The Ninety-Second Video: One Task, One Take, No Introduction

Monday, 06:15, at Tollgate Foods. A ticket arrives from the night shift in the chilled warehouse: "Handheld won't scan again. Watched the video. Couldn't find the bit." The video is "Warehouse handheld: everything you need to know", eleven minutes and forty seconds, recorded by Marcus on the desk with real care. It opens on the company logo, then Marcus's face, then "Hi everyone, Marcus here from IT, and in this video I'm going to walk you through everything about the new handhelds." The thing the night shift needed, reconnecting a handheld to the warehouse Wi-Fi when it drops, starts at six minutes forty and lasts about a minute. Dev, the shift supervisor, says later that nobody on his shift has ever got past the introduction. They are standing in a cold room holding a scanner that does not scan. They did not want everything. They wanted the bit.

A ninety-second video is a recording of one task, from the first action to the confirmation that it worked, with narration or captions and no introduction. It was scripted before it was recorded, recorded in one take, and titled by what it shows. Ninety seconds is not a law. It is about how long one task of eight to twelve actions takes when nothing else is in the recording. A video that runs much longer nearly always contains two tasks or a lecture. This article is part of our Visual & Video Guides series, and it is about making the video the night shift could have used.

It covers why ninety seconds and not eleven minutes. It covers one task, defined by its start and end, then no introduction. Next come narration or captions, and what to say, then scripting before recording and recording in one take. We cover the title that says what it shows, then where the video goes and what sits under it. We finish with a short war story and the video script sheet to copy, in that order.

Why Ninety Seconds and Not Eleven Minutes

The reader of a how-to video is stuck in the middle of a task and wants the next action. And every second of video that is not the next action is a second they spend hunting with the scrubber. That is the entire argument for brevity. A long video is comprehensive from the author's chair, where it was made in one sitting with the whole subject in view. From the reader's chair, at 06:15, it is a haystack with a play button, and the needle is at six minutes forty. They do not have six minutes forty. They have a queue of pallets.

Ninety seconds is what one task takes. Narrate eight to twelve actions at a walking pace, with a beat after each so the reader can follow along on their own device. The clock lands somewhere between sixty and a hundred seconds without anyone trying. When a recording runs to three or four minutes, one of two things has happened: there are two tasks in it, with a "now that that's done, let's" in the middle. Or the narration has stopped describing actions and started explaining background. Both are fixed before recording, by the script, and neither is fixed by talking faster.

The length also sets what the video is for. A ninety-second clip is something a technician can paste into a ticket reply with the words "this shows it". It is something a requester will actually play, because the time shown beside it is shorter than the time they have already spent being stuck. An eleven-minute video is a training session, and training sessions are useful, but they are not help. Help is the bit. (Marcus's video was a very good training session. Nobody was in training.)

One Task, Defined by Its Start and Its End

A task, for the purpose of a video, is anything with a start state the reader can recognise and an end state that proves it worked. The handheld task starts with "the signal icon at the top of the screen shows a line through it" and ends with "scan any barcode; it beeps and the item appears". Everything between those two moments is the video. Nothing before the first is, which rules out the introduction, and nothing after the second is, which rules out "and while we're here, let me also show you".

Tasks that branch are two videos or a video and a sentence. "If you see the login screen, enter your badge number; if you do not, skip to step six" is fine in text, where the reader's eye jumps. It is hopeless in video, where the reader cannot jump and half of them watch a branch that does not apply to them. Record the common path. Put the branch in the text steps beneath the video. Or record the branch as its own clip with its own title if it happens often enough to deserve one. A video that tries to cover every case covers none of them in ninety seconds.

The split test is the title. If you can name the recording with one verb and one object, it is one task. Examples are "Reconnect the handheld to the Wi-Fi", "Reassign a delivery to another driver", and "Add a shared mailbox". If the honest title needs "and", it is two. If the honest title is "everything you need to know", it is a course, and courses are a different article in a different library.

No Introduction: Start on the First Click

The first frame of the video is the screen, in the start state, with the first action about to happen. And the first word of narration is the first action. No logo. No face. No "hi everyone". No "in this video we're going to". No agenda slide listing what will be covered. The reader has arrived at this video by reading its title. So they already know what it shows; telling them again costs fifteen seconds of the ninety and tells them nothing. Nobody has ever watched a help video for the logo.

The same applies at the end. The video finishes on the confirmation, the beep and the item on the screen, with the narration saying what the reader should now see: "The item appears. You are connected." Then it stops. "That's it", at most, and nothing about liking, subscribing, further questions or the desk's opening hours. The article the video sits inside already carries all of that in text, where it can be updated when the hours change. An outro is an introduction that has lost its way.

This is harder than it sounds, because starting cold feels rude. Years of watching other people's videos have taught everyone that a video begins with a greeting. The first time you record one that begins with "Tap the signal icon" it feels like walking into a room and talking without saying hello. Do it anyway. The reader on the warehouse floor does not feel greeted by a logo; they feel delayed. The courtesy they want is the next action, now.

Narration or Captions: Say What You Do as You Do It

Narration is one sentence per action, in the present tense. It uses the name of the control exactly as it is written on the screen, spoken as the action happens. "Tap the signal icon in the top right. Tap the warehouse network. Type the password from the card on the wall. Tap Connect." The sentence and the cursor move together, so the reader hears where to look as they see it. What narration is not is commentary. "Now what we're going to do here is basically go up to the top and find the little signal thing" is five seconds of words for a one-second action. It is how a ninety-second task becomes a four-minute video.

Captions carry the same sentences as text on the screen, and they are not optional. A large share of the people who need the video cannot or will not play sound. Examples include the open-plan office, the shop floor, the person on a phone in a corridor, and the reader who is deaf or hard of hearing. The script you write before recording is the caption text, line for line, so captions cost nothing extra once the script exists. How to produce them, and the transcript that goes with them, is the subject of Accessibility: The Reader Who Cannot Watch. Here the rule is that a video with neither narration nor captions is a video of a cursor wandering about, and is not published.

Leave a beat after each action. The reader is doing the task on their own device while the video plays. They tap, look back at the video, and need it to be at the next action and not three actions on. A one-second pause after each step is the difference between a video people follow and one they pause and rewind eleven times. The table below has the pattern for the sentences themselves.

Script It First

Write the script before you open the screen recorder, and write it as a table. Use one row per action, with what the hands do, the one sentence of narration, and what the screen should show afterwards. This is the same discipline as one action per step in a written article, in Step-by-Step Clarity: One Action per Step, and the reason is the same. A row that contains two actions is a step the reader will fumble. And a row whose narration is longer than one sentence is a step the author has not finished understanding.

Read the script aloud with a stopwatch, performing the actions on a device as you go. The time you get is within a few seconds of the finished video. That means the length argument is settled on paper, before an hour is spent recording. Over ninety seconds? Cut words first: every "now", "basically", "go ahead and" and "what we want to do is" goes. The narration usually loses a third of its length without losing a single instruction. Still over? There are two tasks. Draw a line under the last action that completes the first one, and give the rows below it their own sheet.

The walk-through finds the missing step, which every first draft has. There is always an action the author does without noticing. It might be a dialog that has to be dismissed, a field that has focus on their machine and not on the reader's, or a scroll. Doing the task with the script in one hand catches it on paper. Doing it in front of the recorder catches it too, after the take is ruined. The script sheet at the end of this article is the table. Fill it in, read it aloud, do the task, fix the row you tripped on, and only then press record.

One sheet, three deliverables. The action column is the text steps that go under the video. The narration column is the captions. The rows together are the video. When the screen changes next year, the sheet is what gets edited, and all three come out of it again. A script is not preparation for the video; it is the source the video is generated from.

Record in One Take

Rehearse twice, record once. The first rehearsal is for the hands, finding where the cursor goes and how long each wait lasts. The second is for the voice, reading the narration against the real screen at the real pace. Then the take. If you fumble, if the cursor misses, if a notification pops up, if you say "um", stop, reset the screen to the start state, and begin again from the top. A ninety-second retake costs ninety seconds. Fixing the fumble in an editor costs an afternoon and requires learning the editor. The result still has a visible join where the audio changes tone.

Trim the head and the tail, and nothing else. The head is the second between pressing record and the first action; the tail is the second between the confirmation and pressing stop. Every screen recorder can cut those two moments, and that is the entire edit. No zooms, no transitions, no highlighted cursor effects, no music. The reasons for each are in Cheap Production That Looks Fine, which covers the recording itself: the window, the cursor, the sound and the file. This article is about what to record; that one is about pressing the button well.

Do not narrate over a mistake. "Ignore that, I clicked the wrong thing" is a sentence that will be in the video for as long as the video exists. Every reader will spend the next five seconds wondering what they are supposed to be ignoring. Stop and go again. I have recorded a ninety-second video nine times in a row. The ninth take took ninety seconds. The first eight were the script finding out what it had missed, which is a cheaper way to learn it than the ticket queue.

The Title That Says What It Shows

The title is the task in the reader's words: a verb, an object, and where, if where matters. "Reconnect the handheld scanner to the warehouse Wi-Fi." "Reassign a delivery to another driver." "Set up your e-mail signature." Not "Handheld video 3", not "Scanner tips", not "Everything you need to know about the handhelds", and not the name of the system. The reader does not know the system is called that; they know their scanner has stopped scanning. The title is written in the words the reader would type into search, because for a video the title is the only thing search can see.

Show the length beside the title, so the reader knows before they click that it is short. A reader who has been stuck for ten minutes will play something marked as a minute and a half. They will not play something with no duration shown, on the reasonable assumption that it is eleven minutes long with a logo at the front. Most places a video can be embedded show the duration automatically; where yours does not, write it in the line of text beneath.

Use the same title for the video and for the article it sits in. So a search finds one thing, not two things with different names that turn out to be the same task. Where the video is one of several inside a longer article, each clip's title is its task, and the article's title is the subject. The night shift searched for "handheld won't scan". A video titled "Reconnect the handheld scanner to the warehouse Wi-Fi" is two words away from that search. And "everything you need to know" is on another planet.

Where It Goes and What Sits Under It

The video goes at the top of the article it belongs to, beside the article in the knowledge base. And the full text steps go beneath it. Not in a chat channel, where it scrolls away by Thursday. Not on a shared drive, where its folder is a secret. Not attached to an e-mail, where it reaches the people who were on the list that day and nobody who joins next month. The knowledge base is where the reader already looks. And the article is already titled by the task, so the video inherits a title, a place and an audience by sitting there. Where to keep the file itself and what to call it is in the cheap-production article.

The text steps beneath the video are not a courtesy; they are the version that outlives it. When the vendor moves a button, the text is fixed in a minute and the video is marked for re-recording. In the weeks between, the reader who follows the text gets the right answer. The reader who follows the video gets a note saying which step has changed. That cycle, and the register that drives it, is the subject of Keeping Media Current When Screens Change. The text is also what the reader on the phone, the reader with the sound off and the reader using a screen reader actually use. The video is a kindness to the reader who can watch. The text is the article.

The audience of the video is the audience of its article. A clip of the admin console showing how a technician resets a handheld's enrolment is for technicians. A clip of the handheld showing how a picker reconnects the Wi-Fi is for everyone. If you run QuerySolv, a knowledge base article can be visible to agents only or to everyone. The video goes wherever its article goes. So the admin-console clip stays on the technicians' side without anyone having to remember a separate rule. Whatever the tool, decide the audience on the script sheet, before recording, because it changes what may appear on the screen.

Then tell people it exists, which is the part most desks skip. The best place is the ticket reply: "This video shows it in a minute and a half; the written steps are under it." The second best is the moment of need, where the reader is about to raise a ticket, which belongs to the deflection series. A video nobody is pointed at is a file. Marcus's replacement clip got its first forty views from being pasted into the replies to the next week's "handheld won't scan" tickets. Its next forty came from the pickers sending it to each other, which no announcement has ever achieved.

A Short War Story: The Forty-Eight-Minute Recording at Bellweather Logistics

Bellweather Logistics rolled out a new dispatch system and recorded the one-hour training session. It came to forty-eight minutes once the coffee break was cut, and they posted it in the knowledge base under the title "Dispatch system guide". The dispatchers needed exactly one thing in the first fortnight: how to reassign a delivery to another driver when a van broke down. They could not find that moment in forty-eight minutes and did not try for long. Thirty-one tickets asking the one question arrived in two weeks. Lena on the desk watched the whole recording with a notepad to find the reassignment demonstration. It began at thirty-one minutes and was interrupted twice by questions from the room. She then scripted the task on a single sheet and recorded it in eighty seconds with no introduction. She titled it "Reassign a delivery to another driver", and put the six text steps beneath it. She pasted the link into the reply to every one of the thirty-one tickets. The tickets stopped within the week. Lena wrote the lesson at the top of the recording's page before archiving it. A recording of a meeting is a record of a meeting, and a guide is one task.

The Video Script Sheet

Fill in the sheet before opening the screen recorder. The header settles the task, its boundaries and its audience. The table is the script, one row per action. The timing line is where the length argument is won on paper. The checklist at the bottom is the record that the clip was made the way this article says. Keep the completed sheet beside the article, because it is the source for the captions, the text steps and the next re-recording.

VIDEO SCRIPT SHEET — ONE TASK, ONE PAGE
Title (verb + object + where): ______________________________________
Sits beside article: _________________________   Audience: [ ] everyone  [ ] technicians only
Start state (what the reader sees before step 1): _________________________
End state (the confirmation that proves it worked): _______________________
Device or window to record: _______________   Test account used: [ ]

Step  Action (what the hands do)      Narration / caption (one sentence)     Screen then shows
 1    __________________________      ___________________________________    ____________________
 2    __________________________      ___________________________________    ____________________
 3    __________________________      ___________________________________    ____________________
 4    __________________________      ___________________________________    ____________________
 5    __________________________      ___________________________________    ____________________
 6    __________________________      ___________________________________    ____________________
 7    __________________________      ___________________________________    ____________________
 8    __________________________      ___________________________________    ____________________
 ...  (more than twelve rows: this is two tasks; split the sheet)

Branches that do NOT go in the video (text steps only): _____________________

Timing, read aloud while doing the task: ______ seconds
  Target: about ninety.  Over: cut filler words first; still over: split.

Before recording:   [ ] script read aloud and timed   [ ] task done once from the sheet
                    [ ] rehearsed twice                [ ] notifications off; start state set
Recording:          [ ] one take, restarted on any fumble   [ ] head and tail trimmed only
After recording:    [ ] text steps (action column) pasted under the video
                    [ ] captions or transcript (narration column) attached
                    [ ] title matches the article title; length shown beside it
                    [ ] watched once by someone who did not record it, following along
Recorded by: ____________   Checked by: ____________   Version: ____

Marcus's second video is eighty-four seconds long, begins on the handheld's screen with the line through the signal icon, and ends on a beep. It took him a Tuesday afternoon, most of it on the sheet. The eleven-minute one is still in the knowledge base, retitled "Handheld training session (recording)", for the three people a year who are actually in training. The night shift has not raised a Wi-Fi ticket since, which Dev describes as "the bit, finally".

Frequently Asked Questions