How to make a product demo video: one problem, one flow, one payoff

The tour is the default because it is the easiest thing to plan. Left to right through the navigation, a sentence per area, ending on the settings screen nobody asked about.

The structure that actually holds attention has three parts, and nothing else gets a slot.

Part one, about ten seconds: the situation, in the viewer's words, before any interface appears. The spreadsheet with four tabs and three colour codes. The Thursday afternoon that goes to the same report every week. If the viewer does not recognise the problem, the interface that follows is noise.

Part two, about forty seconds: one flow, start to finish, at something close to real speed. Not the newest feature and not the one that was hardest to build. The flow the job actually needs.

Part three, about ten seconds: the payoff, stated as a changed state rather than a feature list. The report that used to take a Thursday afternoon is done before the coffee goes cold. Then one next step, said once.

The discipline in all of this is subtraction. A second feature added to a demo does not add a little value, it divides the attention the first one had.

The flow worth showing is the one your best-retained customers actually complete in their first week. Not the most impressive one. Look at the product analytics rather than the roadmap.

A tour answers the question does this product do many things. A viewer evaluating you is asking something narrower: can it do my thing, and how much work will that be. A tour answers the wrong question loudly, and confidence in the delivery does not fix a mismatch in the question.

Fake data undoes a demo faster than bad production ever will. Projects called Test 1 and a customer called Acme tell the viewer they are watching a sales asset. Realistic names, a plausible amount of data, and a little mess on the screen buy more credibility than a better microphone.

Show the unglamorous middle if it is short. If a step takes eleven seconds, let it take eleven seconds. Cutting every wait out makes the demo feel like a claim rather than a recording, and the gap is the first thing a new trial user notices.

The two formats are not competing for the same viewer.

A screen recording alone is right for somebody who already knows they have the problem and is evaluating options. Documentation, the help centre, in-app tips, the follow-up after a sales call. They asked, so the interface is the answer.

A talking head alone is right for somebody who does not yet know they have the problem. Cold feeds, ads, a first touch on social. A face and a voice give a stranger a reason to stay three more seconds, and no interface does that.

The pairing usually beats either: a person for the first eight seconds naming the problem, the screen taking over while the voice keeps going, then back to the person for the payoff. The face buys the attention and the screen spends it on proof.

One decision matters more than that choice, though. Narrate. A silent screen recording with text overlays asks the viewer to read and watch simultaneously, and most people end up doing neither. Then burn in captions on top of the narration, because feeds play silent and plenty of people watch a website demo at a desk with the sound off too.

There is no single correct length, but there are honest ranges by placement, and most demo problems turn out to be placement problems in disguise.

The common failure is one demo asked to do every job at once. It ends up four minutes long, sitting on a homepage, aimed at nobody in particular. The fix is not five shoots. It is one recording session and several cuts of it, which is an afternoon rather than a month.

Two other things are true about watch time and worth planning around. Drop-off is heaviest in the first ten seconds and again at whatever moment you change subject, and both of those are structural rather than production problems. And a longer video that keeps people is worth more than a short one that never earned its opening, so cut for structure before you cut for the clock.

The ranges below are a starting point, not a rule. Measure the drop-off curve on your own and let it move them.

The method that survives a normal week is two recordings and no set.

Record the screen once. The whole flow, unedited, narrating as you go, and do not restart when you fumble. A twenty minute recording with three good runs inside it is a better source than four careful takes that each died at minute two.

Record yourself once, thirty to sixty seconds, phone at eye level with a window in front of you: the problem in your own words, then the payoff. That is parts one and three, and it is the half most teams skip.

What is left is the edit, and the edit is where demo videos go to die. Trimming the silences, cutting the restarts, timing the captions, placing the screen footage over the exact sentence that describes it. Twenty minutes of recording becomes an afternoon, and the afternoon is why version two never exists.

The floor costs nothing and is a genuinely reasonable answer: the screen recorder built into macOS and Windows, the caption tools in the posting apps, and one pass to trim the ends.

Above the floor, a tool should finish that edit rather than hand you a timeline. Cutroom takes one spoken take of up to 3 minutes and returns a finished 9:16 MP4: every word transcribed, the hook written, your clips placed on the exact words that mention them, captions timed in one of six styles, silence and filler trimmed, punch-in zooms added. A batch is 100 credits and the export is 20 credits per exported minute, so a finished 30 second cut is about 110 credits. Lite is $19.99 a month for 1,250 credits, about 11 of those, cancel in one click. The limit worth knowing first: it makes vertical 9:16 only, so it covers the social cut and the widescreen version on your homepage stays a separate job.