Virality: what Berger (STEPPS) and Gladwell say, what Duncan Watts proved, and why TikTok changed the rules

Virality describes the phenomenon by which a piece of content spreads rapidly among many people, mainly because the people themselves share it with one another. The definition is simple; what we know about why some content goes viral and other content doesn't is far more contested than marketing culture suggests.

There are two schools of thought on virality—the one that believes it can be designed and the one that holds it's mostly random—with respectable authors and evidence on both sides. Knowing both positions is what distinguishes those who design content with realistic expectations from those who try to replicate the latest viral hit with identical technique and are surprised when it fails.

The "it can be designed" school: Gladwell, Berger, STEPPS

Malcolm Gladwell published The Tipping Point: How Little Things Can Make a Big Difference in 2000. The book introduced the idea that social epidemics (including fads, ideas, viral products) follow identifiable patterns that can be characterized and, by extension, attempted to reproduce. His central model, the Law of the Few, proposed that ideas spread mainly through three types of people:

• Connectors—people with very wide social networks who act as bridges between communities.
• Mavens (informed experts)—people with deep knowledge and a willingness to share it, whom others consult before making decisions.
• Salesmen (persuaders)—people with charisma and the ability to convince.

Gladwell argued that identifying and activating these types of people was the key to getting an idea across its "tipping point" toward mass spread. The book was a bestseller and its thesis entered the marketing lexicon for a decade.

Thirteen years later, Jonah Berger, a marketing professor at the Wharton School of the University of Pennsylvania, published Contagious: Why Things Catch On (Penguin, 2013), which proposed an operational framework for designing content more likely to spread. The framework is known as STEPPS, an acronym for six principles:

Social currency—content people share because sharing it makes them look good (informed, fun, original).

Triggers—content associated with everyday cues in the environment, which makes people think of it frequently.

Emotion—content that activates high-intensity emotions (awe, anger, joy, fear, outrage). Berger found in his research that positive emotion generates more sharing than negative, and high-arousal emotion—regardless of whether it's positive or negative—generates more sharing than low-arousal emotion.

Public—content visible in others' behavior. If sharing or using a product leaves a visible trace, other people replicate it (herd effect).

Practical value—genuinely useful content. People share things that help whoever receives the information.

Stories—content wrapped in a narrative that people retell because the story is interesting in itself, carrying the message "embedded" within it.

STEPPS remains the most widely used virality framework in applied marketing, partly because it's operationally useful for evaluating content before publishing it. The practical question it suggests—"does this piece activate at least three of the six principles?"—is actionable.

The "it's mostly random" school: Duncan Watts

The most rigorous critique of the idea that virality can be designed comes from Duncan Watts, a physicist turned computational sociologist, a professor at the University of Pennsylvania (previously at Microsoft Research and Yahoo Research). His book Everything Is Obvious: How Common Sense Fails Us (Crown, 2011) and numerous academic papers attacked several of the pillars of the Gladwell-Berger theory.

The empirical contributions of Watts and his collaborators:

Viral cascades are rare and poorly predictable. In studies of information spread on Twitter, blogs, and social networks, they found that the vast majority of cascade "initiators" reach only one or two people. Long cascades are extremely rare.

The "Law of the Few" doesn't hold up in real data. When Watts and his colleagues empirically studied who generates cascades, they found that seemingly ordinary people—without Gladwell's connector/maven/salesman traits—can be equally or more successful at spreading information. The quality of the connectors turned out to be far less predictive than Gladwell suggested.

Viral success depends heavily on randomness and initial conditions. In experiments where Watts replicated the same initial spread situation several times (the "MusicLab" song experiment, for example), the same content produced radically different results by pure chance of which first people received it. Quality matters, but less than we assume.

The post-hoc narrative explains everything. When content goes viral, analysts can find reasons why it made sense for it to go viral. But those same reasons exist in hundreds of pieces of content that didn't go viral. The explanation works retrospectively without having predictive power.

Watts's conclusion isn't that quality doesn't matter (it does, at the margin), but that virality as a phenomenon is far less designable than the marketing industry claims. A piece with STEPPS well applied has perhaps a slightly higher probability of going viral than one without it, but most viral success still depends on randomness and timing.

The honest synthesis

The two positions aren't completely incompatible. An honest reading of the evidence suggests:

There are characteristics that increase the probability of sharing. STEPPS captures a good part of them. A piece with high emotional charge, practical value, strong narrative, and associated triggers gets shared more, on average, than one without those characteristics.

But the difference between "gets shared a little" and "goes massively viral" is determined mainly by uncontrollable factors. Who sees it first, at what time of day, in what social climate, with which platform algorithm active, with what contemporary events competing for attention. Randomness dominates the extreme outliers.

Designing for "more sharing on average" makes sense. Designing for "guaranteed virality" doesn't. Brands that pursue the second goal invest resources in the search for something that structurally can't be replicated at will.

The fundamental shift: TikTok and the For You Page

Something that changed the conversation about virality after Berger's book was the rise of TikTok starting in 2018-2020 and the spread of its algorithmic distribution model, which decoupled virality from prior audience size.

In earlier social media models (Facebook, Instagram, Twitter), the number of followers was a critical determinant of initial reach. A new account published into the void. To go viral, you usually had to go through audience-building first.

TikTok changed this with its For You Page, an algorithmic feed that serves content to users based on behavioral signals (what they watch, what they skip, what they save, what they comment on), not on follow relationships. An account with zero followers can have a video with millions of views if the algorithm's first testing signals are positive. For the same reason, an account with millions of followers can publish a video that's barely seen if the algorithm decides it doesn't perform.

Other platforms have adopted similar models (Instagram Reels, YouTube Shorts) that algorithmically approximate what TikTok did first. The operational consequence is:

Virality no longer requires a prior audience. Any video can potentially reach big numbers if it passes the algorithm's initial filters.

Followers matter less as a metric. An account with 50,000 followers can have videos with 5,000 views. One with 500 followers can have a video with 5 million. Follower count has become a vanity metric in many contexts.

The "Berger model" applies with caveats. The STEPPS principles are still useful, but now the first test is passing the algorithm, not convincing humans to share. An emotionally strong video that the algorithm doesn't show to enough people doesn't go viral, even if it meets five of the six STEPPS.

The "Watts model" is confirmed by the data. Algorithmic platforms have made the randomness even more visible. Established creators publish videos that don't reach 1,000 views; anonymous accounts have explosions of a hundred million. Predicting which will be which remains nearly impossible.

Real cases of virality and what they teach

Some famous cases and what each one illustrates:

ALS Ice Bucket Challenge (2014). It accumulated 17 million videos uploaded in August 2014 and raised more than 100 million dollars for ALS research. It meets almost all of STEPPS: social currency (showing that you're participating), triggers (asking others activates the challenge), emotion (a social cause), public (the bucket mechanic is visible), practical value (the money raised), stories (each video tells a small story). Almost a textbook case.

"The Man Your Man Could Smell Like"—Old Spice (2010). An ad that became a phenomenon after its Super Bowl debut. Good content, good timing, a social platform at a receptive moment. STEPPS applied well. But replicating it afterward was extremely hard for Old Spice and for other brands that tried a similar format.

Dollar Shave Club's launch video (2012). "Our blades are f**ing great"* generated 4.75 million views in 90 days, founded the brand, and scaled up to a sale to Unilever in 2016 for $1B. It combined humor, clarity, practical value, and a disruptive product. A real case, but also one among hundreds of companies that tried similar viral videos and failed.

Cases that weren't viral but look like they were in hindsight. The post-hoc narrative selects successful cases and ignores the many attempts that failed. This is exactly what Watts criticizes.

Honest practical application for a brand

If a brand wants to use the logic of virality without falling into false hope:

Design with STEPPS principles without guaranteeing yourself a result. Applying Berger increases probability, it doesn't guarantee an outcome.

Produce enough volume. If virality depends partly on randomness, more opportunities increase the probability of a hit. Brands that occasionally "go viral" usually produce a lot of content consistently; they didn't get lucky with a single piece.

Optimize for sharing, not for virality. High sharing in relative terms is achievable and measurable; massive virality is a lottery ticket. Building content that increases sharing above baseline is strategy; chasing massive virality is illusion.

Accept that you don't control the outlier. If a piece unexpectedly explodes, capitalize on it with follow-up campaigns. But don't build the marketing plan assuming it'll happen.

Invest in your own audience base. As covered in detail in how long it takes to gain followers and see results on social media, building a consistent audience pays off more in the medium term than betting on virality. A genuine audience cushions the randomness of each piece.

Common mistakes in the pursuit of virality

Imitating the format of a recent viral hit without understanding why it worked. Very few viral hits are replicable formula by formula. Novelty and timing were part of the success.

Saturating emotions to force sharing. Manufactured outrage, exaggerated sentimentality, and cheap controversy generate a short-term bounce and medium-term reputational damage.

Confusing vanity metrics with useful virality. A video with 10 million views that doesn't convert or build the brand is glamorous noise, not a result.

Having no plan for when the viral moment happens. Some brands go viral and aren't operationally prepared: the website doesn't scale, support is overwhelmed, the product is out of stock, the messaging is inconsistent. The opportunity is wasted.

Investing everything in a single viral bet. A "one bullet" marketing model. Serious brands diversify; going viral is a bonus, not a plan.

Virality and creative operations

Building content with a reasonable probability of generating sharing requires regular, disciplined production, not one-off bets. Brands that occasionally capture viral moments usually have a sustained editorial calendar for months or years before the visible success. Without that operation, accidental virality is a lottery with no system to sustain it.

That's why this discipline connects to creative operations: the editorial calendar sustains regular production that multiplies opportunities, content production scales the creation of variants with a brand system, and reactive capability in newsjacking lets you capitalize on moments when they happen.

At Polimake, that logic lives across three surfaces: Studio to sustain a consistent calendar that generates opportunities, Studio to produce variants with STEPPS principles applied systematically, and Media as the repository where learnings (which type of content your audience shares most) are documented and reused.

---

If you manage brand content, social media, or growth strategy and you've arrived here looking for an answer about virality, the most useful thing you can take from this article is probably the synthesis of the Berger-Watts debate: massive virality can't be designed, but sharing does improve with recognizable principles. Applying STEPPS increases your odds at the margin; pursuing virality as a central strategy is building on sand. Brands that sustain a significant presence on social media do so through consistency and volume, not through a single viral explosion.

To complement this, the tent-pole effect covers intentional planning around key moments (the opposite and complementary strategy to virality), newsjacking covers reacting to third parties' viral moments, and engagement covers the metric that is manageable.

Quick references

• Engagement—a controllable metric, not dependent on randomness.
• Tent-pole effect—deliberate planning that complements the viral bet.
• Newsjacking—the reaction when something goes viral.
• How long it takes to gain followers—the audience-building that cushions randomness.
• Active listening on social media—to detect viral opportunities.