Virality is a science, not luck

The honest claim

Virality is probabilistic science

Virality is not a secret button and it is not pure luck. It is a diffusion process: one person sees something, decides whether it is worth watching and then may pass it to someone else. Psychology affects the decision. The content affects it. The network affects it. The platform decides how many new people get a chance to respond.

Those variables can be observed, tested and improved. That is what makes virality scientific. But science does not make every outcome predictable.

A controlled experiment published in Science placed 14,341 people inside artificial music markets. When participants could see what other people had downloaded, social influence made success both more unequal and more unpredictable. The best songs rarely performed badly and the worst rarely became winners, but almost any result was possible in the middle.

The correct promise is therefore not “we can guarantee a viral Reel.” It is: we can increase the probability of spread by designing and testing the conditions that research shows matter.

A useful model

Five multipliers behind spread

For practical work, think about a Reel as a chain of five multipliers. This is an operating model, not a published platform formula.

If any multiplier approaches zero, the chain becomes weak. A brilliant story with a slow opening may never receive attention. A strong hook with no payoff may hold two seconds but earn no sharing. A useful video shown only to the same small group may perform well without producing a large cascade.

Attention

The opening creates an opportunity

Short-form platforms expose viewers to more content than they can consciously evaluate. The opening is a selection mechanism: it tells the right viewer that this video may be worth the cost of another few seconds.

TikTok's own creative research reports that the first six seconds are vital. Its research found that 90% of ad-recall impact and 80% of awareness impact can be captured within six seconds. TikTok also reports that suspense early in a story was associated with 16% more watch time, while surprise was associated with 1.7 times the view-through rate.

These are advertising findings, not universal promises for organic posts. They support a narrower and useful conclusion: the first frame, first line and first visual change should be designed as deliberately as the ending.

Curiosity

Open a gap the viewer wants to close

George Loewenstein's information-gap theory describes curiosity as the feeling produced when people become aware of a gap between what they know and what they want to know. A useful hook makes that gap visible.

“Wait until the end” creates no meaningful gap by itself. “We changed one shot and doubled the number of viewers who reached six seconds” identifies a specific unknown. The viewer understands what is missing and why the answer could matter.

The gap must be small enough to feel answerable and important enough to justify attention. Give too much information and there is no reason to continue. Hide everything and the viewer cannot judge the value.

A curiosity hook needs three parts

• Context the intended viewer can recognise immediately
• A missing result, cause, comparison or transformation
• Evidence that the answer will actually appear in the video

Shareability

People transmit value, not views

Jonah Berger and Katherine Milkman studied every New York Times article published over more than three months and analysed which ones reached the most-emailed list. Positive, interesting, surprising and practically useful content was more likely to spread.

Emotion was not simply positive versus negative. High-activation emotions such as awe, anxiety and anger were positively associated with virality, while low-activation sadness was negatively associated. In their fitted model, a one-standard-deviation increase in awe was associated with a 30% increase in the probability of reaching the most-emailed list; anger was associated with 34%, anxiety with 21%, practical value with 30% and surprise with 14%. Sadness was associated with a 16% decrease.

That study is evidence of association in real sharing behaviour, supported by the authors' experiments, not a recipe to manufacture anger. A 2024 replication study also failed to reproduce the narrower claim that incidental physiological arousal alone increases social-media sharing. The more defensible lesson is that people share content for a combination of reasons: emotion, usefulness, interest, surprise, identity and social context.

Build a reason to send

Before posting, finish this sentence: “Someone will send this Reel to ______ because ______.”

If the answer is only “because it looks good,” the transmission motive is weak. More specific motives are easier to design for: it helps a colleague avoid a mistake, names an experience a friend will recognise, reveals a local place, provides a checklist worth saving or expresses an identity the sender wants to signal.

Networks

Popular is not always structurally viral

Researchers Sharad Goel, Ashton Anderson, Jake Hofman and Duncan Watts analysed one billion diffusion events on Twitter. They distinguished a large broadcast from structural virality: content can become popular because one large account distributes it widely, or because sharing continues through several generations of people.

Their surprising finding was that structural virality was typically low. Large events often depended heavily on the biggest broadcast, although real cascades showed many mixtures of broadcast and person-to-person spread.

This matters for brands. A creator collaboration can create reach through broadcast. A highly sendable Reel can create a deeper chain. Both are valuable, but they are different mechanisms and should not be reported as the same achievement.

A field experiment involving 253 million people on Facebook found that seeing signals about friends sharing information made people more likely to share and to do it sooner. Strong ties were individually more influential, but the much larger number of weak ties played a dominant role in spreading novel information.

The practical lesson is to make content legible beyond your existing followers. A stranger should understand the subject without knowing the brand, the previous episode or the people on screen.

Novelty with truth

Surprise helps; deception harms

A Science study traced about 126,000 verified true and false stories, shared by roughly three million people more than 4.5 million times. False stories travelled farther, faster, deeper and more broadly than true stories. They were also more novel and produced more fear, disgust and surprise in replies. The difference was driven by human sharing rather than bots.

This is evidence about misinformation, not permission to copy it. The ethical design lesson is to create novelty through a real result, unfamiliar perspective, sharp comparison, unusual visual or specific local insight—not through a claim the evidence cannot support.

Deceptive reach can damage trust, attract the wrong audience and make future content harder to believe. Scientific marketing measures the downstream result, not just the spike.

The experiment

Turn every Reel into data

Science advances by testing hypotheses. Content should work the same way.

Write one hypothesis

Example: “Showing the finished transformation in frame one will increase two-second retention because viewers can see that a payoff exists.”

Change one meaningful variable

Keep the body, length, topic and call to action stable. Change the opening visual and line as one hook package. If everything changes, a better result teaches you almost nothing.

Choose the success metric before posting

Use two-second retention for a first-frame test, average watch percentage for pacing, completion for the ending, shares per reach for transmission and enquiries or follows for business relevance.

Compare like with like

Compare videos of similar length, subject, audience and distribution. A ten-second tutorial and a forty-five-second story should not share the same completion target.

Repeat before declaring a rule

One successful Reel is an observation. A pattern repeated across several controlled tests is evidence. Keep a record of the hook, length, topic, retention points, shares per reach and final business action.

Scorecard

What to measure and why

Do not use likes as the only definition of success. Likes are visible and easy, but virality requires continued distribution. A Reel with fewer likes and more sends may be the stronger transmission asset.

A repeatable system

What can actually be engineered

Virality is not a format that can be copied frame for frame. It is the emergent result of human decisions moving through a network. The science helps us design better inputs, identify weak links and improve the odds. It also tells us to remain humble about prediction.

The conclusion

Luck is what we call the part of the system we did not control or could not observe. Science does not remove that uncertainty. It replaces superstition with hypotheses, measurement and iteration.

Research

Peer-reviewed papers and platform evidence

• Berger and Milkman, “What Makes Online Content Viral?”, Journal of Marketing Research
• Salganik, Dodds and Watts, “Experimental Study of Inequality and Unpredictability in an Artificial Cultural Market”, Science
• Goel, Anderson, Hofman and Watts, “The Structural Virality of Online Diffusion”, Management Science
• Bakshy, Rosenn, Marlow and Adamic, “The Role of Social Networks in Information Diffusion”, WWW 2012
• Vosoughi, Roy and Aral, “The Spread of True and False News Online”, Science
• Loewenstein, “The Psychology of Curiosity: A Review and Reinterpretation”, Psychological Bulletin
• Prowten and colleagues, two replications of the arousal-and-sharing experiment, Psychological Science
• TikTok Creative Codes: hooks, attention and first-six-second findings
• Instagram Reels insight definitions
• TikTok video play metric definitions