A great deal of professional effort goes into making things spread, and the success rate is poor. That is not incompetence, and there is reasonable evidence that the outcome is inherently unpredictable.
The cumulative advantage problem
The clearest structural explanation, and it comes from research on cultural markets rather than from marketing.
Studies constructing parallel artificial markets, where participants could see what others had chosen, found that the same items achieved wildly different outcomes in different runs.
Quality had some influence — genuinely poor items rarely succeeded, and genuinely good ones rarely failed completely — and within the broad middle, outcomes were largely determined by early random variation amplified by social influence.
An item that happened to receive early attention received more, because attention attracts attention.
Which means success is path dependent. Rerun the same week with the same content and different things would spread.
What this implies
Several uncomfortable conclusions for anybody trying to engineer it.
Retrospective explanations of why something spread are largely storytelling. Any successful item has identifiable characteristics, and so do the thousands of similar items that failed.
Replicating a successful formula does not work reliably, because the formula was not the cause.
And the distribution of outcomes is extremely skewed, with a tiny proportion of content receiving most of the attention, which means average performance is a poor guide to anything.
What does appear to help
Not prediction, but improving the odds.
Volume, straightforwardly. If outcomes are largely random within a quality band, more attempts produce more chances. This is the actual strategy of most successful creators, whether or not they describe it that way.
Being in the quality band at all, since genuinely poor material does reliably fail.
Early seeding to an existing audience, which addresses the initial randomness by ensuring some early attention rather than none.
Emotional intensity, which appears in research on sharing behaviour as one of the more consistent correlates — content producing strong reactions, particularly high-arousal ones, is shared more than content producing mild ones.
And practical utility, which is the other consistent finding. Genuinely useful things get sent to specific people for specific reasons, which is a different and more reliable mechanism than broad spread.
The platform variable
A factor that has grown in importance and that nobody controls.
Distribution is now determined substantially by recommendation systems rather than by social sharing, which changes the dynamics.
Those systems change, without notice, and what they favour shifts. A format that performed for months can stop performing for reasons entirely external to the content.
Which adds a further layer of unpredictability, and it means that even the correlates that hold at one moment may not hold six months later.
What professionals actually do
Given all of this, the sensible strategies are unglamorous.
Produce consistently rather than betting on individual pieces.
Build an owned audience, which converts randomness into something more reliable.
Measure aggregate performance over time rather than reading anything into single results, which are dominated by noise.
And treat a large success as a fortunate event to be capitalised on rather than as evidence that a method has been discovered.
The uncomfortable version
The people who have had one enormous success and now sell advice about how to do it are, on this reading, mostly selling an explanation of a coin toss.
That is not to say nothing can be learned or that skill is irrelevant. Skill moves you into the band where success is possible.
Within that band, the specific outcome is substantially luck, and the honest position is to say so rather than to construct a narrative that flatters everybody involved.
The measurement problem
An additional difficulty that makes learning from outcomes harder than it looks.
Platform metrics count different things under similar names. A view can mean a fraction of a second of a video appearing in a feed, or several seconds of deliberate watching, depending on the platform.
Which means comparing performance across platforms is close to meaningless, and comparing across time is unreliable because definitions change.
Anybody drawing conclusions from these numbers is drawing them from a measure they did not define and cannot audit.
The cost of chasing it
The practical argument against the whole pursuit.
Optimising for spread pushes toward emotional intensity, controversy and novelty, because those correlate with sharing.
The audiences that arrive through those mechanisms are generally transient, and the accounts built on them tend to require escalating intensity to sustain attention.
Which is a poor position to occupy, and it is why the creators with durable audiences are frequently the ones who never had a large breakout at all, and simply produced consistently for people who came back.
Why the advice industry persists anyway
Worth addressing since it is a large industry.
The demand is real. People making things want to believe the outcome is controllable, and an explanation is more comfortable than randomness.
The supply is easy, because any successful case can be analysed after the fact into a set of principles that sound persuasive.
And the claims are close to unfalsifiable, since failure can always be attributed to poor execution of the principles rather than to the principles being wrong.
Which is a stable arrangement that will continue regardless of what the research says, and the useful defence is simply to ask whether the person advising has succeeded repeatedly or once.