A short screen recording versus a written update, timed

A four-minute recording beats a page of text when the point is sequence. Text wins when the point is a single number. Most handovers contain both.

The four minutes that replaced a page

Sam had to hand a half-built feature to Dana four days before he went on leave. She had never seen the code. The feature touched three services.

He wrote an eight-hundred-word note first. Then he deleted it and recorded four minutes of his screen. The note had taken ninety minutes.

Dana watched it twice, asked two questions, and finished the work without a call. The whole handover cost him eleven minutes in total.

What the recording carried that text did not

He said the words the code does not contain. This is temporary, this was a workaround, this will bite us in March. None of that lives in a comment.

He also showed the order of operations. That is the hardest thing to convey in prose without a numbered list. Lists were banned in their documentation.

His hesitation at one point told Dana more than the sentence he eventually said. Text has no hesitation in it. Video carries doubt for free.

Where video fails

Nothing in a recording can be searched, quoted, or compared later. That makes it a poor home for a specification. Specifications get quoted in arguments.

It is also four minutes long whether the viewer needs thirty seconds or all of it. The viewer cannot skip to the part they need. They must guess.

Sam’s recording was right for a handover and wrong for a policy. The difference is whether the reader will return. A returning reader needs text.

The cost measured in both directions

His note took ninety minutes to write. Dana would have spent about forty reading it carefully. Total cost, one hundred and thirty minutes.

The recording took eleven minutes including a false start. Dana spent four watching and six re-watching one section. Total cost, twenty-one minutes.

The saving was real, but it came entirely from the producer’s side. Reading time barely changed. That asymmetry is the reason recordings get made.

Scripting the first thirty seconds

Most recordings lose the viewer in the first half minute. The speaker starts with background, and the viewer has no reason to wait. Dana would have stopped too.

Sam’s rule was to open with what the viewer can do after watching. Then go straight to the screen. He called it the reason first.

He wrote that one sentence down before recording and nothing else. The rest he improvised, which kept the recording short. A script would have doubled it.

Length as a promise to the viewer

A four-minute recording is a promise. It says the whole thing is worth four minutes. That promise has to be honoured or the format dies.

Anything over six minutes gets a written summary at the top. The reader can decide whether to watch at all. Most decide not to.

The summary also serves the person who will never watch. That is most of the people who find it later. They are the majority of the audience.

Transcripts and the search problem

An automatic transcript makes a recording findable. It is also wrong often enough that nobody trusts a quote from it. Names and numbers suffer most.

The team used transcripts as an index rather than a source. That is the only honest use for them. It gets you to the minute, not to the answer.

Anything that will be quoted in a customer conversation was written by hand. That rule was added after one bad transcript. The bad transcript said thirty days.

Recording a decision versus recording a process

A process benefits from video because steps have an order. A decision does not. Decisions need a sentence that can be quoted later.

Dana needed both. The process of finishing the feature, and the reason three earlier approaches were dropped. The recording carried one of them well.

The reasons she copied out into a document. They will be questioned in a review, and a video is a bad thing to quote. Text survives that.

The update nobody rewatched

Six weeks later someone asked why a particular check existed. The answer was in minute three of Sam’s recording. It had been there the whole time.

The person who asked never watched it. He searched the documents, found nothing, and added a duplicate check. The duplicate cost two days of work.

The information existed and was unavailable. That is the specific failure mode of recorded updates. It is not that video is bad, but that video is unfindable.

When text is cheaper for everyone

Numbers, dates, names, and rules are cheaper as text. The reader needs one item, not the whole flow. Text delivers the item in two seconds.

A written sentence can be skimmed and quoted in a ticket without loss. Nothing needs to be transcribed or trusted. The reader stays in control.

Sam’s later handovers used one paragraph of text above the recording. It held exactly those four things. The paragraph was read by everyone; the video by half.

The split the team settled on

The rule became one sentence. Record anything where the order of actions matters, write anything where a single fact matters. Both formats kept the same title.

In practice that meant recordings for demos, tool setup, and handovers. Text for decisions, dates, and prices. The boundary held without much argument.

Both live in the same place, so a reader finds one entry and chooses how to consume it. Choosing is the point. Being forced is not.

What the viewer does with it later

Dana finished the feature in three days. She wrote a two-page note about what she had learned on the way. Nobody asked her to.

That note is now the reference, and the recording is history. That is the correct lifecycle for both of them. Most teams get it the other way round, which costs them.

Recordings age badly because the interface changes. Notes age slowly because they describe intent. Intent outlives buttons by years, and sometimes by versions.

If you had to hand your current work to a stranger tomorrow, would you talk for four minutes or write for ninety?