Why the one metric that could embarrass you next quarter is the cheapest credibility on the market, and the one thing on your website a competitor cannot copy by Friday.

Everyone in your category claims to be fast, accurate and easy to work with. Buyers read those words as noise. What follows is the alternative, with the receipts linked at the end.

1. What does “publish a number you could lose” actually mean?

Putting a measurement on your website that has a real chance of moving against you, in public, on a schedule you announced before you knew what the next reading would say.

Not a testimonial. Not “trusted by 400 brands.” Not a case study where the client is anonymous and the result is called significant. A figure, with a denominator, a definition, and a next-update date.

The test takes four seconds. Could this be worse in ninety days, and would anyone find out? If either answer is no, you published decoration, and the market prices it accordingly.

2. Domino’s told the world its pizza tasted like cardboard. What did that pay?

More than any conventional campaign of that decade, and it is still paying. By 2009 the company had sold roughly the same pizza for fifty years, and focus groups returned verdicts marketing could not spin: crust like cardboard, sauce like ketchup. Domino’s put that research on camera, ran live criticism on a Times Square marquee, and let the public watch it reformulate.

1. Same-store U.S. sales rose 14.3% in Q1 2010, the largest quarterly same-store gain in quick-service history at the time.
2. Domestic same-store sales finished the full year 2010 up 9.9%, off a flat-to-declining base.
3. The stock climbed 130% from December 2009 to December 2010, from around $8.76 a share.
4. Fourth-quarter profit hit $23.6 million, more than double the prior-year period.
5. Shares ran from under $10 in 2008 to over $400 by 2017, roughly a 6,000% decade.

A recipe is copyable in a week. What competitors could not copy was a company president reading his own worst reviews out loud on television, before there was proof the fix would sell. Fifteen years later it is taught at Harvard Business School, which tells you how rarely anyone repeats it.

3. Why does a claim that cannot fail get discounted, and is there research behind that?

Buyers price in the odds you are lying, and history gave them reasons. Edelman found 66% of consumers believe brands overstate their environmental commitments; Terrachoice found over 95% of green products made at least one misleading claim. Business is still the most trusted institution at 62%, but that trust is conditional.

The academic name for the fix is the blemishing effect. Ein-Gar, Shiv and Tormala published four studies in the Journal of Consumer Research in 2012 showing a small dose of negative information, added to an otherwise positive description, makes people more favorably disposed. Three conditions decide whether it works:

1. The negative must be minor. A real flaw, not a catastrophe.
2. It must come after the positive information. Reverse the order and the effect reverses.
3. It works under low processing effort, which describes nearly every real buying moment.

Commerce data agrees. Reevoo: 68% of consumers trust reviews more when good and bad are both present, and 30% suspect censorship when they see no bad ones. PowerReviews: 46% of shoppers are suspicious of a perfect five-star average, 53% among Gen Z; 96% seek out negative reviews at least sometimes, up from 85% in 2018; and shoppers who click into the one-star breakdown convert 108.8% higher. Purchase likelihood peaks between 4.0 and 4.7 stars, then falls toward 5.0.

4. If I refuse to publish the number, does the problem stay private?

No. The best cautionary tale in American business is a soft-serve machine.

McDonald’s has never published an uptime figure for its ice cream machines. In 2020 a developer named Rashiq built McBroken, which places a real online sundae order at every location every twenty to thirty minutes. Accepted, the dot turns green. Rejected, red. That is the whole methodology, and it is more transparent than anything the company has offered.

1. The national broken rate has run from roughly 8% on good days to over 20% in New York on bad ones.
2. When the FTC started asking questions in 2021, McBroken showed about 11% down nationally, over a quarter out in New York.
3. PIRG reported in late 2025 that twelve to fifteen percent are still regularly out of service.
4. Roughly 60% of the McDonald’s dessert menu depends on that one machine.

Then it got worse. In September 2024 Wendy’s licensed McBroken’s live data, named Chicago, Houston, Las Vegas, New York and Los Angeles as the cities suffering most, and ran a national dollar Frosty promotion aimed at people who had just been told no. Wendy’s even publicized the operational detail: machines go down most between 11 a.m. and 3 p.m., averaging two to three hours down.

A competitor built a national campaign out of a metric McDonald’s declined to measure in public. Meanwhile the American Customer Satisfaction Index publishes a McDonald’s score every year regardless: last among limited-service restaurants in 2025 at 70 points, versus 71 in 2024 and 69 in 2023, while Chick-fil-A led for the eleventh straight year at 83. The number you will not publish gets published anyway, by someone with an interest in the framing.

5. Has McDonald’s ever done this right?

Once, brilliantly. “Our Food, Your Questions” launched in Canada in 2012 through Tribal DDB Toronto: ask anything and we will answer. The questions were not gentle. People asked about pink slime, about lips and eyeballs in the nuggets, about why the sandwich in the photo looks nothing like the one in the bag.

1. Close to 6,000 questions arrived in the first months in Canada, past 19,000 as the program spread.
2. The Finnish version answered more than 1,700 questions in under a year. Food quality perception rose 29%, trust score 35%, and sales grew 14% over the campaign.
3. McDonald’s Canada was named Marketer of the Year in 2013 on the strength of it.
4. The 2014 U.S. version hired Grant Imahara of MythBusters, a professional skeptic, and filmed inside the plants.

Note the asterisk. Time argued the program answered the fun questions and sidestepped the ones about antibiotics and cage-free eggs. Fair criticism, and exactly why the next rule exists.

6. Which number do I pick, and why measure it quietly for ninety days first?

The one your buyer already worries about privately and is too polite to raise on the first call.

1. Measure their risk, not your effort. “Average response time” measures how hard you work. “Issues resolved without a second contact” measures whether their day got ruined.
2. It has to be losable. If your realistic floor is 99%, that is a trophy. Publish something with room to fall.
3. It has to be checkable. A customer should be able to catch you fudging it. That possibility is the whole value.
4. It has to be boring. Boring numbers cannot be reframed by a rival’s marketing department.
5. It has to be the fifth-minute question. Whatever a reference call turns to once pleasantries end, that is your number.

Shortcut: call three prospects you lost and ask what made them nervous. They will tell you, having nothing left to lose. Then say nothing publicly for a quarter, because your first measurement will be wrong, your second embarrassing, your third finally true.

1. Definitions start fights. Two people on your team already disagree about what counts as resolved. Settle that privately.
2. Seasonality makes a good month look like a trend and a bad one like a crisis.
3. You need your floor before you promise anything. The promise is permanent. The floor is not.
4. If the result is a disaster, you have ninety days to fix it before anyone hears. Domino’s spent roughly eighteen months reformulating before saying a word. Silence while you fix something is fine. Silence while you sell around it is not.

7. Why does one blended average ruin the whole exercise?

Averages are where inconvenient truth hides. Industry-wide, average drive-thru time improved from 6 minutes 13 seconds in 2022 to 5 minutes 29 seconds in 2024, and order accuracy sits near 87%, meaning about 11% of orders come out wrong. Fine. Now segment it.

In the 2024 study Chick-fil-A posted the slowest total time of ten brands, 479.42 seconds against 375.67 for McDonald’s. The blended figure says Chick-fil-A is slow. But Chick-fil-A had 3.44 cars in line when the shopper pulled up, against 1.48 at McDonald’s. Adjust for that and Chick-fil-A cleared a vehicle every 139.4 seconds. The average told the opposite of the truth.

More from the same data: accuracy improves 26 percentage points when the order confirmation board displays correctly, total time drops 27 seconds when the customer need not repeat the order, and accuracy is worst at dinner. One blended score buries all three. Cut yours three ways, by time period, customer size and product line, and publish the ugliest cut beside the pretty one.

8. Do I really have to publish the misses next to the hits?

Yes, and this is where most companies quietly abandon the project. Reevoo found 95% of customers grow suspicious of a rating with no negative reviews attached. Over half of shoppers, 52% overall and 60% of millennials, click straight to the one-star reviews. They are hunting for proof that any of it is real.

Five columns, no editorializing: target, result, what happened, what you changed, the date you changed it. A miss with a fix and a date reads as an operating company. A miss with a paragraph of explanation reads as a defendant.

9. Where does the number need to live so it actually gets found?

Most companies are polishing the wrong surface. Claude Penland’s analysis at 1000startups.com, in “You Lost the RFP Before Anybody Wrote It,” traces how requirements documents now get assembled: 94% of buyers used an AI assistant somewhere in the purchase, 94% ranked their shortlist before contacting any seller, and 95% bought from a vendor on the Day One list. When the shortlist was pre-ranked, the first vendor called won roughly 80% of the time. When it was not, that fell to 57%. Talking first does not make you the favorite. Being the favorite makes you the first call.

The retrieval mechanics reward exactly this:

1. Pages carrying 19 or more discrete statistics averaged 5.4 citations, against 2.8 for thin ones.
2. Sections of 120 to 180 words between headings averaged 4.6 citations, against 2.7 for longer blocks.
3. “Best X” listicles account for 43.8% of page types ChatGPT cites, per an Ahrefs study of 26,283 source URLs.
4. Branded-query citations go 57% to reviews, listicles and forums, only 12% to product pages. Your website is a minority shareholder in your own reputation.
5. Domains absent from G2, Capterra and Trustpilot average 1.8 citations. Listed domains, 4.6 to 6.3.

Refresh on a calendar, because citations rot. Median half-life across 3.5 million citation events is 4.5 weeks, ChatGPT churns fastest at 3.4, and a URL cited today has roughly a 15% chance of surviving three weeks. The buying cycle averages 10.1 months, so the window opens eleven months out.

10. Can a published number backfire?

Absolutely, and the sharpest warning comes from cardiac surgery. New York State’s Cardiac Surgery Reporting System produced the first physician-specific mortality report ever published. The wins were real: risk-adjusted mortality fell, hospitals performing worse than expected dropped from five to one between 1989 and 1992, and cases handled by very low-volume surgeons fell 25%, mortality in that group down 60%.

Now the bill. Burack found 67% of New York surgeons had refused at least one patient that year, 18% five or more. In Pennsylvania, Schneider and Epstein found 63% of surgeons less willing to operate on severely ill patients afterward, and 59% of cardiologists reported new difficulty finding anyone who would. Narins found 83% of New York interventional cardiologists agreed that publishing scores cut the odds a patient who needed a procedure got one. A later survey found 65% had avoided a medically indicated procedure at least twice, and 40% named gaming the risk factors as the workaround.

Transparency did not fail there. What happened is that a published number becomes an instruction, including to you. Five guardrails:

1. Publish the denominator, not just the rate. A rate alone is a rumor with a decimal point.
2. Publish your exclusions and the case count they removed. Exclusions are where bodies go.
3. Write the definition before you measure, and put it beside the number.
4. Let someone outside the measured team own the count.
5. Publish a paired metric that moves the wrong way if you game the first. Speed alone invites accuracy collapse, the whole story of the drive-thru lane.

11. Why can’t a competitor copy this, and what can I ship in thirty days?

They would have to start their clock today; yours started ninety days ago. Four quarters of readings with a bad quarter in the middle cannot be manufactured retroactively. If their number is worse, publishing costs them deals immediately, so the move is cheap only for whoever is good. Compare the alternative you already run: 166 RFPs a year at roughly 33 hours each, about 5,478 hours annually, at a 39% win rate, with only 13% of losses blamed on the proposal and 55% to 61% on price and criteria set before you arrived.

The thirty-day version:

1. Week one: pick the metric. Write its definition in one sentence, then hand it to your most argumentative colleague to attack.
2. Weeks one to four: instrument it. Backfill ninety days if the data exists somewhere, which it usually does, in a spreadsheet, badly.
3. Week four: build one page. Number, denominator, definition, method, date measured, next update date. Segment three ways, worst one included.
4. Week four: add one miss, with the fix and the date of the fix.
5. Then treat the update date like payroll. Not a marketing deadline, a promise with your name on it.

The whole thing in one paragraph

Anyone can claim expertise, and everyone does, which is why claims stopped working. A metric that cannot go against you is decoration and gets priced like decoration. A metric that can, published on a schedule you set in advance, is a standing bet that you are as good as you say. Domino’s did not win on pizza. It won on saying the number first. The value was never the number. It was the risk that it moves, and that you posted it anyway.

Sources

Claude Penland, “You Lost the RFP Before Anybody Wrote It,” 1000 Startups (retrieval mechanics, 19+ statistics per page, citation half-life, 6sense and Loopio buyer data): https://1000startups.com/2026/09/04/you-lost-the-rfp-before-anybody-wrote-it/

1000 Startups, Proof & Original Research archive: https://1000startups.com/category/proof-original-research/

Ein-Gar, Shiv & Tormala, Journal of Consumer Research 38(5), 2012, the blemishing effect: https://academic.oup.com/jcr/article-abstract/38/5/846/1796852

Stanford GSB summary of that research: https://gsb.stanford.edu/insights/positive-effect-negative-information

Aaron Allen & Associates on the Domino’s turnaround: https://aaronallen.com/blog/dominos-turnaround

UCLA case study, Domino’s (14.3%, 130%, share prices, $23.6M): http://www.econ.ucla.edu/sboard/teaching/tech/dominos.pdf

McDonald’s Canada, “Our Food. Your Questions.” release: https://news.mcdonalds.ca/press-releases/mcdonaldsr-canada-turns-volume-our-food-your-questions

Nord DDB, Finnish results (+29%, +35%, +14%): https://nordddb.com/case/our-food-your-questions/

TIME on the U.S. launch and its critics: https://time.com/3501921/mcdonalds-transparency-campaign/

PIRG on the ice cream machines (twelve to fifteen percent): https://pirg.org/articles/mcdonalds-ice-cream-machines-remain-broken/

NPR on the FTC inquiry: https://www.npr.org/2021/09/02/1033715465/mcdonalds-ice-cream-machines-mcflurries-ftc-investigation

McBroken and the Wendy’s Frosty Fix campaign: https://www.gocanopy.com/news-insights/fixing-mcdonalds-broken-ice-cream-machines

QSR Magazine, 2024 Drive-Thru Report: https://www.qsrmagazine.com/story/the-2024-qsr-drive-thru-report/

Global Payments QSR drive-thru study (87%, +26 points): https://www.globalpayments.com/insights/qsr-drive-thru-study

Restaurant Dive, 2025 ACSI restaurant study: https://www.restaurantdive.com/news/acsi-restaurant-satisfaction-survey-2025-chick-fil-a-mcdonalds/750843/

PowerReviews on rating and conversion (46%, 108.8%): https://www.powerreviews.com/average-rating-impact-on-conversion/

PowerReviews consumer survey (96%, 52%, 60%): https://www.powerreviews.com/survey-importance-product-reviews-drive-purchase-behavior/

Reevoo findings via SiteTuners (68%, 30%): https://sitetuners.com/blog/4-ways-to-leverage-user-reviews/

2026 Edelman Trust Barometer: https://www.edelman.com/trust/2026/trust-barometer

Hiebing on Edelman data (business most trusted at 62%): https://www.hiebing.com/blog/state-of-trust-2026-what-brands-need-to-know

JACC, New York State Cardiac Registries (Burack 67% and 18%, Narins 83%): https://www.jacc.org/doi/10.1016/j.jacc.2011.12.051

Hospital outcomes profiling critique (Schneider & Epstein 63%, 59%): https://arxiv.org/pdf/0710.4622

Annals of Thoracic Surgery, report cards review: https://www.annalsthoracicsurgery.org/article/S0003-4975(01)03222-2/fulltext

ScienceDaily on report cards and risk avoidance (65%, 95%): https://www.sciencedaily.com/releases/2018/05/180510133042.htm