A product video is filmed quickly and integrated slowly. On mobile pages the share of pages with a video element has risen by 32% compared to 2022 (Web Almanac) – the technology behind it has moved more slowly. This article shows which formats work in an online shop, what a video costs in load time and which requirements for captions and alternatives apply to videos published after 28 June 2025 (BFSG).
Why product videos belong in the shop
The frame is manageable: online retail in Germany recently grew to 92.3 billion euros (HDE), and the online share of total retail stands at 13.5% (HDE). Within that frame a purchase is decided on the product page – on an image, on a text and, increasingly, on a video. A video does not replace touching the product, but it shows size, material and handling in a way an image gallery cannot.
Moving images are part of everyday life for your audience. 33% of internet users in the EU subscribed to film, series or sport streaming services in 2025 (Eurostat), and 59% of 16- to 74-year-olds in Germany used social media privately (Destatis). Anyone explaining a product is therefore not competing with a block of text but with a viewing habit.
That does not mean every product needs a video. It pays off where text reaches its limits: proportions, operating steps, material and workmanship. Where the product page already performs well, the effort is often better spent on load time and structure – the usual reference figures are in the article on conversion benchmarks. Which pages qualify is shown by a look at your own numbers, not by an industry recommendation.
The product in motion
Size, material and handling can be shown instead of described – where a still image reaches its limit.
Captions are part of it
Success criterion 1.2.2 of the WCAG requires captions for prerecorded content at level A (W3C) – as part of delivery, not as an add-on.
Load time in view
A video must not push Largest Contentful Paint beyond the target of 2.5 s (web.dev).
How common video on the web really is
The impression is deceptive: video is rarer on the web than it feels. On mobile pages, 6.7% of pages contain a video element (Web Almanac). Anyone integrating a product video properly therefore stands out – not because video is new, but because it is rarely done carefully.
The attributes tell the rest of the story: 23% of video elements carry autoplay, but only 8% are muted (Web Almanac). Mobile browsers do not start a video with sound on their own – a video set to autoplay without the muted attribute simply sits there. If you use autoplay, you mute it, or you do not use it.
The gap is even clearer in accessibility. A caption track is found on only 0.5% of desktop pages with video (Web Almanac). 95.9% of the home pages examined had detectable WCAG failures (WebAIM), and pages with an embedded platform video – 6.2% of the sample – had 9.4 more errors than average (WebAIM).
| Requirement | Reference value | Basis |
|---|---|---|
| Captions for prerecorded video | Required, level A | WCAG 1.2.2 |
| Alternative for time-based media | Required, level A | WCAG 1.2.3 |
| Audio description | Required, level AA | WCAG 1.2.5 |
| Caption synchronisation | within 100 ms | EN 301 549 |
| Caption speed | 160 to 180 words per minute | BBC |
| Display time per word | around 0.3 seconds | BBC |
The median desktop page already loads 1,054 KB of images and 613 KB of JavaScript (HTTP Archive Web Almanac). A video comes on top of that, not underneath. If you embed one, plan the space beforehand: a poster image instead of autostart, loading only on demand, fixed dimensions against layout shift.
Video formats that work in a shop
Not every format suits every assortment. Four basic forms cover most of the ground; they differ in effort, length and in what they can show at all.
Product demonstration
The product demonstration shows the product in use: unpacking, assembling, operating. It is the most demanding form because it needs an action, and the most useful where operation is not self-explanatory. For assortments that need explaining it is usually the first candidate.
360-degree views
A 360-degree view is not a video in the narrow sense but a sequence of images the visitor rotates through. It answers exactly one question – what does the product look like from every side – and costs less bandwidth than a film if the individual images are delivered in a modern image format.
Explainer and how-to videos
Explainer videos address visitors who want to understand a product before they compare. They are often watched without sound: for some online content subtitle usage rises to 35% (BBC). An explainer video without captions therefore loses part of its audience before the question of accessibility is even raised.
Livestream formats
Livestream formats move consulting into a fixed broadcast slot. They are the smallest but fastest-growing part of the field: in 2025 video commerce accounted for 43.22% of the social commerce market share (Mordor Intelligence). For most shops this is an expansion step, not an entry point – the demands on staff and planning are the highest in this list.
Length, pace and load time
There is no evidenced ideal length, but there is an evidenced reference value for pace: captions should run at 160 to 180 words per minute, and a caption stays up for around 0.3 seconds per word (BBC). That limits how much text fits into a minute – and therefore how long a video has to be to carry its content. Technically the familiar targets apply: 2.5 s for Largest Contentful Paint, 200 ms for Interaction to Next Paint and 0.1 for Cumulative Layout Shift (web.dev).
How far that carries is shown by the distribution: in June 2025, 48% of pages achieved good overall Core Web Vitals scores (HTTP Archive). A video that loads unchecked is one of the fastest ways to slip from the better half into the worse one. The basics are in the article on Core Web Vitals.
- A poster image instead of autostart: the video loads only on click or when scrolled into view.
- Fixed width and height on the video element so no layout shift occurs – target 0.1 for Cumulative Layout Shift (web.dev).
- autoplay only together with muted and playsinline, otherwise nothing starts on mobile.
- Ship the caption track as a separate file, synchronised within 100 ms (EN 301 549).
- Keep controls reachable by keyboard and answer interactions within 200 ms (web.dev).
- A transcript next to the video so the content is readable without playing it.
Integration into the shop
In Shopware a video sits in the same media pool as a product image and can be placed in dedicated blocks via shopping experiences. That keeps delivery in your own hands: format, compression and poster image are part of your own configuration, not that of an external player. The same principle applies to WooCommerce and Magento.
For a video to appear in search it needs its own page on which it is embedded, and that page must be indexed; the video must not be hidden behind other elements and needs a thumbnail at a stable address (Google Search Central). Structured data, video sitemaps and the Open Graph protocol are evaluated – a video that is only loaded after a user action will not be found by search. How a page becomes discoverable overall is covered under SEO.
Anyone embedding a video from an external platform brings its scripts and its storage access onto their own page. Storing information on the user device and accessing it is only permitted with consent (Section 25 TDDDG). In practice that means: either the video sits in your own hosting, or it loads only after consent – a placeholder in front of it is the usual solution. How measurement and consent fit together is covered in the article on server-side tracking.
Load time is not a side issue. In one analysis retail conversions rose by 8.4% after load time was improved (Deloitte); one retailer reported 53.37% more revenue per visitor after working on Core Web Vitals (web.dev). Both are individual cases and not a promise – but they show which way a video that is too heavy pulls.
Returns: where a video helps
Returns are everyday business in Germany: on average about every fourth parcel delivered came back, more precisely 24.2% (bevh). For 2021, around 530 million returned parcels are estimated (bevh). That customers may send goods back is not goodwill but law: the withdrawal period is 14 days (Section 355 of the German Civil Code). The question is therefore not whether returns happen but what causes them – there is a separate article on returns management.
The costs can be quantified: a return shipment costs almost twenty euros on average (bevh), adding up to more than 5.46 billion euros a year (bevh). The categories named most often are clothing and shoes at 86.3% and electronics at 27.5% (bevh). The ecological weight matters too: for 2018 the impact was estimated at 238,000 tonnes (bevh).
More than 56% of customers attribute the bulk of their online returns to poor fit or unreliable sizing information (bevh). That is exactly where moving images help: a video showing a garment on a person in motion, or placing the dimensions of a piece of furniture in a room, answers the question before the order. For size guidance the savings potential is estimated at 25% of returns across the market (bevh) – an estimate for the market, not a promise for an individual shop.
What you can measure
Whether a video works shows not on the video but on the page. Play starts, watch time and the path from video to cart can be measured; so can whether the page became slower after the integration. Anyone who does not separate the two mistakes a load-time regression for an effect of the content.
A measurement needs a comparison. Without a before value and without a control group every change remains an assumption; how a robust comparison is set up is described in the article on A/B testing.
The second measurement concerns delivery itself: how many visitors start the video at all, on which devices, and how many drop off in the first few seconds. A video that is barely started is not a question of content but of placement – and that can be fixed without a new production. What to watch out for on mobile is covered in the article on mobile commerce.
Building a video strategy
A video strategy starts with selection, not with the camera. Which products have the highest return rate, which generate the most support questions, which carry the highest order value? The intersection of those three lists is the production list for the first quarter.
- Pick candidates: products with a high return rate, many support questions and a high order value first.
- Choose the format: demonstration, 360-degree view or explainer – depending on which question is open.
- Plan captions and a transcript from the start, do not retrofit them.
- Integrate technically: poster image, fixed dimensions, loading on demand, plus structured data or a video sitemap for discoverability (Google Search Central).
- Measure before and after, with a control group, and only then scale up.
The effort does not scale linearly: the first production costs setup, the tenth almost only recording and editing. It therefore pays to define visual language, length and caption format once and then stick to them. We support selection, integration and measurement – from consulting to implementation in the shop.
Livestream shopping as a market segment
The social commerce market as a whole is put at 2.11 trillion US dollars for 2026, with an expected compound annual growth rate of 29.12% through 2031 (Mordor Intelligence). Livestream formats are a part of that, not a synonym – the figure describes the entire social commerce market, not the revenue of individual retailers.
For a mid-sized shop a livestream is above all a staffing question: it needs a date, a host and a follow-up in which the recording is reused as a regular product video. Anyone who does not plan for that produces once and is left with nothing that stays on the product page. How social channels and a shop can be connected is covered in the article on social commerce.
The requirements for captions and alternatives are set out in the WCAG (success criteria 1.2.2, 1.2.3 and 1.2.5, W3C) and in EN 301 549. Directive (EU) 2019/882 explicitly names the provision of subtitles for instructional videos in Annex II. In German law the BFSG implements these requirements; for services provided using products already in use before 28 June 2025 a transition period runs until 27 June 2030 (BFSG). For discoverability in search the Google Search Central documentation applies. Market figures come from the HDE, Eurostat, the Federal Statistical Office, the bevh and Mordor Intelligence, technical values from the HTTP Archive and web.dev.
Frequently asked questions
For prerecorded content the WCAG require captions at level A (success criterion 1.2.2, W3C). The BFSG exempts prerecorded time-based media published before 28 June 2025 – so new videos are covered. For breaches the law provides for fines of up to 100,000 euros (BFSG). Captions are therefore not a comfort feature.
There is no evidenced ideal length. What is evidenced is the reading pace: captions run at 160 to 180 words per minute, and a caption stays up for around 0.3 seconds per word (BBC). That determines how much content fits into a minute – a more useful yardstick than a blanket number of seconds.
It can. The target for Largest Contentful Paint is 2.5 s (web.dev); in June 2025, 48% of pages achieved good overall Core Web Vitals scores (HTTP Archive). A video that loads only on click and brings fixed dimensions, by contrast, costs the page almost nothing.
There is no robust figure for that link. What is evidenced is the starting point: more than 56% of customers attribute the bulk of their returns to fit or unreliable sizing information (bevh), with a parcel return rate of 24.2% on average (bevh). A video can answer that question before the order – a statement about the effect in an individual shop cannot be derived from it.
Only with consent, if information is stored on or read from the user device (Section 25 TDDDG). The usual approach is a placeholder that loads the video only after consent. Delivering the file yourself avoids the question and keeps control over format and compression at the same time.
Worldwide, 430 million people need rehabilitation because of hearing loss (WHO); among those older than 60, over 25% are affected by disabling hearing loss (WHO). In Germany a good 7.8 million people lived with a severe disability at the end of 2025, which is 9.4% of the population (Destatis); in the EU 90 million people live with a disability (European Commission).