Everyone copied the unhinged playbook. Here is the measured record of who it worked for.

Duolingo’s owl created a genre. Between the end of 2022 and the end of 2024, mentions of the word “unhinged” roughly tripled on social platforms, and mentions of “unhinged marketing” grew twenty-five-fold. That second number is the important one: it measures not a mood but an industry deciding, more or less at once, that this was now a strategy you could buy.

So by now there is a record. Brands copied the playbook with real budgets and real agencies, some for years. This piece is about what that record shows: one clear success with third-party sales data behind it, one campaign where the brand and the analysts genuinely disagree about what happened, a middle tier of partial evidence, and the strange, instructive fact that the inventor itself tried to retire the format and got punished for it.

Nutter Butter: the copy that worked, and how we know

In August 2024, the TikTok account of a peanut butter cookie born in 1969 had about 3,100 followers. Its team began posting content that made Duolingo look restrained: surreal, occasionally menacing videos with recurring characters, no visible product logic, and captions that explained nothing. By 11 September it had 400,000 followers. By 13 October, 1.1 million. By June 2025, 1.6 million.

The dates matter because they come from different, independent outlets: the September and October figures were reported by CNN in an article whose headline (“Nutter Butter, are you OK?”) was itself part of the phenomenon, and the June 2025 figure by Marketing Brew. This is not a growth curve reconstructed from one press release.

Behind the account: three named people. A senior social media manager at Dentsu Creative, a brand manager at Nutter Butter, a digital lead at Mondelēz. Weekly brainstorms, low production budgets, zero paid media on the viral posts. Their own description of the method is more disciplined than the output suggests. The Dentsu manager, on why the videos refuse to explain themselves: “we want to leave room for interpretation because that’s what causes people to ask questions.” A former team member, on the routine: “So what if a post flops? Try again and take notes. Don’t delete a low-performing post just because it didn’t perform how you wanted it.”

Now the part almost no imitator can show. Stackline, a retail intelligence firm with no stake in the campaign, measured a sales bump that began in May 2024 and, in Adweek’s words that autumn, “has yet to go away”. Ad Age put a number on the Amazon slice of it: sales up 190% from January to June. And a comment under one video, quoted by the brand manager, compresses the whole mechanism into one line: “I bought Nutter Butter today. I haven’t bought one in 30 years.”

One caution before anyone copies the copy: Mondelēz’s earnings calls in 2024 and 2025 never mention the account. A campaign can move a cookie’s Amazon sales 190% and still be invisible at the scale of a $36 billion snack company. The unhinged playbook moved a brand, not a corporation.

Dunkin’: the campaign where both sides are right

Halloween 2024. A spider character “takes over” Dunkin’s social accounts for a week, posts increasingly odd content, leaks the seasonal menu, and is publicly “fired” on 1 November. The brand’s award entry claims 2.5 million engagements, 700,000 shares, 100,000 new followers and 2.9 billion earned impressions in about a week, on a production budget under five thousand dollars, and adds that it was the best-selling year ever for the spider donut the character promoted.

Pulsar looked at the same account six months later and reported something else: the engagement spike was real, and the account’s baseline after the campaign was no higher than before it. A week of applause, then the old normal.

Both can be true, and that is the useful lesson, not a contradiction to resolve. A character stunt is a campaign: it spikes attention, may well sell donuts, and ends. Duolingo and Nutter Butter did something categorically different: they changed the account’s permanent voice, which is why their baselines moved. If you borrow the aesthetic for a week, you get a week. Dunkin’ seems to have understood its own results perfectly, incidentally: it brought the spider back the next Halloween, as a seasonal bit, not as an identity.

The middle tier: evidence, but only their own

Scrub Daddy runs its smiling sponge with the same menace-adjacent humor and 4.2 million TikTok followers, and reports that mascot videos earn 128% more impressions than its other content on Instagram, 95% more on TikTok. Pop-Tarts fed its mascot to a toaster bowl-game audience, won a Cannes Grand Prix for it, and says the flavor it pushed became the country’s fourth best-selling. Jack in the Box raps. Mr. Peanut lurks. All of this is brand-reported: real signals, no third party attached.

The precedent that predates all of it is Wendy’s, whose roast persona dates from 2017, four years before the owl. Its 2023 National Roast Day campaign on TikTok reported 154,000 new followers, a 37% lift in ad recall, and a 4.5% increase in store visits, numbers that originate from TikTok’s advertising case studies, so from the platform selling the format, but at least not from the brand grading itself.

The evidence against, including from the inventor

A Sprout Social survey in mid-2025 asked consumers what makes a brand bold on social. Half said honesty. Twenty-three percent said unfiltered or unhinged content. That is the saturation problem in two numbers: a format that worked by being unexpected is now expected, and an audience that has seen every snack account have a breakdown is harder to surprise.

Pulsar’s analysis names Oreo and Reese’s as brands whose unhinged-adjacent posts underperform their ordinary content, published without figures, so worth exactly an analyst’s word. The strongest counter-evidence comes from Duolingo itself, and it cuts both ways. In 2025 and 2026 the company shifted budget from its unhinged voice toward paid creator accounts, and its growth guidance for 2026 halved from the 40-plus percent quarters of the unhinged era. Mark Ritson’s verdict in The Drum was that abandoning a distinctiveness asset this valuable deserved its own category of error: “Duolingo stupid.” Our companion case study covers what really drove that slowdown (a trust crisis about AI, more than format fatigue), but the direction of the evidence is the same: the format’s value was real, and the company that proved it is the one that stopped.

What actually separates the winners

Reading the successes against the failures, the discriminating variable is not audacity. Everyone was audacious; it was 2024. Three things separate the accounts whose baselines moved.

Nothing to protect. Nutter Butter’s marketing lead was told the brand had “no brand meaning” to lose, and one strategist called that its actual advantage. Duolingo was a free app with an owl. Dunkin’, Oreo and McDonald’s carry decades of accumulated meaning, and their unhinged turns read as costume. The format transfers to underdogs and blank slates, not to institutions.

Identity, not campaign. The baselines that moved belong to accounts that changed permanently. The spikes that faded belong to stunts with an end date. This is the single most practical filter: if the plan has a wrap date, expect the results to have one too.

A tiny team with standing permission. Three people at Nutter Butter, one then four at Duolingo, weekly brainstorms, no per-post approvals, low budgets: under five thousand dollars produced Dunkin’s entire spider week. The costs that are never published are the retainers and the humans: Duolingo’s original social lead left in 2025 after a medical leave she attributed to the pace. The playbook is cheap in money and expensive in people, and nobody puts the second part in the award entry.

If you run a small brand and are tempted

The transferable parts are unglamorous. Answer your comments and mine them for ideas, because every account in this piece treats the comment section as the brief. Move in days, not quarters. Pick a voice you can hold for years, at an intensity you can staff. And measure like Nutter Butter got measured: follower counts date-stamped by outsiders, sales data from someone with no stake, and a working assumption that your own campaign recap, like everyone’s, is an award entry.

The unhinged era’s honest summary is that a handful of accounts converted weirdness into durable attention because the weirdness was the identity, and everyone else bought a week of noise. The follower curves and the survey numbers now say the surprise is spent. What is not spent, on the evidence of every success in this piece, is the underlying mechanism: being genuinely present in the conversation, at conversational speed. That part never depended on the owl.