When should a small business redesign their website

Redesign a small-business website when an operating trigger fires: conversion collapsed while sessions still arrive, the job cannot finish on a phone, the CMS will not let you publish the site you now need, the offer on the page is no longer the offer you sell, or the photos and trust assets stopped matching the company. Calendar age is not a trigger. “It has been three years” is a meeting, not a diagnosis.

This spoke sits under Websites That Feel Like Films. I have shipped hundreds of production sites. The expensive mistake is buying a new look because the anniversary arrived, while the leak is a buried tap-to-call, a form that never hit the inbox, or photos of trucks you sold.

The short answer

• Conversion collapse. Sessions hold or rise. Calls, forms, bookings, or checkout fall. Tracking is proven. The fold has no single job.
• Mobile fail. A thumb cannot name the offer, complete the primary action, or read the proof on a real phone. Desktop looking fine does not count.
• CMS lock. You cannot add the pages, fields, or performance the business now needs without a developer for every comma — or you cannot leave with your work.
• Offer change. New service line, new geography, product to service, shop added, or a line you no longer sell still sitting in the hero.
• Trust / photos stale. People who left, vehicles you sold, work you no longer do, licenses that expired, reviews frozen in a past year.

Trigger	What you can point at	Redesign if	Do not redesign if
Conversion collapse	GA4 + ledger: sessions vs calls / forms / jobs	Path and fold are the leak after tracking is proven	Events are broken, ads died, or GBP still gets the calls
Mobile fail	Thumb test on a mid-range phone, cellular	Primary action is off-screen, tiny, or competing	Desktop Lighthouse is the only complaint
CMS lock	Failed editor task, export, Core Web Vitals inside the tool	Theme owns every layout; campaigns cannot ship	You are bored of the typeface and the editor still works
Offer change	Current services vs homepage copy	The site sells a company you no longer run	You added a bullet and the IA still fits
Trust / photos stale	Staff, fleet, before/after, badges, dates	Proof would embarrass you in a sales call	Photos are old but still true and labeled

If you cannot fill the “what you can point at” cell with an artifact, you do not have a trigger yet. You have a feeling.

What does conversion collapse actually look like?

Conversion collapse is a split: traffic still shows up, money-making actions do not. If both traffic and actions fell, you may have a demand problem, a ranking problem, or a tracking problem. That is not automatically a redesign.

Name the action the site exists to produce. For a trade company it is usually a call, a book, or a quote form. For a shop it is add-to-cart then paid order. For a studio it is a qualified inquiry. Then look at that action against sessions for the same period, on the same device split.

Signal	Collapse	Not collapse
Sessions	Hold or rise	Fell with the actions
Primary action (calls, forms, bookings, orders)	Down in the business ledger, not only in a dashboard	Dashboard down; inbox and calendar still full
Device split	Mobile sessions fine; mobile actions cratered	Desktop and mobile fell together after an ads pause
Fold	Cannot say the offer out loud from the first screen	Offer is clear; the calendar is seasonally quiet
CTA	Four peer buttons, or the real ask is below a video	One ask, and people still complete it on phones

A trade homepage that hides the number behind a “learn more” stack will leak even if Search Console looks healthy. An artist homepage that is a gallery with no listen / tour / contact path will leak even if the stills are beautiful. Same trigger. Different ask.

How this affects conversion: a redesign only helps if the leak is composition, proof, or path. It does not restore a paused campaign, a dead phone number in the header, or a form that posts nowhere. Fix those first. Then decide whether the remaining leak is a design job.

• Sessions for 90 days vs the prior 90, mobile and desktop split
• Primary action counted in the ledger (call log, CRM, inbox, Stripe), not only as a GA4 event
• Form test from a phone on cellular — thank-you state, human inbox, spam folder checked
• Header phone is a real tel: link and matches Google Business Profile
• One primary CTA on the first mobile screen

If the ledger is healthy and the founder is bored, that is not conversion collapse. Do not spend a redesign budget on a mood.

False collapses I want named before anyone opens a new Framer file:

What you heard	What it often is	First move
“Leads died after the new site”	Tag renamed; form to staging; tel: dropped	Reproduce one lead end-to-end
“Nobody calls from the website”	GBP is the call surface; site never had a job	Give the site a job, or stop blaming the fold
“Traffic is up, jobs are down”	Wrong query class, or the fold now attracts browsers not buyers	Check Search Console queries vs the offer sentence
“Mobile conversion is terrible”	Desktop-only CTA, or INP wrecked by a chat embed	Thumb test before a visual redesign
“We used to rank and now we don’t”	Index / move / cannibalization, not a typeface	Search Console coverage and URL map first

A redesign affects conversion when the leak is recognition, path, proof, or offer truth. It does not resurrect a campaign you paused in Ads. Write which of those four you are buying. If you cannot, you are shopping.

When is a mobile fail a redesign trigger?

A mobile fail is not “the site looks smaller.” It is: a person with one thumb cannot finish the job the business is paying the site to do. Google indexes with a smartphone crawler under mobile-first indexing. A desktop-only premiere is not the site Google or a job-site customer actually uses.

Do this with the mouse unplugged, on a phone that is not this year’s flagship, on cellular:

1. Land on the homepage.
2. Say the offer out loud from the first screen.
3. Complete the primary action (call, book, listen, buy, inquire) without pinch-zoom.
4. Confirm a success state a human would trust.
5. Note every extra tap, extra field, and extra decision.

If step 2 fails, you have a fold problem. If step 3 fails, you have a path problem. If step 4 fails, you have a launch problem. Any of those, stuck for months, is a redesign trigger. A 12-column desktop grid that merely squashes is a fail even if the brand still likes the Figma file.

Fail state	What it looks like	Redesign move
Offer not on the first screen	Hero is a slider, a reel, or a paragraph	One headline, one ask, still visible at ~390px
CTA not tappable	Tiny text link, stacked under a chat bubble	One primary button, thumb-reach, no overlay
Proof unreadable	White type on a pale still, no scrim	Contrast that holds on the real photo (WCAG 2.2 1.4.3: 4.5:1 normal text)
Tap targets too small	Phone, email, and nav all fight in one bar	Size that meets WCAG 2.2 2.5.8 (24×24 CSS px minimum; larger is kinder on a job site)
Performance as cheapness	Blank hero, late LCP, layout jumps	Image and motion budget; LCP ≤ 2.5s, INP ≤ 200ms, CLS ≤ 0.1 at the 75th percentile
Form as a novel	Eight fields before a human will talk	Name, contact, one qualifier; the rest after

Core Web Vitals are a shared language, not a religion. I use Google’s “good” bars in production because they are published and checkable in Search Console. I will not invent a conversion lift for passing them. Slow cinema still feels cheap. That is the claim. The percentage is yours to measure after, not mine to fabricate here.

Desktop-only pride is how trade sites leak weekend emergency calls. If the phone path is the business, mobile fail is enough to redesign even when the founder’s monitor looks “premium.”

The job on the phone is not the same in every lane. Score mobile against the actual ask:

Lane	First-screen job on a phone	Fail that forces a redesign
Local trade	Trust + tap-to-call or book	Number in an image, or below a slider
Shop / merch	Product truth + a cart path that does not fight identity	Atmosphere hero, no price or stock truth, tiny add-to-cart
Studio / services	Offer + one inquiry path	Four peer buttons, form that asks for a novel
Artist / catalog	Identity, then listen / tour / contact	Mute gallery; play control off-screen

Same trigger. Different primary action. Do not copy a studio’s inquiry fold onto a trade site that lives on the phone number. Do not copy a trade tap-to-call onto an artist home whose first job is listen.

When is CMS lock the real reason to redesign?

CMS lock is not “I do not enjoy the editor.” Lock means the tool will not let the business ship the site the offer now requires — or will not let you leave with the work. A redesign that stays on a locked theme is a costume change. The ceiling is still there on Tuesday.

Test	Pass (inconvenience)	Fail (lock)
Weekly copy	Owner or manager edits a field and publishes	Every headline is a developer ticket
New service / project	Duplicate a recipe, fill required fields	Theme has no collection; you clone a whole page and break the nav
Campaign landing page	Same system, new URL, locked components	New page means a new purchased template
Performance	You can swap the hero image and keep LCP honest	The theme injects sliders, chat, and three fonts you cannot remove
Exit	Export or transfer under your accounts	Content lives in a vendor login you do not own

WordPress with a page-builder soup, Squarespace or Wix with a theme that owns every composition, a Webflow site whose client editor is a blank canvas, a Framer file nobody on the staff can open — those can all be locks. The brand of the CMS is not the diagnosis. The failed task is.

I have watched owners buy a “redesign” that is a new premium theme on the same locked stack. Six weeks later they still cannot add a service-area page without cloning junk. That is not a visual problem. That is a system problem wearing a new typeface.

• Who publishes next Tuesday, named, with a live login
• One real edit completed in the meeting: headline + image + publish
• Image field that enforces a ratio, or at least rejects a 4000px PNG dump
• Homepage cannot accept “any section”
• Domain, hosting, and CMS billing sit under the business, not the freelancer
• You can describe the exit: export, transfer, or rebuild on something you own

If three of those fail, stop picking color. The trigger is lock. A sprint can still be the right buy when the lock is the fold and the editor, and the stack can be made honest. A rebuild is the right buy when the platform is the ceiling. This post is the trigger list. The scope split is a different conversation. Name lock first so you do not buy the wrong invoice.

What lock sounds like in a kickoff, translated:

1. “Our guy handles updates” — there is no editor. There is a bottleneck. Score lock at least 1.
2. “We bought a new theme last year and still cannot add a location” — theme is the ceiling. Score 2.
3. “We’ll move to Webflow as part of the redesign” — that is a platform change. Treat it as lock-plus-migration, not as a paint pass.
4. “The last designer has the login” — you do not own the site. Ownership is the first job; visuals wait.
5. “It is fine, we just want it to look current” — not lock. Do not let lock language sneak in later as scope creep.

If sentence 3 or 4 is true, write it on the statement of work as its own workstream. A “redesign” that quietly includes a host move and an account transfer will blow up tracking, DNS, and the editor the week you announce.

When does an offer change force a redesign?

An offer change is a business fact the current site still contradicts. New service line. New city. You stopped residential and went commercial. You added a shop. You killed a SKU that is still the hero. The site is lying, even if the type is current.

That is a redesign trigger because the jobs of the pages changed. A new coat of paint on a lie is still a lie. It is not a trigger because you added a footnote in the footer.

Offer delta	What must change on the site	What you can keep
New primary service	Homepage title card, primary CTA, proof that matches	Domain, most URLs if the old service still exists as a secondary page
Dropped a line of work	Hero, nav, service index, ads landing pages	Blog posts that are still true if labeled as archive
New geography	Service-area proof, NAP, GBP alignment, fake city clones cut	Brand system if it still fits
Product → service (or the reverse)	Fold job, form vs cart, proof type	Photography only if it still depicts the work
Shop added	Merch or product path that does not hijack the primary ask	Identity pages, if the shop is a scene, not the title card

Do not “redesign” by adding twenty thin pages because a competitor’s footer has them. Inventory jobs, not a magic sitemap — that is the argument in how many pages a small business needs. An offer change often reduces the honest page count: kill the lines you no longer sell, give the new line a real job, keep Home / services / proof / contact.

1. Write the offer you sell in the next 90 days in one sentence a stranger can repeat.
2. Screenshot the current first viewport. Read it out loud. If it names a different company, you have the trigger.
3. List every nav item. Mark keep, demote, or kill.
4. Map proof to the new offer. Old awards for a dead product line are not proof.
5. Align name, address, phone, and services with Google Business Profile. Mismatch is a quiet tax.

A Divine Toke-style shop whose site still leads on a product you do not stock is an offer-change trigger. An AllCity HVAC-style trade site whose hero still says “residential only” after the book of work went commercial is the same trigger. Pretty is irrelevant until the sentence is true.

When do stale photos and trust assets force a redesign?

Stale trust is not “the site looks like 2019.” Stale trust is evidence a skeptical buyer can catch you in. Staff who left. Trucks you sold. Before/after from a quality standard you no longer meet. License badges that expired. A reviews widget frozen on a year nobody on the crew remembers. The founder’s headshot from a different decade while the about page claims a current team.

Visual age without a lie is a refresh conversation. A lie in the proof is a redesign trigger because the composition has to make room for true assets, and the old gallery is often load-bearing.

Asset	Stale (redesign the proof)	Dated but true (maybe a photo pass only)
People	Named humans who do not work there	Same owners, older photo, still them
Fleet / rooms / merch	Vehicles, interiors, or products you cannot show in person	Same assets, less flattering light
Work samples	Jobs you would not take / cannot repeat	Older jobs you would still stand behind
Reviews / ratings	Widget that does not match GBP or the inbox	Real reviews, not freshly designed cards
Credentials	Expired licenses, dead associations	Current credentials, ugly PDF badges
Dates	“Est. 2014” as the only proof; last blog 2018 used as if it were news	History that is labeled as history

Lindgaard and colleagues showed people form a stable visual-appeal judgment of a webpage in as little as 50 milliseconds; Nielsen Norman Group treats that snap call as the start of a halo on credibility. A first viewport of stock handshakes and a team that does not exist spends that 50 milliseconds on a fiction. I will not attach a conversion percentage to replacing those photos. I will say the sales call gets shorter when the site and the truck match.

• Every face on the site still works there, or is labeled as founder / archive
• Every vehicle, room, and product still exists in the business
• Before/after pairs are jobs you would still show a skeptic
• Licenses and insurance dates are current
• Review counts on the site do not contradict Google Business Profile
• Alt text describes the real thing, not “team image 3”

If four of those fail, you do not need a new gradient. You need a proof system. That is a redesign of the evidence scenes, not a theme swap.

Why “it’s been three years” is not a trigger

Calendar age is a reminder to inspect, not a reason to buy. A three-year-old site that still converts, still edits, still tells the truth, and still works on a phone may need new photos and a copy pass. A fourteen-month-old site with a buried number and a slider hero already needs a redesign.

Agencies sell anniversaries because anniversaries are easy to invoice. Triggers are harder. They require a ledger, a phone, and an editor login.

Calendar story	What to inspect	Likely job
“It’s been three years”	Triggers table above, in order	Whatever actually leaks — often nothing structural
“Competitors look newer”	Their fold job vs yours; their phone path vs yours	Craft or proof, not a clone of their theme
“We have budget this quarter”	Rank the five triggers; spend on the loudest	Smallest honest scope, or wait
“The platform version end-of-lifed”	Can you still publish and leave?	That can be lock — still not “three years”
“Our logo changed”	Where the old mark still ships (PDF, OG, app icon)	Identity pass; redesign only if page jobs also changed

Boredom is not a diagnostic. Neither is a fiscal-year leftover. If none of the five triggers fire, keep the engine. Update the stills. Do not burn the URL tree for a feeling.

A site can be visually tired and still be the wrong thing to rip up. If calls still arrive, the CMS still cooperates, the offer is true, and the photos are honest, you are looking at maintenance. Maintenance is cheaper than a premiere. Spend the premiere when a trigger is on fire.

Inspect on a cadence. Buy on a trigger. A practical inspection rhythm that does not become an anniversary invoice:

Cadence	What you inspect	Buy a redesign only if
Monthly	Form test, tel: test, GBP vs site NAP	A path is dead
Quarterly	Sessions vs ledger, mobile split, Search Console queries	Collapse with sessions holding, after tags are proven
Twice a year	Proof inventory (people, fleet, work, credentials)	The proof would fail a sales call
When the business changes	Offer sentence vs title card	The site sells a past company
When you cannot publish	One real CMS edit	The tool is the ceiling

The calendar here is a reminder to look. It is not a coupon that expires at year three.

How do you score the five triggers?

Score them in writing before anyone opens Figma. One loud trigger sets the job. Two loud triggers still get sequenced. Five “kind of” scores are a bored founder with a credit card.

Use a 0–2 scale: 0 = cannot point at an artifact, 1 = annoying but the job still finishes, 2 = the job fails or the site is lying.

Trigger	Score 0	Score 1	Score 2
Conversion collapse	No numbers, or tracking unproven	Soft dip; path still obvious	Ledger down, sessions hold, fold/path is the leak
Mobile fail	Phone path works on cellular	Ugly but completable	Cannot name offer or complete the ask with a thumb
CMS lock	Owner published a real change this month	Slow editor, still possible	Cannot ship the pages or exit you need
Offer change	Homepage sentence still true	Secondary line outdated	Hero sells a company you no longer run
Trust / photos stale	Assets would survive a sales call	A few old but true stills	Proof would embarrass you in the room

Add them. Interpretation is not a formula I will pretend is science. It is a forcing function so the kickoff cannot skip evidence.

1. Fill the table with artifacts in the cells, not vibes.
2. Circle every 2. That circle is the scope of v1.
3. If conversion is a 2 and tracking is unproven, stop. Prove tracking first (next section).
4. If mobile is a 2, the first viewport and the CTA are in scope even if the founder prefers desktop comps.
5. If CMS lock is a 2, do not buy a theme on the same ceiling.
6. If offer change or stale proof is a 2, content and media are in scope before motion.

A 2 on conversion plus a 2 on mobile is usually one redesign: phone-first fold, one ask, proof that holds. A 2 on lock plus a 0 everywhere else is not a paint pass. A 0 across the board and a “three years” speech is a no.

Write the scores where the client can argue with them. If they cannot argue, you do not have a brief. You have a mood board.

How do you prove tracking before you call it conversion collapse?

A quiet phone is not a fold problem until the measurement is honest. I have seen “conversion collapsed” resolve into: the call tracking number expired, GA4 lost the form event in a tag rebuild, the form still posted to a staging inbox, or Google Business Profile was getting the calls while the site was treated as the only channel.

Prove the ledger and the instrumentation on the same week. Then talk about design.

Source	What it is allowed to claim	What it cannot claim alone
Business ledger (call log, CRM, jobs, Stripe)	Whether money-making actions happened	Which page caused them
Google Analytics 4	Sessions, events you actually tested	“Leads” if the event never fired on a phone
Google Search Console	Queries, index, Core Web Vitals, mobile usability	Revenue
Google Business Profile insights	Calls and direction requests from the listing	That the website is the leak
Inbox / calendar	Whether the form or book button reached a human	Traffic quality

Procedure:

1. Submit the site form from a personal phone on cellular. Watch the inbox and the spam folder. Confirm the thank-you URL.
2. Tap the header phone. Confirm it dials. Confirm it matches the listing.
3. In GA4, complete the primary action once. Confirm the event and the timestamp.
4. Pull 90 days of sessions vs 90 days of ledger actions, mobile split.
5. Pull the same window for Google Business Profile calls if you are local.

If GBP calls are healthy and site forms are dead, you may not have a “website conversion” crisis. You may have a site that was never the call-getting surface. Sequence the owned site for proof, services, and fallback — that split is Google Business Profile vs website. Redesign the site when the site’s job (proof, depth, a URL Google cannot suspend) is failing, not because the map pack is doing its job.

If sessions fell with the actions, hunt demand and index first. If events are missing, hunt tags first. If the ledger fell and sessions held and the thumb test fails, now you have conversion collapse as a redesign trigger.

Do not brief a designer on a number nobody can reproduce. That is how you buy cinema for a broken webhook.

Failure mode: redesigning because the founder is bored

The failure is specific. The site still books work. The editor still publishes. The offer is true. The photos are honest. Mobile completes the ask. The founder has seen a reel on Instagram and wants “a new site.” The agency agrees because the anniversary invoice is easy.

What breaks:

• URLs that had search equity get invented slugs “while we are in there.”
• The phone number moves behind a form because the new theme’s hero has a default “get a quote.”
• Tracking is reinstalled wrong. The post-launch report says conversion died. Nobody can prove whether the site failed or the tag did.
• Photos get replaced with generated atmosphere that does not match the truck.
• The CMS the staff knew is swapped for a prettier one they will not open.

What it costs: weeks of calendar, a confused Google index, a week of missed calls if the tel: link dies, and a staff that now emails the designer for every comma. What you do instead: run the five-trigger score. If it is zeros, buy photography and a copy pass. Keep the engine.

Bored-founder ask	Translate	Honest response
“Make it feel more premium”	Not a job	Name fold, proof, CTA, or stack — or refuse
“Everyone else has video in the hero”	Fashion	Still + optional play; autoplay as LCP is a mobile fail waiting to happen
“Can we redo the URLs too?”	Second project	Not in the same launch
“We’ll know it when we see it”	No brief	Write the 90-day offer sentence first
“Just refresh it but also move hosts”	Mixed launch	That is how tracking and DNS die together

I will not name a fake lift that “rebrands” produce. I will name the failure: you spent a premiere budget on boredom and imported new ways to miss the phone.

If you already know you need work and you are arguing paint vs engine vs new factory, that is a scope question. This spoke stays on when. Boredom is not a when.

How does this affect conversion?

Conversion moves when you shorten the distance between recognition and action, on the device people actually use, with proof that survives a skeptic, for the offer you actually sell. A redesign affects conversion only through those four. Theme fashion does not.

Lever	If it is the leak	If it is not
Recognition (fold + brand)	Stranger cannot repeat the offer	They can; they still do not act
Action (path)	Extra taps, tiny targets, form novel, dead tel:	Path is short; demand is the issue
Proof	Stale or missing evidence	Proof is true; the ask is weak
Offer truth	Site sells a past company	Copy is current; traffic is the wrong query

A redesign that only changes type and color, while the phone stays in the footer and the hero stays a slider, should not be expected to move the ledger. I will not quote a made-up “2x conversion” for a new palette. Measure the lever you changed.

What to measure after a trigger-driven redesign — same window length, same device split, after tracking is proven:

• Primary actions in the ledger (not vanity events)
• Mobile completion of that action
• Form delivery rate (submits that reach a human)
• Search Console: queries you care about still land on the same useful URLs
• Core Web Vitals on phones, 75th percentile, not a single lab run on a laptop

Give it enough time to see a pattern. A week of noise after launch is not a verdict. A quarter with a broken event is not a design indictment. Tie the metric to the trigger you claimed. If you redesigned because of mobile fail, the scoreboard is mobile completions. If you redesigned because of stale proof, the scoreboard is whether sales calls still trip on “that person does not work here.”

Conversion is not a vibe. It is a named action on a named device against a named baseline.

What should you ask a designer about this?

Ask them to diagnose against the five triggers, not to sell you a reel. A designer who cannot tell you which trigger they are being hired to remove will sell you a look.

Question	You want	Walk if you hear
Which trigger is in scope?	One or two from the table, with artifacts	“We’ll modernize the brand online”
What stays (URLs, CMS, phone path)?	A written keep-list	“We’ll clean that up while we’re in there”
How will mobile be tested?	Named phones, cellular, thumb path	“It is responsive” with no device
Who publishes next Tuesday?	Live editor demo in the sales process	“We’ll train you at launch”
What happens to tracking?	GA4 events and form tests in the launch checklist	“SEO is included” with no property
How do we know it worked?	Metrics tied to the trigger	“You’ll love how it feels”

Bring this packet to the call:

• 90-day sessions vs ledger actions, mobile split
• Screenshot of the current first viewport on a phone
• One failed CMS task, or a note that the editor works
• Current offer sentence vs homepage headline
• Proof inventory: people, fleet, work, credentials, reviews
• Google Business Profile vs site NAP check
• Whose names are on domain, hosting, and analytics

If they start picking typefaces before they can fill that list, they are decorating. Decorating is not the job when a trigger is on fire.

Ask who owns the accounts. A pretty redesign on a freelancer’s Webflow workspace is a future lock. Ownership is a trigger you can create by accident.

When is a custom site worth it for this?

Custom is worth it when the current stack cannot host the trigger you have to remove. It is not worth it because custom sounds expensive, and expensive sounds serious.

Trigger	Custom (or a real authored system on Webflow / Framer) is worth it when	A constrained template can still work when
Conversion collapse	You need a fold and path the theme cannot compose	You can kill the slider, write one ask, and keep URLs
Mobile fail	The theme’s header, chat, and sticky bars cannot be tamed	You can ship a phone-first fold inside the tool
CMS lock	You cannot add the content model or you cannot leave	The editor works; you only need fields locked down
Offer change	New jobs need templates the theme does not have (tour, fleet, shop, areas)	The new offer fits existing page recipes
Trust / photos stale	You need a proof system with enforced crops and required alt	You can replace media in current slots

“Custom” here means an authored brand system — compositions, type, motion, CMS rules as one production — not custom CSS sprinkled on a rented card grid. Framer, Webflow, or a custom Astro shell can all carry that bar. The tool is secondary. Whether you can remove the trigger is the product.

Worth-it test:

1. Name the trigger with a 2 on the score table.
2. Attempt the smallest fix inside the current CMS for a week (fold copy, CTA, photos, form).
3. If the tool blocks the fix, you have lock. Now stack choice matters.
4. If the tool allows the fix and the ledger still dies, you have a demand or offer problem, not a custom-site problem.
5. If you need a unique path (listen-first artist home, phone-first trade, proof-heavy studio) and every theme fights you, authored is the spend.

Hundreds of production sites later, the pattern is boring: the stack that ships, stays editable, and can take the LCP hit of a real photo beats the stack that won a screenshot. Custom is the right answer when the screenshot stack cannot pass the thumb test or the editor test. It is the wrong answer when the trigger was stale photos and the CMS already had image fields.

What should you skip if you only have a week?

A week is a diagnostic and a stop-the-bleeding pass. It is not a redesign. Do not pretend otherwise. Skip cinema. Skip a new CMS. Skip URL invention. Skip a blog. Skip service-area clones. Skip autoplay.

Do this, in order, and write down which trigger remains:

Day	Do	Skip
1	Thumb test, form test, tel: test, NAP vs GBP	Mood boards
2	Sessions vs ledger, 90 days, mobile split; GA4 event proof	“Analytics cleanup” that changes every event name
3	Homepage sentence vs live offer; kill nav lies	New service pages
4	Proof inventory; pull down faces and trucks that are gone	Generated hero atmospheres
5	One fold: headline, support, one CTA, one true still	Motion, sliders, chat widgets
6	CMS: can a non-designer publish that fold?	Platform migration
7	Score the five triggers; decide redesign / wait / different job	A Friday domain cutover

Checklist for the week:

• Phone number works and matches the listing
• Form reaches a human
• First screen tells the truth about the offer
• Obvious stale proof is removed, even if the hole looks empty
• Chat overlay is off the primary CTA
• Hero video is not the only LCP candidate
• Scores written; next spend named

If the week reveals conversion collapse plus mobile fail and the stack can host a new fold, you have a redesign brief. If it reveals lock, stop decorating. If it reveals zeros, you just saved a five-figure premiere. That is a successful week.

Skip anything that changes look, URLs, CMS, and tracking in the same seven days. You will not know what worked. You will only know that something is on fire.

When is this not worth doing yet?

A redesign is not worth doing yet when the five triggers are zeros, when tracking is unproven, when the offer is still being argued inside the company, or when Google Business Profile is the actual call engine and the site has no named job beyond “we should have a website.”

It is also not worth doing yet when you cannot name an owner for domain, hosting, and the inbox the form hits. You will launch into a void.

Situation	Redesign now?	Do this instead
Scores are zeros; founder is bored	No	Photos, copy, wait
Tracking unproven	No	Form, tel:, GA4, ledger — then re-score
Offer still in a fight between partners	No	One 90-day sentence in writing
GBP gets the calls; site is a brochure nobody uses	Not until the site has a job	Sequence listing vs owned proof; give the site a job or pause it
CMS lock is real but cash is tight	Maybe a fold sprint on the current stack; not a vanity custom	Unlock editor, or save for an authored build
Seasonal dip every winter	No, if it matches last year	Compare like-for-like windows
You cannot staff replies to new leads	No	Capacity first; a louder fold will drown you

Not worth doing yet is a legitimate outcome. The point of a trigger list is to protect you from anniversary shopping. If the site is true, tappable, editable, and the ledger is fine, keep it. Put the money into the work the site is supposed to sell.

When a trigger is a 2, waiting is also a decision. A buried phone number does not become less buried in Q4. Stale staff photos do not become true. A lying hero does not age into honesty. Delay has a cost. Just do not confuse delay-cost with calendar-cost. Three years is not a fee. A missed weekend call is.

FAQ

When should a small business redesign their website?

Redesign when an operating trigger fires: conversion collapse with sessions still arriving, a mobile path that cannot finish the job, a CMS that locks the pages or the exit you need, an offer the current site no longer sells, or trust assets that would embarrass you in a sales call. Do not redesign because the site turned three. Inspect on a schedule; buy on a trigger.

How do I measure whether a small-business website redesign is working?

Measure the named action the trigger was about, in the business ledger, on the same device split, against a baseline you recorded before launch. If you redesigned for mobile fail, watch mobile completions; if you redesigned for stale proof, watch whether sales calls still trip on dead staff and dead trucks. GA4 events only count after you have proven they fire. I will not hand you a universal lift percentage.

What usually fails first when teams try this?

Tracking and the phone path. The new theme drops the tel: link, the form posts to staging, or GA4 events are renamed so the report looks like a crash. The second failure is mixing URL changes into the visual launch so nobody can tell what broke. Prove the ledger, keep the slugs unless a page is truly retired, and test the thumb path on cellular before you announce.

How long does this take to show results?

Stop-the-bleeding items (dead phone, dead form, lying hero) show up as soon as a human can complete the job again — often immediately after you ship those fixes. Fold and path changes need enough traffic to read a pattern; a quiet week is not a verdict. Search-visible URL changes take longer because Google has to recrawl. If you cannot see the primary action in the ledger within a normal business cycle after a proven launch, you either shipped the wrong trigger or you are still measuring theater.

What should I skip if I only have a week?

Skip a new CMS, new URLs, autoplay heroes, chat widgets, and a twenty-page sitemap. Spend the week on the thumb test, form and tel: proof, sessions vs ledger, offer truth, and pulling down stale proof. Write the five-trigger scores on day seven. A week is how you learn whether a redesign is the next spend, not how you fake one.

When is this not worth doing yet?

When the five triggers score zero, when tracking is unproven, when the company cannot agree on the 90-day offer, or when you cannot name who owns the domain, the host, and the inbox. Also wait if the site has no job and Google Business Profile is already getting the calls — give the site a job first, or pause it. A redesign without a trigger is an anniversary purchase.

CTA

If a trigger is a 2, stop waiting on the calendar. Name the leak, keep the URLs unless they are dead, and buy the smallest job that removes it.

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