Daily Dose Of Internet's best Short got 7.3M views. Its worst got 127K. Here is the difference.

Daily Dose Of Internet is about as faceless as a channel gets. Same format every time, more than 20 million subscribers, nobody on screen to carry a weak clip. That makes it a clean experiment: when one of its Shorts gets 57 times the views of another, the only things that changed are the clip and the title.

So we took the channel’s last 50 Shorts and divided each one’s views by the channel’s own median Short, which sits at about 643,000 views. Above 1x, a Short beat the channel’s normal. Below 1x, it lost to it.

The best Short did 7.3 million views. The worst did 127,000. That is a 57x gap on a channel where nothing about the production changes. Here is what separates the two lists.

1. The winning titles tell you nothing

Exactly the Right Amount. Of what? Did Not See That Coming. See what? There Was a Hidden Surprise. Where? The three best-performing titles share one property: you cannot know what happens without watching.

The losing titles hand over the payoff before you press play. Dinosaur Plays His Theme Song. Jurassic Park Ballet. Blasting Tunes on a Tesla Coil. Jupiter and Venus are Getting Closer. Each one is an accurate summary, and that is the problem. The viewer already has the whole video in seven words.

2. Music clips flopped

Four of the ten worst Shorts are about music: the dinosaur’s theme song, the guitar, the ballet, the Tesla coil. It is five if you count I Don’t Remember it Sounding Like This. None of the ten best are.

Our guess at the reason: a music clip needs the sound on and a few seconds to land, and in the Shorts feed the swipe comes first. It is a guess. What is not a guess is the split, five against zero.

3. People beat things

Six of the ten best titles have a person as the subject: This Guy, He Made it Rain, Camera Man, Baby. Only two of the ten worst do. The losing list is mostly things: a machine, a Tesla coil, two planets, a car that parks itself.

4. Likes did not predict reach

The 7.3 million view winner had an engagement rate of 2.5%, counting likes and comments against views. Two of the flops were at 5% or higher: Jurassic Park Ballet at 5.5% and the Jupiter and Venus clip at 5.0%. A small audience loved them, and it made no difference to how far they travelled.

If you judge your own Shorts by the like ratio, this is worth sitting with. On this channel, the most-liked clips were among the least seen.

What we would not read into it

Seven of the ten worst titles carry an emoji and only two of the best do. It is tempting to call that a finding. But every emoji title in both lists was published before mid July and none after, and eight of the ten flops come from those same three weeks. The channel seems to have dropped emoji around then. So emoji and timing are tangled together, the data cannot separate them, and we are not going to pretend it can.

How the score works

The baseline is the median of the channel’s 20 most recent Shorts that are at least two days old, so a video still collecting its first views does not drag the number down. Each Short’s score is its views divided by that median. A Short that is itself part of the baseline is measured against the median of the other nineteen, so a hit cannot inflate its own yardstick.

The median matters more than it sounds. An average would be pulled up by the 7.3 million view hit, and every other Short would look like a failure against it. The median answers the question a creator actually has: what does a normal upload do here?

What to do with this on your own channel

1. Find your own median first. A Short at 40,000 views is a hit on one channel and a flop on another. Until you know your normal, you cannot tell which of your videos deserve a sequel.
2. Put your top three next to your bottom three. Read only the titles. If the winners hide the payoff and the losers give it away, you have found the cheapest fix there is.
3. Look for a topic that only appears on one side. For this channel it was music. For yours it will be something else, and you will not see it until the two lists are side by side.

The numbers above were frozen on the day of writing. The live scoreboard for Daily Dose Of Internet will have moved since, which is its own small lesson: an outlier score is a snapshot, and the channel’s median moves with every upload.