Word of mouth is the only distribution channel nobody can buy outright

The theory of word of mouth is well covered elsewhere. This is about the version you meet in practice.

What advertising can and cannot buy

A campaign can reliably buy awareness, meaning that a large number of people know a title exists and roughly what kind of thing it is. It can buy trial to a degree as well, particularly alongside convenience, since a title placed in front of someone at the right moment gets sampled. What it cannot buy is the moment one person tells another that the thing was worth their evening, which is a different transaction entirely.

That recommendation stakes the recommender's own credibility as collateral, and credibility is not available for purchase at any volume. This is why campaigns behind weak titles reliably produce a strong opening period and almost nothing after it.

Why recommendation converts

A recommendation from someone you know contains an implicit model of your taste, which no campaign can replicate at any scale. The recommender has already filtered the title against what they believe you enjoy, so the message arrives pre-qualified for the recipient. It also carries a small social cost if it turns out wrong, and that cost is exactly what makes the signal trustworthy.

Structurally, anything that removes the cost, such as an incentive for sharing, degrades the signal almost immediately and fairly permanently. Communities where this has happened tend to discount recommendations from within them, which is an entirely rational adjustment.

The shape of the curve

Recommendation spreads inside groups that already share taste before it crosses between them, producing a long flat period and then a sharp rise. That early flatness is regularly misread as failure, and titles get pulled or deprioritised before the crossing has had time to happen.

Crossing between groups requires people who belong to several at once, and there are far fewer of those than there are enthusiasts. Anything that makes a title easy to discuss across groups, such as a simple premise or a distinctive image, speeds the crossing considerably. This is a large part of why apparently modest titles occasionally overtake heavily supported ones over a longer period.

Availability is the constraint

Recommendation only works if the person receiving it can act on it quickly, and the window between hearing and watching is genuinely short. A title that is hard to find, restricted to one territory or awkward to reach loses most of the value the recommendation generated.

Structurally, this is one of the strongest arguments for wide availability, and it is regularly outweighed by the money offered for exclusivity. The cost is invisible on any ledger, because nobody records the recommendations that failed to convert into an actual viewing.

Distribution decisions that look sound in isolation can quietly disable the channel that would have carried the title furthest.

Why it cannot be manufactured

There is a substantial industry devoted to simulating recommendation, and audiences have become fairly good at detecting when it is happening. The detection is rarely conscious, but coordinated enthusiasm has a texture that differs from the uneven way people actually recommend things. Real recommendation includes hedges, conditions and comparisons, while simulated recommendation tends to be uniformly and suspiciously positive.

Once a community suspects manipulation, it discounts all enthusiasm from that source, including the entirely genuine kind. The attempt therefore risks destroying the channel it was trying to use, which is an unusually direct sort of penalty.

Attribution is difficult on collaborative work, and a credit is not the whole story.

What this means for how titles are released

Releases built for a large opening and nothing else are betting entirely on paid awareness, which is a bet with a known ceiling. Releases that leave a title available, findable and easy to describe give recommendation the conditions it needs to operate properly. The second approach is slower and harder to report internally, since its effects appear weeks later and resist clean attribution.

That reporting problem, rather than any disagreement about the mechanism, is why the first approach persists as widely as it does. The channel is free, and it is routinely disabled by decisions taken for reasons that have nothing to do with it.

The takeaway

Awareness is purchasable; a recommendation between friends is not, and it decides what lasts.

The premise gets you in. The structure decides whether you stay.