Every piece of advice about going viral assumes a normal world, where better work does better. Short video does not work like that, and the gap between what it rewards and what people are told to do is where most of the frustration in this field comes from.
Outcomes in short video follow a power law. Most clips do modestly, a small number do very well, and a tiny number do numbers that dwarf everything else the account has ever posted combined. Any account with a real archive looks like this: a flat field with two or three towers in it.
That distribution is produced by the mechanism itself. A clip is shown to a small group, measured, and either promoted to a larger one or dropped - and promotion compounds, so a clip that clears the bar three times in a row is not three times bigger than one that cleared it once, it is orders of magnitude bigger. Small differences at the start become enormous differences at the end. Nothing about that requires the winner to be better.
In a power law the mean is dragged around by the largest observation, so "my clips average four thousand views" describes no clip you have ever posted. The number worth watching is not the average outcome but how often you produce an outcome at all - because the tail can only be drawn from clips that exist.
The most reliable thing about a breakout is that its follow-up underperforms it, badly, even when the follow-up is better and the audience is already waiting. One of the cases on our page about two overnight breakouts is exactly this: ten million views for the first video, and a fraction of that for the sequel everybody had asked for.
Editing changes the probability that a clip clears the first bar. It does not change the fact that clearing it four times in a row is what produces a tower, and that whether any given clip does is mostly the draw. Ten decent clips have ten chances at a tower; one excellent clip has one. This is arithmetic rather than an opinion about craft.
None of this says the work is irrelevant. It says the work sets your odds per draw, and the number of draws is a separate decision that most people never make deliberately.
What moves the odds is narrow and well understood: where the clip starts (the hook is a starting point, not a line), how long it runs (length is a trade, not a rule), whether it is legible without sound, and whether it is riding something the machine is already pushing (what a trend actually is). If a specific clip stalled and you want the diagnosis rather than the theory, that is why a short gets no views.
If the number of draws is the variable, the cheapest place to find more of them is the recording you have already made. An hour of talk contains six to fifteen self-contained moments; most people post one of them.
SnapDub takes the long recording and returns those moments as finished vertical clips in one pass - the cut chosen rather than sampled, the frame following whoever is speaking, captions burned in, your look applied from a template. One evening of filming becomes a fortnight of entries instead of one.
And when the draws pay per thousand views regardless of whether one of them becomes a tower, the arithmetic stops being a lottery altogether: how clipping campaigns work.
There is no threshold, which is part of the confusion. What people mean by viral is a clip that broke out of its own account's normal range - so on a small account it might be forty thousand views and on a large one four million. The shape is the same either way.
Because the first audience batch is drawn from whoever happens to be on the app at that moment, and a clip's fate is decided largely by how that particular batch behaved. This is the single most common experience in short video and it is structural, not a judgement on the work.
Yes, and it is a short list: where the clip starts, how long it runs, whether it reads without sound, and whether it is riding something already being pushed. All of it moves the probability per clip. None of it decides any individual clip.
In a power-law distribution the tail can only be drawn from entries that exist, so more entries is the one lever with no ceiling. The practical limit is that the entries have to be genuinely different - the same cut reposted is one entry wearing four hats.
Far less often than people expect. A view is a moment of attention, not a decision to follow, and the gap between the two is its own subject.