Twenty years of digital advertising have produced thousands of campaigns that underperformed and precisely a handful that nobody can stop citing. The difference is not the size of the budget or the loudness of the backlash. The campaigns that stuck around are the ones that broke in a way the industry had not yet named.
Key takeaways
- Pepsi pulled its Kendall Jenner spot the day after its 4 April 2017 release.
- McDonald’s promoted #McDStories for under two hours in January 2012.
- Chevrolet’s March 2006 Tahoe contest invited edits it could not moderate.
- Google Glass sold a 2013 prototype narrative it could not deliver by 2015.
Four cases, four distinct failure modes
Grouping these campaigns as generic “fails” hides what makes them useful. Read as failure modes, they stop being anecdotes and start being a checklist.
| Campaign | Date | What actually broke |
|---|---|---|
| Pepsi, Live For Now Moments Anthem | April 2017 | Borrowed standing in a movement |
| McDonald’s, #McDStories | January 2012 | Open prompt, no moderation path |
| Chevrolet, Tahoe ad builder | March 2006 | Publishing tool handed to critics |
| Google Glass | 2013 to 2015 | Prototype marketed as product |
Borrowing standing you do not have
Pepsi released Live For Now Moments Anthem, the two-minute film with Kendall Jenner, on 4 April 2017. It was pulled the next day, 5 April, with a statement that the brand had been “trying to project a global message of unity, peace and understanding” and had “missed the mark”.
The creative problem was never the casting or the edit. It was the claim underneath: that a soft drink could resolve a confrontation whose real-world stakes belonged to people who were not in the room when the ad was greenlit. Audiences read that as a brand renting credibility it had not earned, and no amount of production value covers that.
What makes the case durable is the speed. Less than 48 hours from release to withdrawal, on a spot that had presumably cleared several rounds of internal approval, which tells you the approval process was measuring the wrong thing.
Handing over the microphone without a plan
Two campaigns, six years apart, failed identically. In March 2006, Chevrolet ran a cross-promotion with NBC’s The Apprentice: visitors to ChevyApprentice.com could assemble their own 30-second Tahoe commercial from supplied footage and their own captions, with prizes for the best entries. Created with agency Campbell-Ewald, the mechanic worked exactly as designed. A share of entrants used it to make environmental and fuel-consumption arguments against the vehicle, and those versions travelled furthest.
In January 2012, McDonald’s promoted #McDStories on Twitter hoping for fond memories. Social media director Rick Wion later confirmed the hashtag was promoted for less than two hours; in that window it was used more than 1,600 times, mostly not as intended.
Both campaigns assumed participation would be goodwill. Neither had a plan for the version where it is not, and the tell is identical in both: an open prompt with no defined path for what happens when the responses turn.
Selling the demo, not the product
Google Glass is the odd one out because the marketing was arguably brilliant and the product was the problem. The Explorer Edition began shipping in April 2013 at $1,500 as a developer-facing device, and the surrounding narrative promised a merged digital and physical world. Google ended the Explorer programme on 19 January 2015, and the widely anticipated cheaper consumer edition never arrived at volume, with cost, privacy and limited functionality all cited. Glass then pivoted to enterprise use, where it has been considerably more useful.
The lesson is not that ambitious storytelling is bad. It is that a narrative sets an expectation the product has to meet on the same timeline, and the gap between them is where public scepticism grows. Anyone building creative around a roadmap rather than a shipped feature is running this risk knowingly or otherwise.
What we check before a campaign ships
These four cases have quietly become our pre-launch pass. None of it is exotic; all of it would have caught at least one of them.
- Standing test. Does the brand have a genuine, documentable relationship to the subject it is invoking? If the connection only exists inside the ad, cut it.
- Hostile-input test. For anything user-generated, write the worst plausible submission yourself. If there is no answer to it beyond deleting, the mechanic is not ready.
- Delivery test. Every capability shown must exist in the shipping version on the launch date, not in the roadmap.
- Kill-switch test. Who can pull the campaign, in how long, across every placement including partner channels? Pepsi managed it in a day; that is a process, not luck.
- Room test. Was anyone in the approval chain part of the audience being depicted? If not, that absence is the risk.
The tone-level advice that usually accompanies these case studies, be authentic, know your audience, is not wrong, but it is untestable. The five checks above produce a yes or a no, which is what a launch decision actually needs. The same logic sits behind the psychology behind scroll-stopping ads: attention mechanics are measurable, good intentions are not.
Questions we get asked
Which digital campaigns are most often cited as failures? Pepsi’s Kendall Jenner spot, McDonald’s #McDStories hashtag and Chevrolet’s Tahoe ad builder come up in almost every list, usually alongside Google Glass as a product-marketing example rather than a campaign one.
Is a fast withdrawal a sign of good crisis management? Partly. Pulling within a day limits paid amplification, but it does nothing about organic reach, and the withdrawal itself becomes news. Speed is damage control, not a fix.
Can a brand recover from this kind of failure? Routinely, yes. None of the companies here suffered lasting commercial harm. The reputational cost is measured in becoming a permanent teaching example, which is cheap in revenue terms and expensive in internal credibility.
Does user-generated content still carry this risk in 2026? More than in 2006, because the distribution is faster and the archiving is permanent. The mitigation has not changed: curated submission, human review before amplification, and a pre-agreed response to hostile entries.
The reason these campaigns stay in circulation is not schadenfreude. It is that each one isolated a variable the industry had not yet learned to test for, and testing for it is now standard. That is a reasonable definition of a useful failure.
Working out why your creative is not landing?
Audience mismatch is the most common cause, and the youngest cohort is the least forgiving of it.
Sources: Pepsi statement and press coverage of the withdrawal of Live For Now Moments Anthem, April 2017; L.A. Times statement by Rick Wion, McDonald’s, January 2012; NPR and trade coverage of the ChevyApprentice.com contest, March and April 2006; Google announcement ending the Glass Explorer programme, January 2015. Reviewed August 2026.
