Should my homepage use a background video

Default to a still plus optional play. A homepage should use a background video only when the motion is the brand — a trailer, a live cut, a product that moves — and the first viewport still converts on a mid-range phone with the loop off. For trades and local SMBs, the answer is almost always no: the fold job is trust and a phone number, not atmosphere. This is a yes/no on the title card, not a codec tutorial. It sits under Websites That Feel Like Films. If the site already stutters, read Lighthouse Without Killing Design before you add a decoder.

The short answer

• Still first. A graded frame that arrives on time beats a soft loop that arrives late. Optional play is the upgrade path.
• Artists and film brands can earn a muted loop when the footage is the work and the listen / tour / contact path still wins without it.
• Trades and SMBs should not. Tap-to-call, service area, and a real truck beat drone wallpaper over the phone number.
• If you cannot pause it, do not loop it. WCAG 2.2 Success Criterion 2.2.2 requires a pause, stop, or hide control for moving content that lasts more than five seconds.
• I will not invent a conversion lift. Keep the file only if your own CTA clicks, field LCP, and phone path hold. Lab green on a desktop is not evidence.

What job does the homepage fold have if you add video?

The fold has one job. Video does not add a second job. It either serves the first job or it steals it. Lindgaard and colleagues showed people form a stable visual-appeal judgment in as little as 50 milliseconds, and Nielsen Norman Group treats that snap call as the start of a halo on credibility. A decoder that paints late is a title card that missed the premiere.

Name the job in writing before you encode anything.

Fold job	Video helps when	Video hurts when
Listen / watch the work	The clip is the catalog	Wallpaper under a headline nobody can repeat
Book a sprint / inquiry	Atmosphere supports a clear ask	The CTA waits on JS and a first frame
Tap-to-call / get a quote	Almost never	The number sits under a loop and a scrim
Buy / pre-save	Motion proves the release	Autoplay fights the player you actually need
Orient a stranger in five seconds	The still already does that	The loop is the only way the brand reads

If you cannot fill the middle column with a sentence a non-designer would repeat, you do not have a video homepage. You have a still with extra tax.

• Cold visitor can name the offer with the loop paused
• Primary CTA is visible on a 390px-wide phone without the video
• The still alone could ship as the title card
• Secondary actions (tour, merch, services) are not competing for the same eye-line

A homepage that needs the loop to “feel expensive” does not have a brand system yet. It has a mood board.

Run a five-second title card test with the file off. Do not argue in the room.

1. Show the first viewport on a 390px-wide phone. Loop paused or deleted.
2. Hide it after five seconds.
3. Ask only: what do they sell, who is it for, what should you do next?
4. Write the answers down. If they diverge, you have a strategy problem. Video will not unify them.
5. If they remember a texture and not an offer, you built a mood board.

Test result	What it means	Video decision
They name the offer and the next step	Title card already works	Video is optional seasoning
They name a vibe, not an offer	Atmosphere without a job	No. Fix the headline
They ask how to unmute	They think it is a player	It is a player. Give it controls
They cannot find the CTA	The loop (or the scrim) ate the ask	Still + louder primary
They wait for the clip to “start”	You trained them that the fold is a preload	Still must paint first

The loop is not allowed to be the only way the brand reads. If the still fails this test, encode nothing.

Artist/film vs trades/SMB — who should even consider it?

Same production values. Different asks. I will not flatten Arkayla, Foxtide, or Matt Coffey into the same hero as AllCity HVAC. Cinema on a trade site is restraint. Cinema on an artist site is identity — still first, motion as a scene.

Lane	Default	When video can earn a yes	Death state
Artists / musicians	Still + Listen / play	Footage is the work (live cut, video stills, trailer) and play is optional	Full music video as CSS wallpaper, no pause, mute gallery
Film / premium brand	Still + optional play	Motion is the product (reel, lookbook that must move) and reduced-motion has a designed cut	40-second intro that blocks the offer
Trades / local SMB	Still. Phone in the title card	Almost never. A 4-second truck pull can live below the fold on tap	Drone loop over the emergency number
Studio / services	Still + one CTA	Behind-the-scenes only as a later scene, not the LCP	Mood reel instead of named work

Run this as a decision, not a taste argument:

1. Write the primary action in one verb (listen, call, book, request).
2. Put that action on a still. If the still cannot sell it, video will not save it.
3. If the still can sell it, video is optional seasoning — click-to-play, not wallpaper.
4. Only then ask whether a muted loop on large viewports is worth the decoder.

Hundreds of production sites later, the split is boring and true: artist identity can survive a short loop. A plumber’s homepage cannot survive hiding the phone. If you serve both fans and bookers, pick one primary for the homepage. A trailer that sells two movies sells neither.

“The motion is the brand” is a footage test, not a compliment.

Footage	Counts as the work	Does not count
Live cut, official video, trailer, lookbook that must move	Yes — optional play	Wallpaper under the title
Product spin / garment / bottle on turntable	Yes if the product is the offer	Generic stock loop of “particles”
Real truck arriving, real crew, real job site	Below the fold on tap, maybe	Hero wallpaper over the phone
Stock aerials, city flyovers, handshake B-roll	Never	That is rented atmosphere
Slow pan across a still with Ken Burns	That is motion on a still	Do not decode a video for it
Full music video, uncut	A player on a work page	Homepage CSS background

Dual-audience rule I will not bend:

You serve	Homepage primary	Where the other audience goes
Fans + bookers	Identity + listen	Booking / EPK as a later scene or a dedicated URL
Homeowners + commercial	Phone / quote for the next 90 days	Commercial as a path, not half the title card
Buyers + wholesale	The public offer	Wholesale behind a named door
Studio clients + recruiting	The paid offer	Careers is not a second hero

If both audiences need the first viewport, you do not have a homepage problem. You have two products. Video will not merge them.

When does a still plus optional play beat a loop?

Almost always. Optional play is the pattern that keeps cinema without making every visit pay the video tax. The still is the premiere. The file is a bonus reel.

Hero media	Use it when	Do not use it when
Graded still (AVIF/WebP)	Default for every marketing home	You only have a muddy screengrab from a 4K timeline
Still + tap-to-play	The motion is the product and the visitor asked	You need it playing to “feel alive” on first paint
Muted desktop loop, still on mobile	After the still already converts and field LCP holds	You ship the same autoplay file to every breakpoint
Full autoplay background	Almost never	“Competitors have video”

Ship it in this order:

1. Lock the still as the title card — crop, scrim, type, CTA.
2. Add a visible play control that starts the file on tap. Poster stays until then.
3. Honor prefers-reduced-motion: reduce by never fetching the file. MDN and web.dev both treat it as a user preference you query, not a nice-to-have.
4. Only if desktop field data stays inside Google’s LCP bar after that, consider a muted loop on large viewports — still with a pause control.

The cheap-feeling version is the inverse: the loop is the hero, the still is an afterthought, and play is hidden because “it’s a background.” That is not film. That is wallpaper.

The play control has to be a real control, not a rumor in the mock.

• Visible on the still, not only after hover on a desktop
• Large enough for a thumb (the 24px glyph in the corner is a miss)
• Label matches the job: Play, Watch, Listen — not a mystery icon
• Pause is as obvious as play if anything loops past five seconds
• Reduced-motion visitors never see a control that fetches a file they asked not to play
• Focus ring exists. A custom cursor is not a substitute
• Hitting play does not move the headline or the primary CTA (CLS)

Pattern	Honest?	Notes
Still + labeled Play	Yes	Default recipe
Still + tap anywhere to play	Only if the CTA is not the tap target	Do not steal the ask
Autoplay with a tiny pause in the corner	Only if 2.2.2 is actually met	Most “background” themes fail this
No control, CSS background	No	Fails pause; fails reduced-motion
Unmute button as the only CTA	No	You built a player and hid it

If the composition cannot afford a play control, it cannot afford a video. The still already had the job.

How does background video affect conversion?

I will not invent a lift. There is no honest “background video increases conversions by X percent” that I can cite for your homepage, and I will not attach fake percentages to AllCity HVAC, Divine Toke, Foxtide, or anyone else in the reel. Portfolio names prove what kinds of sites get built. They are not a license to fabricate outcomes.

What I will say: conversion on a brand site is the distance between recognition and action. A loop that delays the largest paint, covers the ask, or burns a job-site phone makes that distance longer. Google’s Core Web Vitals are the shared scoreboard, not a religion: LCP at 2.5 seconds or less, INP at 200 milliseconds or less, CLS at 0.1 or less, each at the 75th percentile, mobile and desktop split. Search Central states that trio. A decoder that wakes on first load is a common way film-grade comps miss those bars.

Signal	What to watch	Keep the video only if
Primary CTA clicks / tap-to-call	Event in analytics, same window as the fold	Rate holds or rises vs the still-only cut
Field LCP (p75, mobile)	CrUX / Search Console, ~28-day window	Stays at or under 2.5s
INP	Same field report	Decoder + motion library do not blow 200ms
Bounce / engagement	Directional only — not a conversion proof	You can explain a change without blaming seasonality you did not measure
Reduced-motion / data-saver visits	Qual + the static cut	Those visitors still get a complete title card

How to run the comparison without lying to yourself:

1. Ship the still-only fold to production first. That is your baseline.
2. Add optional play. Do not autoplay yet.
3. If you still want wallpaper, A/B or time-box a muted desktop loop with a pause control. Leave mobile on the still.
4. Read field LCP and the primary event after a full CrUX window, not after one Lighthouse run on office Wi-Fi.
5. If the ask gets worse or LCP leaves the good bar, revert. The still already shipped.

Pretty traffic with no next step is a different failure. Fix the path before you commission another reel.

Conversion traps I see on video homes — none of these get a percentage, because I did not measure your funnel:

Trap	What visitors actually do	Still-plus-play alternative
Waiting for the “real” site to load	Bounce or stare at a blank card	HTML still paints immediately
Hunting for pause / mute	They treat the fold as a player	Give them a player, or don’t
Missing tap-to-call under a loop	They leave and Google the next plumber	Number in the title card
Mute-by-default music video	They never hear the work	Listen CTA that they started
Chat widget + autoplay	Two things moving, no ask	Kill chat on the fold

If a stakeholder asks “but won’t video convert better,” the honest reply is: I don’t know until we have a still-only baseline and a field window. Anyone who answers with a stock industry percentage is selling a reel.

What should I ask a designer about a video hero?

Ask questions that force a decision, not a mood. If the answers are adjectives, you are buying atmosphere. If they name the fold job, the still, the pause control, and the mobile cut, you might be buying cinema.

• What is the one action this homepage exists to get in the next 90 days?
• Can the still carry that action with the file deleted?
• Where does the video file live — first-party, or a Vimeo/YouTube embed stealing the LCP slot?
• What is the mobile cut: still, tap-to-play, or the same autoplay file?
• How do we honor prefers-reduced-motion — no fetch, not animation: none on a hero that was opacity: 0?
• Where is the pause / stop control if the loop runs past five seconds?
• What is the LCP node — the poster still in HTML, or a CSS background discovered late?
• Who owns encoding, captions if the clip carries meaning, and the reduced-motion still in the CMS?
• What do we cut first if a mid-range Android stutters — the loop, the webfont, or the chat widget?

Write the answers in the brief before anyone opens After Effects. A designer who cannot answer the pause and mobile questions is designing a trailer, not a homepage. If they answer with a reel and no still, they are pitching a mood board.

If they say	Translate it to	Your move
“It needs to feel premium”	They do not have a still that holds	Commission photography or a grade, not a loop
“Everyone in the genre uses video”	Peer pressure	Watch those sites on a phone on cellular
“We’ll optimize it later”	Video is the plan; performance is a wish	Still ships first
“Background so it doesn’t distract”	Hidden controls, WCAG miss	Require a pause or kill the loop
“We’ll use a YouTube embed on mute”	Third-party player as the hero	First-party still; embed belongs below

If you are hiring this as a website sprint, those nine checks are the intake. Taste comes after.

What usually fails first when teams autoplay a homepage?

The phone path. Then accessibility. Then LCP. Desktop Lighthouse is usually the last thing to complain, which is why teams ship the failure.

Concrete failure I will not dress up: a trade homepage puts a looping aerial over the fold. The headline is white on pale sky. The tap-to-call button sits under a sticky bar and a chat bubble. On a job-site Android the first paint is a blank card while the MP4 starts. Reduced-motion users get a stuck overlay because the type was keyed off the timeline. Cost: missed calls. Fix: still of the truck, number in the title card, video (if it exists) as a later scene on tap.

Artist version of the same failure: the full music video is the CSS background. Listen is a tiny glyph. There is no pause. Mobile burns data. The brand looks expensive in a screenshot and cheap in the hand.

What breaks	What it costs	What you do instead
Phone CTA covered or delayed	Missed calls, missed listens	Still + CTA in the first viewport; loop optional
No pause on a >5s loop	WCAG 2.2.2 fail; trapped motion	Pause control, or no loop
prefers-reduced-motion ignored	Users who asked for less motion get more	Never fetch the file; instant still
Video as LCP	Late title card, “cheap” feel on cellular	Poster still in HTML carries the paint
Third-party embed as wallpaper	Extra JS, cookies, INP	First-party file or no file
Sound-on autoplay	Browser blocks it; you look broken	Muted or user-started. There is no third option

Fix in this order — do not reshuffle:

1. Restore a complete still title card in HTML. Headline, support, CTA, visual. Visible with JS off.
2. Put the primary action on screen at 390px width.
3. Add pause if any motion runs past five seconds.
4. Gate the file behind tap, or behind a viewport and reduced-motion check.
5. Re-measure field LCP. If it is still late, the loop is not the brand. It is the bug.

Bravery is not a 12MB homepage.

Stakeholder script when the CEO forwards a competitor reel:

1. Open that site on a phone on cellular. Time the first useful paint by feel.
2. Try to complete their primary action with one thumb.
3. Turn on Reduce Motion. See if their offer survives.
4. Only then talk about whether your still is weaker than their loop.

Most of those reels fail the phone test. You do not copy a failure because it looked expensive on a monitor.

What do browsers and WCAG actually require?

This is a yes/no gate, not a polyfill. MDN’s autoplay guide is blunt: media with an audio track is blocked until the user has interacted, the media is muted, or the site is allowlisted. Chrome also scores how often a visitor plays media on a site (a Media Engagement Index) before it will even consider sound-on autoplay. Muted autoplay is the only pattern you can treat as generally allowed, and even that can fail on battery saver, data saver, or a background tab. Mux’s autoplay guide says the same operational thing: you must be prepared for play() to reject.

Safari still wants playsinline for inline muted autoplay. None of that is a reason to autoplay. It is the list of ways autoplay fails in production.

Rule	What it means on a homepage	Default
Muted (or no audio track)	Sound-on wallpaper will not start	Mute, or tap-to-play with sound
User gesture	Unmuted play waits for a tap	Optional play is the honest UI
playsinline	iOS will otherwise fullscreen	Required if you even try autoplay
WCAG 2.2.2 Pause, Stop, Hide	Moving content >5s needs a pause/stop/hide	No loop without a control
WCAG 2.3.3 (AAA)	Motion from scroll/hover/click must be disableable unless essential	Reduced-motion cut is designed, not hoped
Contrast on the still	4.5:1 normal text, 3:1 large — WCAG 1.4.3	Scrim on the real frame, not only in Figma

Decision procedure:

1. If the clip carries speech or music the visitor must hear, it is not a background. It is a player. Give it controls and captions.
2. If the clip is decorative, it cannot be the only way the offer reads.
3. If it loops past five seconds, budget a pause control in the composition — not a hidden ARIA afterthought.
4. If any of those conflict with the art direction, the art direction loses. Ship the still.

A CSS background-image plus a hidden <video> with no controls is the classic miss: it looks authored in the desktop mock and fails the pause criterion the moment it runs.

Decorative vs informative is a yes/no, not a caption later:

The clip…	Treat it as	Homepage default
Carries lyrics, dialogue, or a claim you need	Informative player	Captions, controls, not wallpaper
Is B-roll that adds no meaning	Decorative	Still preferred; if it loops, pause is required
Is the product moving (garment, bottle, live set)	Optional play	Still must work without it
Is a third-party embed with cookies	A vendor on your title card	No. First-party still

Save-Data, Low Power Mode, and background tabs are not edge cases. They are how a lot of real phones arrive. If autoplay rejects, the still has to already be the page. Design for the reject path first.

Condition	Typical browser behavior	Homepage implication
Unmuted autoplay	Blocked without gesture or MEI	Do not plan on sound
Muted autoplay	Often allowed, not guaranteed	Still must stand alone
play() promise rejects	You must catch it	Poster stays; no blank stage
Background tab	Playback may wait	Do not assume the loop is running
Reduced motion	You must not fetch/play	Static cut is the product

Is a YouTube or Vimeo embed a homepage background?

No. An embed is a player from another company sitting in your title card. It brings its own JS, cookies, and often a play UI you do not control. Mute-and-loop hacks (?autoplay=1&mute=1) are not a brand system. They are a way to lose the LCP node to a third party.

Source	Homepage background?	Where it belongs
First-party <video> + poster	Only after the still-plus-play gate	Home, if you passed every other yes
YouTube embed as hero	No	Work page, or below the fold on tap
Vimeo background parameter	Still a third-party player	Same as YouTube: not the title card
Instagram / TikTok embed	No	Never as LCP
GIF pretending to be video	Usually no	A still, or a real file you can pause

Procedure if someone pastes an embed into the hero:

1. Screenshot the fold with the embed blocked (tracker blocker, or JS off).
2. If the offer disappears, the embed was doing the selling. That is a no.
3. Replace with a still you host. Link or tap-to-play to the player below.
4. Keep the vendor player on a work or watch URL if the clip is the catalog.

A homepage that needs YouTube to look alive does not have photography. Budget a still. Then decide if optional play is even worth a first-party file.

When is a custom site worth it for this media choice?

A template can ship a still-plus-play hero. You do not buy a five-figure custom build because you want video. You buy custom when the interaction model, performance budget, or CMS rules cannot live in the builder you already have. The wider custom-vs-template test is the same as the rest of this pillar: pay for authorship when the stack cannot host the recipe. This section is only the media slice.

Constraint	Template / builder sprint	Custom (or heavily rebuilt Framer/Webflow)
Still + tap-to-play	Usually enough	Overkill if that is the only ask
Muted loop, pause, reduced-motion, no fetch	Possible if the builder lets you own the markup	Worth it when the theme owns the hero and will not let you pause
Mobile still, desktop optional loop	Often a custom breakpoint rule	Worth it when the theme ships one file to every device
First-party file, no YouTube-as-hero	Required on either	Custom if the CMS cannot store a poster + src without an embed
Theme forces autoplay wallpaper	Sprint to kill it, or leave	Rebuild if you cannot turn it off

Worth-it checklist:

• The fold job is already written. Video is not a substitute for an offer.
• You can leave the platform with the stills and copy. A locked hero widget is not an asset.
• Someone on the team can replace the poster without a developer every time the single drops.
• You are paying for authorship — type, crop, motion budget — not for “we added a reel.”

If the current site cannot pause a loop or swap a poster, that is an ownership problem wearing a cinema costume. Fix the hero recipe. Do not buy a museum renovation when you needed a storefront.

Builder tells I actually use when the question is “do we need custom for video”:

Stack	Can ship still + optional play	When I still leave
Framer	Yes, if the hero is a component you own	Theme templates that autoplay a CMS video with no pause
Webflow	Yes, native <video> or a component	Interactions that start the file with no reduced-motion branch
Squarespace / Wix video background	Often a wallpaper toggle	If you cannot turn it off, the stack is the no
Custom Astro / similar	Best for HTML-first stills	Only if you need it — not because video is fashionable

• Can an editor swap the poster without touching the timeline?
• Can you ship the fold with JS disabled and still see the offer?
• Is the pause control in the design file, not a “we’ll add ARIA” ticket?
• If you cancelled the agency tomorrow, does the still still exist as a file you have?

If three of those are no, you are not deciding about video. You are deciding about ownership.

How do I measure whether the video is working?

Measure the fold job, not the vibe. “It feels more on-brand” is not a result. Use a window long enough that field LCP exists. Google’s CrUX / Search Console Core Web Vitals reports typically need a rolling ~28-day window at the 75th percentile. A two-day Lighthouse streak on Wi-Fi is a debug tool.

Window	What you can honestly claim	What you cannot claim
Same day, lab only	The file starts, the pause works, the still paints	Conversion, SEO, “users love it”
One week	CTA event direction, obvious JS errors, mobile qualitative	A Core Web Vitals pass
~28 days field	LCP / INP / CLS vs the still-only baseline	A universal lift you can put on a sales deck
After a release / tour / season	Need to separate campaign traffic from the hero change	“Video did that” without a holdout

Procedure:

1. Name the primary event (tap-to-call, form start, listen click, booking). One event.
2. Capture a still-only baseline on production. Screenshot the Search Console CWV card.
3. Ship the candidate (optional play, or muted desktop loop with pause). Leave a revert path.
4. Re-read the same event and the same CWV card after a full field window.
5. Kill the loop if LCP leaves the good bar, INP blows, or the event falls and you cannot name another cause.
6. Keep optional play if the still remains the title card and the file is a visitor-started scene.

I will not tell you “results show in two weeks.” Field LCP does not. CTA direction can. Do not mix those clocks.

Metrics I will not accept as a yes:

Metric	Why it lies on a video home
Desktop Lighthouse Performance	Lab machine, fat pipe, often no Reduce Motion
Time on page	A stuck loop can inflate it
Scroll depth	They may be hunting for pause
“The client loved the reel”	Taste is not the fold job
Views of the MP4	Proves the decoder ran, not that anyone called

Holdout rule: if you cannot keep a still-only path (or a revert) for a field window, you are not measuring. You are hoping.

What should I skip if I only have a week?

Skip shooting a hero reel. A week is enough to kill a bad loop, grade a still, and put the ask on a phone. It is not enough to storyboard, shoot, encode, caption, and prove a background file in the field.

• Delete autoplay wallpaper. Replace with the best still you already have.
• Put the primary CTA on the first screen at 390px.
• Add a pause control to any leftover motion that runs past five seconds — or remove the motion.
• Honor reduced-motion with a static end-state. Do not fetch the MP4.
• Kill third-party hero embeds (YouTube/Vimeo as background).
• Confirm tap-to-call or the form on cellular, not only office Wi-Fi.
• Write the fold job in one sentence. Put it in the brief for whatever comes next.
• Do not buy stock aerials to “hold the place.” A phone still of the truck, the room, or the stage beats a rented sky.

If the brand insists on motion this week, the honest version is tap-to-play on a compressed clip below the fold, or a short transform / opacity entrance on the still. web.dev’s compositor guidance still holds: those two properties can stay cheap. A decoder is not a one-week project.

A week is a direction sprint on the title card. Treat it that way.

What a week is not:

Request	Fits a week?	Do instead
Kill autoplay, ship a still, put the CTA on a phone	Yes	That is the week
Grade and crop existing frames	Yes	Phone stills of the truck / stage beat stock
Shoot a hero reel, encode, caption, prove in CrUX	No	Schedule a later scene
Custom WebGL prelude	No	Still + optional play
A/B a loop with statistical swagger	No	You do not have a field window yet

• One decision-maker on the fold job
• No new footage unless it is a still you already shot
• No new third-party player
• Revert path named before anyone uploads an MP4

When is homepage video not worth doing yet?

Not yet if the still cannot sell the offer. Not yet if the phone path is broken. Not yet if you cannot pause the loop. Not yet if the only argument is that a peer’s reel looked good on a 27-inch monitor.

Blocker	Why it kills the yes	Do this first
Offer is unclear on a still	Video will not write the headline	Rewrite the title card
Phone CTA missing or below a chat bubble	Trades especially: the site exists to be called	Fold sprint, no media spend
No pause budget in the design	You will fail 2.2.2 the moment it loops	Redesign the control, or no loop
No reduced-motion still	You will ship opacity: 0 leftovers	Author the static cut
No field baseline	You cannot tell if the file helped	Ship still-only, wait for CWV
Theme owns the hero and forbids a poster	You cannot implement the default recipe	Change the recipe or the stack
You wanted video to hide weak photography	The loop is a cover	Shoot or select real frames

If three rows in that table are true, the answer is not “maybe later.” The answer is no, for this homepage, until the blockers move. Background video is a scene you add to a working trailer. It is not how you start production.

Go / no-go scorecard. Score honestly. Four or more “no” means the homepage does not get a background file.

Gate	Yes	No
Still sells the offer in five seconds
Primary CTA on a 390px screen without the loop
Pause control in the composition if anything exceeds 5s
Reduced-motion cut does not fetch the file
Mobile is still or tap-to-play
First-party file (no YouTube-as-hero)
Field LCP baseline exists
Footage is the work, not rented sky
Trades/SMB: phone still wins without atmosphere
Someone can swap the poster without a developer

Artists can pass more of those than a furnace company. That is the point of the lane split. Do not borrow an artist yes for a trade homepage.

How many pages does this decision actually touch?

Almost none. The homepage is the trailer. Do not copy the loop onto About, Services, and every landing page because the file exists. Extra URLs with the same autoplay wallpaper split attention and delay launch. Inventory pages by job — how many pages a small business needs is the IA spoke. Video is not a reason to add pages.

Page	Video default	Why
Home	Still + optional play, or still only	Trailer. One title card.
Work / selected projects	Play on the project, not as wallpaper	The clip is evidence, not atmosphere
Services / trades	Never as background	Proof and the phone
About	Still of the people / room	Faces over B-roll
Contact	None	Friction is the enemy
Campaign landing	Still unless the offer is the reel	One job, one visual

Checklist before you paste the same <video> into a layout:

• Does this URL have a different job than Home?
• Would a still of the work / truck / product do that job faster?
• Is the file user-started, with captions if it carries meaning?
• Can the CMS swap the poster without copying a 4K timeline into five templates?

If the sitemap is already a junk drawer, adding cinema to every template makes the drawer louder. Cut pages. Then decide the home title card.

Artist vs trade media map (same files, different jobs):

Asset	Artist home	Trade home
Best still of the work / truck	Title card	Title card
Official video / live cut	Optional play, or work URL	Not the hero
Drone / aerial	Rarely. Mood, not identity	Never over the phone
Customer / fan clip	Work page with consent	Reviews as proof, not a loop
Embed from a platform	Work or watch URL	No

Do not grow the sitemap to “have somewhere to put the reel.” If the clip is the catalog, it needs a work / watch job, not five cloned heroes.

FAQ

Should my homepage use a background video?

Usually no — default to a still plus optional play. Use a background loop only when the motion is the brand, the fold still converts with the file off, and you can pause anything that runs past five seconds. Artists and film brands can earn that yes; trades and local SMBs almost never should.

How do I measure whether a homepage background video is working?

Measure the fold’s primary event and field Core Web Vitals against a still-only baseline. Keep the file only if CTA or tap-to-call holds and mobile LCP stays at or under 2.5 seconds at the 75th percentile. I will not invent a conversion percentage for you. Lab Lighthouse on desktop is not the test.

What usually fails first when teams try this?

The phone path: the ask gets covered, delayed, or pushed below a chat bubble while the decoder starts. Next is the missing pause control on a loop longer than five seconds. Desktop screenshots stay pretty, which is why the failure ships.

How long does this take to show results?

A still-plus-CTA fix can show up in the same week in your own analytics. Field LCP and INP need a rolling window on the order of ~28 days in CrUX and Search Console. Do not claim a Core Web Vitals win after two Lighthouse runs. I will not give you a fake “video ROI by day fourteen.”

What should I skip if I only have a week?

Skip shooting a hero reel. Kill autoplay, ship a graded still, put the primary CTA on a 390px screen, add a pause or remove leftover loops, and honor reduced-motion by not fetching the file. A week is a title-card sprint, not a production shoot.

When is this not worth doing yet?

When the still cannot sell the offer, the phone path is broken, you cannot pause a loop, or you have no field baseline. Peer pressure and “it needs to feel premium” are not reasons. Fix the title card first. Background video is a later scene.

CTA

If the homepage needs a loop to feel expensive, the title card is unfinished. Ship the still, keep the ask, and treat video as optional play.

Explore /websites or book a Website sprint at /contact?intent=websites-sprint.