The first 30 seconds of a YouTube script, by retention data

Every retention graph on YouTube has the same shape. A near-vertical drop in the first 5 seconds as the casual scrollers bail, a softer slope from 5 to 30 seconds as the rest of the audience makes the decision to commit, and then a long curve from 30 seconds onward that reflects the actual quality of the video. Most of the avoidable retention loss happens in that 5-to-30-second window.

Below are the patterns we see consistently across long-form retention graphs on the channels we operate, and the cold-open structures that hold the most viewers past the 30-second mark.

The 0-5 second drop is mostly unavoidable

You will lose 15-25% of your viewers in the first 5 seconds no matter what you do. These are people who clicked on the wrong recommendation, mis-tapped, expected a short, or are checking whether the audio quality is good enough to bother with. There is no script change that prevents this drop. The viewers leaving in the first 5 seconds were never going to watch the video.

The mistake operators make is reading this drop as a script problem and trying to fix it. The actual lever in the first 5 seconds is whether the audio kicks in immediately and whether the first frame is consistent with the thumbnail. That is all the first 5 seconds are doing.

The 5-30 second window is where scripts win or lose

By second 5, the people watching are giving the video a real shot. The script decides whether they keep watching past second 30. We see four patterns that consistently hold this window. Each one works for a different content shape.

Pattern 1: The specific opening fact

The cold open is a single specific data point that is hard to look away from. Not a question, not a tease, not a meta-statement about the video. A fact.

Example: "In 1987, Reebok sold more pairs of one specific sneaker than the entire NBA had ever sold combined. That shoe is no longer made."

Why it holds: the brain registers the specific number and the specific date as "this person has done research," which earns the next 30 seconds of attention almost automatically.

When it fails: when the fact is interesting but not actually relevant to the rest of the video. If you open with a specific fact and then the next 90 seconds drift into generic context, retention collapses in the next window.

Pattern 2: The hypothetical with stakes

The cold open establishes a what-if that has real consequences if true. Not abstract philosophy; concrete cause and effect.

Example: "If you parked one container ship outside the port of Long Beach for 72 hours, three Fortune 500 companies would miss their quarterly numbers. This actually happened in 2024 and nobody talks about it."

Why it holds: the hypothetical creates a knowledge gap (what was the ship, which companies, why didn't anyone notice) and the second sentence promises the gap will be filled.

When it fails: when the hypothetical is too speculative. "Imagine if AI made every job obsolete" is not a hypothetical, it's a vibe. Viewers leave.

Pattern 3: The cold artifact

Open on something concrete on screen. A document. A photograph. A receipt. A timestamped frame from another video.

Example: "This is a receipt from a 1994 hotel room in Las Vegas. The line items add up to $12,847. The name printed at the top is a person who is currently being sued for $40 million."

Why it holds: the artifact gives the viewer a visual anchor in the first frame, and the narration walks them through what it means. The pacing pulls them along because every sentence reveals a new piece of the artifact.

When it fails: when the artifact isn't actually interesting. If the receipt doesn't hint at a story, the viewer learns to distrust the cold-open setup for the rest of your channel.

Pattern 4: The contrarian summary

Open on the conclusion of the video, stated in a way that contradicts what the viewer expects.

Example: "The fastest-growing fast-food chain in America is profitable on every single store. The reason isn't the food. It isn't the price. It isn't the locations. It is one specific accounting decision that the founder made in 1986 and never told anyone about."

Why it holds: it tells the viewer the topic and signals there is a payoff worth waiting for. The list of negations is doing the work of building suspense.

When it fails: when the accounting decision turns out to be boring. The contrarian setup raises the bar on the payoff. If the payoff is generic, the viewer feels misled and won't click the next video.

Patterns we've stopped using

Three cold-open patterns that show up everywhere on YouTube and consistently underperform in our retention data:

The date-led open. "On March 17, 1994..." or "Back in the summer of 2003...". The date is doing no work. Viewers immediately tune out because the opening sentence has not promised them anything. Whatever fact you were going to lead with after the date, just lead with the fact. The date can come in sentence two.

The question open. "Have you ever wondered why...". The brain has been trained to ignore this exact phrasing because it has been overused in clickbait for a decade. If your hook requires a question to land, write the question as a declarative statement and trust the viewer to register the implicit question.

The meta-statement. "In this video, we're going to look at...". The viewer who clicked already knows what the video is about. The meta-statement burns the first 8 seconds of the 30-second window on information the viewer already has.

The 30-second pivot

Whatever cold-open pattern you use, around second 30, you need a pivot. The pivot is a single sentence that re-states the stakes of the video in fresh language. It is not a summary of what was just said. It is a promise about what is about to come.

Example, after a 25-second cold open about a sneaker: "The reason that specific shoe disappeared from store shelves has nothing to do with the sneaker itself. It has to do with a phone call between two executives that happened six months before the product launched."

The pivot does two things. It rewards the viewer for staying past 30 seconds (they get a payoff sentence, not just continued setup), and it sets up the next 90 seconds of the video with a fresh promise.

Without the pivot, the retention graph shows a second drop around second 45 as viewers who survived the cold open start drifting because nothing new has been promised. With the pivot, the second drop moves to around the 90-second mark, which is where the next 90-second re-hook should land.

What this means for your script template

If you write a single fixed cold-open template for every video, retention will be inconsistent because different topics need different patterns. The script template that works across topic types is structural, not phrasal:

[Cold open: one of patterns 1-4, 5-25 seconds]
[Pivot sentence: 30-second mark, one sentence, fresh language]
[First chapter: 30-90 seconds, delivers the first payoff]
[Re-hook: ~90 seconds, restates stakes in fresh language]

This is what the script generators we ship are tuned to produce, because forcing the model to pick a pattern instead of defaulting to "date-led open + drift" is the single biggest retention lift we can deliver per script.

The 30-second cliff is the easiest retention bug to fix on most YouTube channels. The reason it persists is that operators look at their retention graph, see the drop, and assume their content isn't interesting enough. Usually the content is fine. The cold open is just wasting the first 30 seconds.