compliance deadline, the Eighth Circuit vacated it in Custom Communications v. FTC. Not on the merits of whether trapping people is acceptable. On procedure, because the FTC had skipped a required preliminary regulatory analysis and so had not given anyone a proper chance to comment on the alternatives.⁷7 8th Cir. · 8 Jul 2025

The story did not end there, which is where most of what you will read online is out of date. The FTC did not appeal. It started again. An advance notice of proposed rulemaking went for review in January 2026, was announced in March, and the comment period closed on the 13th of April 2026. No replacement rule has been proposed yet. In the meantime, enforcement continues under ROSCA and Section 5 of the FTC Act, and state automatic renewal laws still apply, several of them going well beyond the federal baseline.⁸8 FTC ANPRM · Mar–Apr 2026

0606Why companies actually do this

It is tempting to picture a room of people deciding to trap you. That is almost never what happens. What happens is a quarterly review where churn is up two points and somebody is asked what they intend to do about it.

Here is the part that matters. Churn is measured every week, and it has an owner and a target attached to it. The resentment of the person who eventually got out is measured nowhere and belongs to nobody. So when a product manager adds a retention offer to the cancellation flow and churn drops eleven percent, that is a promotion. The customer who left anyway, and who spends the next four years telling people not to sign up, never appears in the same spreadsheet and usually never appears in any spreadsheet at all.

Nobody builds a maze either. Somebody adds one screen because a test showed it saved cancellations. Somebody else moves the button because the page was cluttered. A third person routes the final step to phone support because the retention team closes better on calls than in chat. Each of those is defensible on its own, and the maze is what you get when you put four defensible decisions in a row across three years. That is why the people inside the company are often genuinely surprised when you describe the experience back to them. They have never seen it end to end, because no single person built it.

Friction pays this quarter and costs over the following three years. Quarters are what get reviewed.

Any organisation that reports quarterly will drift toward sludge on its own, without one person intending it, because the gain lands inside the reporting window and the damage lands outside it.

0707Nudging is weaker than the industry admits

In early 2022, Stephanie Mertens, Mario Herberz, Ulf Hahnel and Tobias Brosch published a meta-analysis in PNAS concluding that choice architecture is an effective and widely applicable behaviour change tool. It was the headline the industry wanted. PNAS issued a correction to it that May, after it emerged the dataset had included a retracted study along with coding errors.⁹9 Mertens et al. · 2022

Later the same year, in the same journal, Maximilian Maier, František Bartoš, Tom Stanley, David Shanks, Adam Harris and Eric-Jan Wagenmakers published a letter reanalysing the corrected data. Their title was ‘No evidence for nudging after adjusting for publication bias.’ They found strong evidence of publication bias across nearly every subdomain, food being the one exception, and once they adjusted for it the effect largely disappeared.¹⁰10 Maier et al. · 2022

The exchange is genuinely contested, and Mertens and colleagues replied in turn. A fair objection is that the original analysis pooled wildly heterogeneous studies, which makes any single average effect close to meaningless in either direction. So the honest summary is not that nudging does not work. It is that the confident, universal version of nudging the industry sells is much shakier than the industry admits.

0808The arbitrage

Which leaves a gap sitting in plain sight, and it is the most useful thing in this piece if you run a business.

Nudging says: I can change what you choose by rearranging how the choice is presented. Sludge removal says: there is a barrier between a person and a thing they have already decided they want, and if I take the barrier away, more of them will get there.

One of those is contested across a decade of meta-analyses. The other is close to arithmetic. If four hundred people a month start a process and a hundred and ten finish it, the two hundred and ninety who stopped are not a persuasion problem. They already wanted it. Something got in the way, and you can go and find out what.

0909Friction is not distributed evenly

Everything above treats sludge as an inconvenience. For a lot of people it is not an inconvenience, it is a wall. Anandi Mani, Sendhil Mullainathan, Eldar Shafir and Jiaying Zhao published a study in Science in 2013 called ‘Poverty Impedes Cognitive Function’. Part of it was run with sugarcane farmers in Tamil Nadu, who receive most of their annual income in one lump after harvest. The researchers tested the same farmers twice: once before harvest, when money was tight, and once after, when it was not.¹¹11 Mani et al. · 2013

Same people, same tests. Performance before harvest was substantially worse. For a sense of scale, the authors describe the effect in their laboratory arm, run with shoppers in a New Jersey mall, as comparable to losing a full night of sleep, and they put the effect measured out in the fields at about three quarters of that.

Their explanation is that financial strain is not only a shortage of money. It occupies mental bandwidth. Worry runs in the background and consumes the capacity you would otherwise use for everything else.

The same form is not the same form for everybody.

A twenty minute process costs a lawyer twenty minutes. It costs someone already running at capacity a great deal more, and sometimes it costs them the outcome entirely, because they abandon it.

1010The obligation is daily

Which is the uncomfortable part, because it means good intentions are not doing the work you think they are doing.

If the drift is structural rather than malicious, then a company full of decent people, none of whom would ever design a trap, will still produce one. Nobody has to agree to anything. The incentives point one way, the quarters keep arriving, and three years later the exit is a maze. Goodness that lives in people’s characters cannot catch that, because character is not what is making the decisions. The measurement system is. So the obligation has to live somewhere the measurement system can see it, which means it has to live in the calendar.

That looks unglamorous, which is rather the point. Somebody owns the exit journey the way somebody owns the signup journey. Time to cancel sits on the same dashboard as time to convert, and gets read out in the same meeting by the same person who has to explain the conversion number. When a test shows a screen saved eleven percent of cancellations, somebody is expected to ask what it saved them from, and whether the answer is a problem you could fix or simply a person who wanted to leave.

None of that asks anyone to become a better person. It asks the thing a good person would do anyway to be the thing the system requests, every week, rather than the thing you write on a values page once a year and hope holds.

1111So what about us

1&O builds funnels. A funnel is choice architecture. If I am going to write about people engineering friction, I should say where I sit.

The distinction I would defend is the one from the middle of this piece, applied honestly to my own work. Friction that protects the person is good design. Friction that protects the number is sludge, whatever we call it in the deck.

And I am not going to pretend the line is always obvious from the inside. A confirmation step before a large purchase protects the customer. A retention offer on a cancellation flow can be genuinely useful, if someone is leaving over a problem you can actually fix. The same screen becomes sludge the moment it is there to exhaust rather than to help, and the honest test is whether you would be comfortable explaining its purpose to the person looking at it.

If you would not say it out loud to your customer, you already know what it is.

1212What to actually do on Monday

Sunstein’s practical proposal is a sludge audit, and in 2024 the OECD published a set of good practice principles for running one, drawn from case studies of governments that had already tried. You do not need a government to do this.¹²12 OECD · 2024

Take your own most important journeys and time them. Not estimate, time them, with a stopwatch, on a phone, on a normal connection, as a person who does not work at your company. Count the steps. Count the fields. Count how many of those fields you already have the answer to.

Then do the one almost nobody does. Run the exit. Cancel your own subscription. Request your own refund. Ask your own support team a question from an unfamiliar email address and see how long it takes.

The gap between those two numbers, the entrance and the exit, is the most honest metric in your business. It is not on any dashboard, and it will tell you more about how your customers actually experience you than a quarter of survey data.

Most companies will spend more on their next campaign than it would cost to fix what that audit finds. The campaign has to win attention from strangers. The audit is working on people who already decided they wanted you.

You do not have to guess what a company thinks of its customers. Try to leave one.

1313Sources

1. Thaler, R.H. & Sunstein, C.R. (2008). Nudge: Improving Decisions About Health, Wealth, and Happiness. Yale University Press. ↩
2. Johnson, E.J. & Goldstein, D.G. (2003). Do Defaults Save Lives? Science, 302(5649), 1338–1339. ↩
3. Sunstein, C.R. (2022). Sludge Audits. Behavioural Public Policy, 6(4), 654–673; and Sludge: What Stops Us from Getting Things Done and What to Do about It (MIT Press, 2021). ↩
4. Moynihan, D., Herd, P. & Harvey, H. (2015). Administrative Burden: Learning, Psychological, and Compliance Costs in Citizen-State Interactions. Journal of Public Administration Research and Theory, 25(1), 43–69. ↩
5. Thaler, R.H. (2018). Nudge, not sludge. Science, 361(6401), 431. ↩
6. Mathur, A., Acar, G., Friedman, M.J., Lucherini, E., Mayer, J., Chetty, M. & Narayanan, A. (2019). Dark Patterns at Scale: Findings from a Crawl of 11K Shopping Websites. Proceedings of the ACM on Human-Computer Interaction, CSCW. ↩
7. Custom Communications, Inc. v. FTC, No. 24-3137 (8th Cir., 8 July 2025), vacating the FTC Negative Option Rule on procedural grounds. ↩
8. US Federal Trade Commission, Advance Notice of Proposed Rulemaking on negative option marketing, announced 11 March 2026, published 13 March 2026, comment period closed 13 April 2026. ↩
9. Mertens, S., Herberz, M., Hahnel, U.J.J. & Brosch, T. (2022). The effectiveness of nudging: A meta-analysis of choice architecture interventions across behavioral domains. PNAS, 119(1). See also the PNAS Correction, May 2022. ↩
10. Maier, M., Bartoš, F., Stanley, T.D., Shanks, D.R., Harris, A.J.L. & Wagenmakers, E.-J. (2022). No evidence for nudging after adjusting for publication bias. PNAS, 119(31). ↩
11. Mani, A., Mullainathan, S., Shafir, E. & Zhao, J. (2013). Poverty Impedes Cognitive Function. Science, 341(6149), 976–980. ↩
12. OECD (2024). Fixing frictions: ‘sludge audits’ around the world. OECD Public Governance Policy Papers No. 48. ↩