Say · Do · Feel — Consumer Case Study

Your exit survey says
“price.” It was the ick.

When a customer quietly stops coming back, the exit survey records a tidy reason: too expensive, slow service, weak coffee. Give them a blank box and they’ll even own up to a grim bathroom. But the single biggest trigger for actually walking — a crowd they’d rather not be seen in — never reaches a form at all: not one of 233 named it. We gave 235 digital twins twelve small turn-offs and measured three things at once: what they’d say, how it felt, and what they’d actually do. The gap between them is where churn hides.

235

Digital twins

Customer icks tested

Reads each — say·do·feel

9,600+

Reasoned responses

01 — The puzzle

The reason they gave is not the reason they left

Ask a lapsed customer why they stopped coming and they’ll hand you a list of sensible reasons. Most of them are real. The one that actually moves the most feet isn’t on the list — and never will be.

We asked the standard exit question with no forced choices at all — a blank box: what would make you stop going to a coffee chain you used to like? Across 233 twins, the spoken reasons line up exactly where you’d expect, and people are franker than a tick-box survey assumes. They’ll happily name a dirty bathroom; a quarter will raise the company’s ethics unprompted:

What they say would make them quit — unprompted	Share who mention it
Service / staff / rudeness	72%
Price / fees	46%
Cleanliness / hygiene — the dirty bathroom	43%
Coffee quality / taste	30%
Ethics — dishonesty, worker treatment, politics	25%
Identity — “a crowd I’d rather not be associated with”	0%

Open-ended; twins gave more than one reason, so shares sum past 100. “Crowded / busy” came up (19%), but as over-crowding and anxiety — not once as who was in the room.

Zero of 233 named the identity ick — even with a blank box and no list to hide behind. It is the single biggest driver of actual defection in this study (63%), and the one reason nobody will say out loud.

This is the trap, and the fair test sharpens it rather than softening it. People will admit the embarrassing-but-acceptable icks — a filthy bathroom, a greedy CEO. What they won’t admit is the socially awkward one: “I left because I didn’t want to be seen with the people in there.” That sentence makes the speaker sound like a snob, so it gets laundered into “it got too expensive” or “service slipped” before it ever reaches your dashboard. So we stopped asking twins to explain their behaviour and started measuring it instead.

02 — Method

Three reads on the same moment: say · do · feel

We built a panel of 235 digital twins — synthetic respondents modelling a real US population — and walked each one through twelve everyday coffee-shop icks: a grim bathroom, an aggressive upsell, staff on their phones, shrinkflation, a barista’s eye-roll, a forced app download, a cringe CEO, greenwashing, a guilt-trip donation screen, a place that’s suddenly too popular, and a crowd you’d rather not be associated with. Every twin saw every ick, so we can follow the same person across all three reads.

Say

The exit-survey answer — asked as an open, no-forced-choices question: the reasons they’d give out loud for cutting back.

Feel

The gut reaction. How strong the ick is on a 1–10 visceral scale — the cringe, the disgust, the discomfort.

Do

The behaviour. What they’d actually do, on a six-rung ladder from “back as normal” to “stop and tell everyone.”

The Do ladder is the spine of the study, because it separates the customer who grumbles and stays from the one who quietly disappears:

Rung	What the twin does after the ick
0	Back as normal — no effect
1	Back, but it niggles
2	Goes a bit less, quietly
3	Switches to a rival for a while
4	Quietly stops going — says nothing
5	Stops and tells them / posts about it

We count rung 3 or higher as defection — real lost revenue. And we flag a special, dangerous case: a twin who defects (3+) but doesn’t voice it (anything below 5). That’s the silent quitter — gone, with no complaint, no review, no signal you could ever have caught.

A note on trust. Every twin wrote out its reasoning, so we can read why each answer landed where it did. In this study the exported ratings were unusually clean — the felt-ick number matched the twin’s own words 98% of the time, and the behaviour ladder is a plain categorical choice — so we report the numbers as measured, with the reasoning as the receipts.

03 — The core finding

Feeling the ick barely predicts leaving

If strong feelings drove behaviour, the gut reaction would tell you who walks. It doesn’t. Across all twelve icks, the correlation between how strongly a twin felt the ick and whether they actually left is just r = 0.30.

That weak link is the whole point. Some icks make people recoil and change nothing. Others barely register as a feeling and quietly end the relationship. A survey that measures sentiment — “how did that make you feel?” — is reading the wrong dial. Sort every twin’s feeling against their behaviour and four groups appear:

Quadrant	What it looks like	Share
Fine	Low feeling, stays. The ick washes over them.	64%
All bark	Strong feeling, stays anyway. Cringes — and comes back tomorrow.	16%
Loud & gone	Strong feeling, leaves. The churn a survey would catch.	11%
Silent quit	Little feeling, still leaves. Walks out without a flicker.	9%

A quarter of your at-risk customers are mis-read by sentiment alone: 16% feel it and stay, 9% feel almost nothing and leave. Emotion and exit are two different dials.

The All bark group is the false alarm — they light up every sentiment tracker and cost you nothing. The Silent quit group is the real danger: no spike in feeling, no complaint, just a customer who stops showing up. Behaviour is the only instrument that sees both.

04 — The ick league table

The icks that disgust aren’t the icks that empty the till

All twelve icks, ranked by what they actually do — the share of twins who defect (switch, quietly leave, or leave loudly). Feel is the average gut rating (1–10); Strong feeling is the share rating it 7+; Defect is the share who’d actually cut their visits. Watch how badly feeling and behaviour disagree.

The ick	Feel (1–10)	Strong feeling (7+)	Defect
A crowd you’d rather not be associated with Silent quit	5.8	27%	63%
A dirty, grim bathroom	7.4	99%	61%
Aggressive upsell at the till	5.9	12%	20%
Staff on their phones, ignoring you	5.2	6%	19%
Shrinkflation — smaller cup, same price All bark	6.7	72%	18%
A barista’s eye-roll	5.8	26%	17%
Forced app download to order	5.9	22%	13%
A cringe CEO in the news	4.3	8%	10%
Obvious greenwashing All bark	6.0	37%	9%
An in-app upsell wall	5.7	12%	8%
A checkout guilt-trip donation screen	5.1	2%	2%
The place is suddenly too popular	2.7	0%	5%

Strong feeling (rated 7+) Would actually defect Both bars on the same 0–100 scale — where they disagree is the gap.

Look at the top two rows. The dirty bathroom is the most felt ick by a mile — 99% rate it a strong 7+ — yet it ranks a close second on actual defection. The wrong crowd barely registers as a feeling (5.8, only 27% strong) and still tops the table at 63% defection. Disgust is loud and forgivable; identity is quiet and final.

All feeling, little exit

Shrinkflation outrages everyone (72% feel it strongly) and shifts almost no one (18%). It’s the textbook All bark ick — a great way to fail a sentiment tracker and keep every customer.

Little feeling, real exit

The wrong crowd is the textbook Silent quit ick — modest disgust, maximum churn. People don’t rage about it; they just don’t come back, and they don’t tell you why.

05 — The quiet quitters

Quiet quitting, for your brand

The phrase belongs to burnt-out employees who do the minimum and say nothing. Your customers do the consumer version: they stop coming, and they never file a complaint.

When we read the reasoning of the twins who defect, the loudest theme is silence. Among everyone who’d walk over the wrong crowd (n=147), 81% say they’d simply go elsewhere — and 35% explicitly say they wouldn’t voice it at all. For the dirty bathroom defectors (n=144), it’s starker still: 93% are driven by disgust, 94% would switch — and roughly half would say nothing on the way out. No review. No survey. No warning.

“I’d just switch to somewhere else for a while. I don’t want to deal with dirty tables and a grim bathroom, so I’d go somewhere cleaner and not say anything.”Silent quitter — dirty bathroom

“I’d just go somewhere else for a while, quietly. I don’t want people thinking I’m part of that group.”Silent quitter — the wrong crowd

“I’m not gonna make a post or tell them, I’d just kind of drift away.”Silent quitter — the wrong crowd

This is why the exit survey is structurally blind. The customers who would have told you something are the rare Loud & gone 11% — the ones who post the one-star review. The vast majority drift, switch and stay polite. Your most accurate churn signal is the one your customers have quietly agreed never to send you.

Across the worst single ick, more than a third of leavers would never say a word. They don’t complain. They just stop being your customer.

06 — The breaking point

Icks don’t add up. They stack — then they detonate.

Real customers don’t meet one ick in a vacuum; they accumulate. So we stacked them, one layer at a time, on the same 235 twins, and watched defection climb. The curve isn’t a smooth slope — it has cliffs.

Stacked experience	All — defection	Millennial	Gen X	Boomer
L1 — greenwashing alone	9%	3%	11%	15%
L2 — + shrinkflation +26 · first tip	35%	31%	37%	42%
L3 — + barista eye-roll +8 · shrugged off	43%	44%	41%	47%
L4 — + dirty bathroom +23 · hygiene cliff	66%	70%	59%	70%
L5 — + the wrong crowd +23 · identity cliff	89%	92%	87%	90%

Two cliffs and one dud. The first tip comes at L2: a single annoyance (greenwashing) is shrugged off at 9%, but add shrinkflation and a third of customers are gone. The eye-roll at L3 adds almost nothing (+8) — interpersonal friction is forgivable on its own. Then the two big cliffs: the hygiene cliff at L4 (+23, disgust is a physical limit) and the identity cliff at L5 (+23, who’s-in-the-room is the final straw). By the full stack, nine in ten have defected.

Some pairings detonate. Some don’t.

Two icks together are rarely the sum of their parts. We compared each pairing’s observed defection against what you’d predict if the two acted independently:

Combination	Predicted (if independent)	Observed	Verdict
Dirty bathroom × wrong crowd	86%	95%	Detonates +9
Shrinkflation × greenwashing	26%	35%	Detonates +9
Dirty bathroom × wrong crowd × eye-roll	—	98%	Ceiling
Wrong crowd × eye-roll	—	62%	Identity dominates
Dirty bathroom × shrinkflation	68%	67%	As predicted

Predicted if the two icks acted independently Observed together Observed bar past the predicted bar = detonation.

A grim bathroom next to a crowd you distrust isn’t additive — it’s a 95% wipe-out, well above the 86% you’d predict. “This place isn’t for people like me” and “this place doesn’t care” compound into a verdict about the brand itself. Meanwhile some pairings are sub-additive — once identity has decided you’re leaving, an extra eye-roll changes nothing. You can only lose a customer once.

07 — Who breaks, and on what

Different people, different breaking points — except one

The stack hides as much as it shows, because not everyone breaks on the same rung. Splitting the panel reveals which ick belongs to which audience:

Greenwashing skews old

Boomers defect at 15% over greenwashing; Millennials at just 3%. The younger cohort assumes corporate spin and barely blinks; older customers take the broken promise personally.

Hygiene hits the old hardest

A dirty bathroom pushes 75% of Boomers to defect — versus just 56% of Millennials, the most forgiving cohort. Disgust standards rise with age: older customers walk over a grim bathroom; the young grumble and stay.

Men forgive early, snap late

Men are more forgiving through the early layers, then break hardest of all at the identity cliff — 95% defection at L5 vs 88% for women. Slow to anger, absolute once they go.

The one ick that ignores all of this: the wrong crowd defects 62–65% across every generation. Identity discomfort is universal — not a young thing, not an old thing, a human thing.

We expected the identity ick to be generational — a Gen Z / Millennial sensitivity about brand tribes. It isn’t. Boomers, Gen X and Millennials walk away from “the wrong crowd” at almost identical rates. That’s a finding we didn’t predict, and it’s the kind a topline average would have buried: the strongest churn driver in the study is the one demographic cut can’t explain.

08 — Why twins, not a survey

The reasoning is the product

A survey gives you a number and a guess at the reason. The twins give you the number and the reason, person by person, with the contradiction left intact. That’s what let us catch the silent quitters at all — a customer who rates the ick a calm 5 and then explains, in their own words, that they’re quietly done.

Mining the reasoning surfaces the mechanics a topline can’t. Defectors over the wrong crowd talk about image and belonging (31%) and value for money (32%) in the same breath — “if this isn’t my place, why am I paying a premium for it?” Greenwashing defectors are driven by broken trust (48%), not the environment. And the customers who stay are held by one thing above all: habit and convenience — cited by 32% of stayers versus 20% of leavers. Loyalty here isn’t love; it’s inertia, and inertia is exactly what an ick overrides.

A survey gives you the polished self-report — the version a customer is willing to put on a form. Brox works the other way: it reads the individual, reasoning and all, and keeps the contradictions a topline would average away. The ick only shows up if you can see the gap between what someone feels, what they say, and what they do. That gap is the insight — and it lives at the level of the single customer, not the aggregate.

09 — What it means

Stop measuring satisfaction. Measure defection.

Feeling and leaving are different dials. If your retention programme runs on sentiment, you are chasing the 16% who feel it and stay, and missing the 9% who feel nothing and go.

For anyone who owns churn — a retail chain, a subscription, an app, a hospitality brand — the lesson generalises past coffee. Most of the reasons customers give on the way out are honest; the single most decisive one is the one they’ll never say, because saying it would make them look bad. And the triggers compound: a tolerable annoyance plus a hygiene failure plus an identity threat is a 90% wipe-out, not three small dings. You cannot manage what your customers have quietly agreed not to tell you — and you only see it by measuring what they do, not what they report.

Model behaviour, not mood

The gut reaction and the exit decision correlate at just 0.30. Measure what people do, or you’ll optimise the feeling and lose the customer.

Hunt the silent quitters

A third of your leavers will never tell you. Find the low-feeling, high-defection icks — they’re the ones bleeding you in silence.

Read the reasoning

The why behind a defection is where the fix lives. Identity, trust and disgust need three different responses — the number alone can’t tell them apart.

Method & caveats

How to read these numbers

The panel. 235 digital twins — synthetic respondents modelling a US population (skewing female, with Millennials, Gen X and Boomers as the analysable cohorts; Gen Z and the Silent generation were too small to cut). Each twin answered every ick across say, feel and do, with written reasoning.

Within-subject. The same twins ran every scenario, paired on person ID — so the say–feel–do gaps are measured inside each person, not across different samples.

The Say arm is open-ended. An early version offered a forced four-way choice (service / quality / price / other), which artificially suppressed every reason not on the list. We re-ran it as a blank-box question (233 twins) so any reason could surface freely. Hygiene and ethics duly appeared; identity still did not — that null is from the fair test, not the constrained one.

One stimulus needs pinning. “A crowd you’d rather not be associated with” was read differently by different twins — some political, some a general “unsafe vibe.” The 63% defection is robust; the reason spans a range we report through the reasoning rather than over-claiming a single trigger. A pinned re-run is the obvious next step.

US-localised wording. Every hygiene scenario was re-run with “bathroom” in place of “toilet” — the natural US term — to keep the stimulus idiomatic for the panel. The ranking held: a dirty bathroom remains the most-felt ick (99% strong) and the No.2 defection driver (61%), just behind the wrong crowd.

Directional, not decimal. At n=235, lean on the large gaps — the r=0.30, the league-table ordering, the L4/L5 cliffs — not on single points. The benchmark here is behavioural realism, not a public poll.

Your exit survey says“price.” It was the ick.

The reason they gave is not the reason they left

Three reads on the same moment: say · do · feel

Say

Feel

Do

Feeling the ick barely predicts leaving

The icks that disgust aren’t the icks that empty the till

All feeling, little exit

Little feeling, real exit

Quiet quitting, for your brand

Icks don’t add up. They stack — then they detonate.

Some pairings detonate. Some don’t.

Different people, different breaking points — except one

Greenwashing skews old

Hygiene hits the old hardest

Men forgive early, snap late

The reasoning is the product

Stop measuring satisfaction. Measure defection.

Model behaviour, not mood

Hunt the silent quitters

Read the reasoning

How to read these numbers

The reason they gave isn’tthe reason they left.

Your exit survey says
“price.” It was the ick.

The reason they gave isn’t
the reason they left.