Your Demo Video Does Not Expire All At Once

Fast, but unevenly. In DORA's 2025 State of AI-assisted Software Development report, based on nearly 5,000 technology professionals, 16.2% of teams deploy on demand multiple times per day and another 21.9% deploy between once a day and once a week (DORA, 2025).

That is deployment frequency, not UI change frequency. Most weekly deploys are backend work, bug fixes and flag flips. The visible surface of a product moves far less often than the release notes imply.

Ask your design lead which screens they plan to redesign in the next nine months. Normally two or three. Build the video so those are the only fragile parts.

Four things, and only four.

UI screens. Any frame showing the live product ages the moment a component moves. A new sidebar, a repositioned primary button, a color token update, and the shot reads as old software.

Product and feature names. Workspaces becomes Projects. Insights becomes Analytics. The rename is a Tuesday for the product team and a re-record for you.

Pricing screens. Plan names, tiers, seat counts. These change more often than the interface does, and a stale price on screen is worse than no price.

Dated copy. "New in 2026." "Now with AI." Anything that timestamps itself decays from the day you publish.

Everything else, and everything else is most of the runtime.

The problem framing in the first ten to fifteen seconds does not touch your UI. It depends on what your buyer struggles with, and that moves on the timescale of your market, not your sprint. "Cuts month-end close from five days to one" is true whether the button is blue or grey.

This is why the framing work in an explainer video outlives the product footage sitting inside it.

The structural mistake is using product footage as connective tissue. A frame of UI here, a frame there, a picture-in-picture under the whole voiceover. When the interface changes, every one of those moments is a separate edit, and a refresh starts costing what a new video costs.

Put the product in defined blocks.

1. Problem framing. No UI.
2. Positioning and category. No UI.
3. Product block one. One workflow, one screen family.
4. Bridge or customer proof. No UI.
5. Product block two. One workflow.
6. Outcome and close. No UI.

Now a redesign of one workflow touches one block. You replace 12 seconds and the other 78 stay locked. The voiceover on untouched sections is never reopened, which matters more than it sounds. Re-recording a full track means rebooking the voice artist and re-timing every cut against a new performance.

We build this way across the work in our portfolio. It is the main reason a two-year-old video is still running.

A screen recording is a flat file. Pixels, baked. If the sidebar changes there is no layer to edit. You go back to staging, seed the same data, match the same cursor path, then re-cut and re-check every downstream edit built on the old timing.

A UI rebuilt in motion design is a project file. The sidebar is a group. The button is a shape with a live text layer. Changing it is a text edit and a color swap. Timing, camera moves and audio sync stay untouched.

That difference is the whole argument. Same result on screen, different cost curve on change two, three and four.

Screen recordings also capture whatever the product did that day: a test account name in the corner, an empty state nobody seeded, a spinner that runs four seconds. A rebuilt UI shows the same flow without that noise.

Treat the deliverable as two things: the exported video, and the layered project that produced it. Only one has a future.

A handover built for maintenance contains:

• The layered project file, UI grouped per screen and named properly, not "Layer 47".
• Every text element live and editable, never flattened into an image.
• Fonts and brand colors saved as reusable styles, not typed in per layer.
• Voiceover delivered as separate stems, line by line, not one bounced track.
• A shot list mapping every timestamp to the screen it shows.

The shot list is the item most studios skip and the one that saves the most time. When a feature gets renamed, someone has to find the four affected timestamps in under a minute. Ask about all five before you commission. It is a standard question in our FAQ, and a studio that cannot answer it is selling a render, not an asset.

A refresh replaces assets inside the existing structure. Timing, script and narrative stay. A reshoot rebuilds from the script up.

Refresh when the interface changed but the workflow did not, when a feature was renamed, when the palette moved, or when you added logos to the integration wall.

Reshoot when the workflow itself changed, so the story no longer matches how people use the product, or when your ICP or positioning moved.

Teams get this wrong in one direction. They see one dated screen and condemn the whole asset. Demand does not pause while they rebuild: 96% of people say they have watched an explainer video to learn about a product or service, and 39% of video marketers have made a product demo (Wyzowl, 2026).

Six rules we apply to every SaaS script.

1. Describe the job, not the feature name. "Routes the ticket to the right team" outlives "Smart Assign".
2. Never say a plan name in the voiceover.
3. Never put a year on screen.
4. Keep numbers in the customer's outcome, not your spec sheet.
5. If a feature name must appear, put it on screen as a text layer and keep it out of the voiceover.
6. Do not name integration partners in the voiceover. Show logos where they can be swapped.

Rule five pays for itself. A script that keeps product nouns on screen and human nouns in the audio absorbs a full rename of your feature set without anyone touching the voice track. The same rule applies to the short-form cuts you pull from the master, where one stale label kills a clip.

We redesign every quarter. Is video worth commissioning at all? Yes, if the fragile blocks are isolated. A redesign that touches one workflow costs one layer swap, not a new production. What kills the economics is product footage running continuously under the whole edit.

How long does a refresh take compared to a new video? A refresh with a clean layered file is normally days, because nothing gets re-approved except the changed frames. A new video restarts script, storyboard, voiceover and review.

Should we just wait until the product settles? Products do not settle. Teams that wait for a stable UI wait two years, then commission under launch pressure. Commission now and refresh on your own schedule.

Who owns the project files? You should, and get it in writing before production starts. If you are weighing whether to keep this capability internally, in-house video team vs agency covers how ownership and idle capacity trade off.

Does this apply to founder content? The economics flip. Founder content is quick to record and short-lived by design, so engineering it for shelf life is wasted effort. Founder-led video covers that model, and SaaS video marketing covers how the two fit together.