On this page
- The sitemap is a keyword decision that happens before anyone opens a design tool
- The speed numbers we hold ourselves to, in public
- Accessibility is a legal exposure before it is a ranking question
- What this looked like on a two-office medical practice
- Schema has to describe what a human can actually see
- A form nobody answers is a broken form
- Sites rot, and the plan for that is part of the build
Most small business sites I get called about are on their second or third rebuild. Not because the design aged badly. Because the first one was a brochure with a phone number on it, and a brochure cannot be retrofitted into something that earns traffic.
I have been building and buying this work since 2011. The pattern is consistent. A business pays for a site, likes how it looks, waits a year for the phone to ring differently, then pays a second time for someone to fix the structure underneath. The second build costs more than the first one would have if it had been done in the right order.
So here is what we put into a site before it goes live. Not a wish list. The actual gate.
The sitemap is a keyword decision that happens before anyone opens a design tool
The first artifact on any build we do is not a homepage mockup. It is a page list, and it comes out of search demand.
The rule is one page per thing a person actually searches for and buys. If a plumbing company sells water heater replacement, that is a page. If it also sells repiping, that is a second page, not a bullet on the first one. Google cannot rank a bullet. A single "Services" page listing nine offerings will lose to nine competitors who each gave their offering a page.
The same rule governs geography, and it is where most sites overreach. City pages exist only for markets the business genuinely serves and can drive to. Our own locations rollout is 15 pages against a backlog of 90, because pages for cities you do not serve are the exact structure Google's spam policies describe as doorway abuse.
Three things fall out of getting this right at the sitemap stage. Navigation matches how buyers think instead of how the org chart is drawn. Internal linking has somewhere real to point. And nobody has to renegotiate the information architecture after the design is approved, which is the change that quietly doubles a build budget.
The speed numbers we hold ourselves to, in public
Core Web Vitals are three published thresholds, measured at the 75th percentile of real page loads, split by mobile and desktop. We hold every build to them before launch, not after a client complains:
| Metric | Target | |---|---| | Largest Contentful Paint | 2.5 seconds or less | | Interaction to Next Paint | 200 milliseconds or less | | Cumulative Layout Shift | 0.1 or less |
Hitting them is mostly a discipline of subtraction. Layout shift is caused by images without dimensions, fonts that swap late, and ad or embed slots that reserve no space. Paint time is killed by hero video, a carousel nobody scrolls past slide two of, and six font weights when two will do.
What we will not build: a homepage slider. It loads the heaviest assets on the page before anything else, it pushes the actual offer below the fold, and every client who asks for one is really asking for a way to avoid choosing a primary message. Choosing is the job.
There is a trade-off here worth being honest about. Speed is a threshold, not a lever. Once a site is comfortably inside those numbers, making it faster does very little for rankings. Teams that spend a further month shaving 300 milliseconds are optimizing the thing they can measure instead of the thing that matters, which is usually the content.
Accessibility is a legal exposure before it is a ranking question
We build to WCAG 2.1 Level AA. That is the W3C Recommendation, most recently republished on 6 May 2025, with three conformance levels defined as A, AA and AAA. Level AA is the practical bar, and it includes a text contrast ratio of at least 4.5 to 1 for normal text.
The reason it is not optional for a business open to the public is on ADA.gov. The Justice Department's guidance states that "Title III prohibits discrimination against people with disabilities by businesses open to the public" and that the Department "has consistently taken the position that the ADA's requirements apply to all the goods, services, privileges, or activities offered by public accommodations, including those offered on the web." Businesses currently choose how to get there, and the same guidance points to WCAG as helpful for doing so.
That exposure is not limited to large companies. A public accommodation is a public accommodation with four locations or four hundred, which is why we build to the standard by default instead of quoting it as an upgrade. The accessible version costs almost nothing extra when it goes in at the start: real heading order, labels on form fields, alt text that says something, keyboard focus you can see, and contrast that passes.
The ranking-adjacent part is a side effect. Semantic headings, descriptive link text and captioned media are the same signals a crawler uses to understand a page. You do not get a bonus for accessibility. You avoid the penalty of being illegible to both a screen reader and a parser.
What this looked like on a two-office medical practice
A specialty practice and its sister surgery center in the Inland Empire signed with us on 10 July 2026 for two sites at once: a full redesign of the practice site, roughly 27 pages, and a new site for the surgery center.
The constraints made every decision above concrete.
Two offices, one in Corona and one in Lake Forest, which means two location pages with genuinely different content, two profiles, and referral paths that differ by market rather than a duplicated address block.
Three languages: English, Spanish and Vietnamese, chosen from who actually walks through the door. Multilingual is where cheap builds break, because translation gets bolted on as a plugin that machine-renders the page and produces three thin versions of every URL. We built the language structure into the page model instead.
WCAG 2.1 AA across both sites, for a patient population that skews older and includes people with low vision. In healthcare this stops being an abstract compliance line very quickly.
HIPAA-aware forms, which mostly means being careful about what a public form is allowed to collect and where the submission goes. A contact form that asks about symptoms and drops into a shared inbox is a problem waiting for a complaint.
WordPress with Kadence, so the practice can edit its own pages. That is a deliberate downgrade from a fully custom build, and I will defend it. A site the client cannot update is a site that goes stale in nine months and comes back to us as a change order for a paragraph.
The honest limitation: neither site has launched as of this writing, so there are no ranking results to show you. The SEO retainer and the city page program start after launch. When there are numbers, they go on those two city pages, and not before.
Schema has to describe what a human can actually see
We ship structured data on every build, and we keep it boring on purpose. Google's general structured data guidelines are explicit: "Don't mark up content that is not visible to readers of the page" and "Don't mark up irrelevant or misleading content, such as fake reviews or content unrelated to the focus of a page."
That rules out the most common shortcut in this trade, which is pasting an aggregate review rating into schema that appears nowhere on the site. It is a manual action waiting to happen, and the rich result it buys disappears the moment anyone looks.
Our default set is small. Organization or LocalBusiness with the exact name, phone and address treatment used on the profile and every directory. Service markup on service pages. FAQPage only when the answers are visible on the page and match the markup word for word. If an FAQ accordion and its schema block drift apart during a content edit, that is a bug, and we treat it as one.
A form nobody answers is a broken form
The last mile is where most builds quietly fail. A site can rank, load fast and pass every audit, and still produce nothing, because the lead goes into an inbox that gets checked on Thursdays.
Every site we launch posts its form to an endpoint we control, writes the lead to a CRM record with source and page attribution attached, and sends an email alert at the same time. Call tracking goes on before launch rather than three months later, so first-month numbers mean something. And the standard we hold on our own contact page is the one we ask clients to hold: a human replies the same business day.
Attribution is the part clients underrate. When a lead record carries the page it came from, the city page program stops being a matter of opinion within about 60 days. You can see which markets produce work and which ones produce nothing, and you can stop paying for the second kind.
Sites rot, and the plan for that is part of the build
A launched site is not a finished site. Plugins go out of date, the platform ships breaking changes, images accumulate, a staff change leaves a dead phone number on three pages, and Core Web Vitals drift as content gets added by people who were not in the original build.
We handle that under a care plan: hosting, backups, security patching, uptime, quarterly speed checks against the same three thresholds, and small edits handled by the team that built the site rather than by a ticket queue. Clients who skip it are the ones who call a year later with a broken form or a white screen.
Who we turn away: a business that wants a site and nothing else, with no intention of adding a page or answering a lead for the next two years. We will build it, it will look good, and it will underperform, and then it becomes a case study nobody wants. That work belongs with a template vendor, and I would rather say so on the first call.
One more honest limitation, because it is the thing agencies oversell. Everything on this page makes a site eligible to rank. None of it makes a site rank on its own. Rankings still come from content depth, a real Google Business Profile, reviews, and other sites linking to you. Build order matters because the structural work is expensive to add later and cheap to add now. It is a foundation, not a result.
If your site is due for a rebuild and you want a straight read on whether the structure underneath it is worth keeping, tell us what is not working. Same business day reply, and the answer is sometimes that you should keep the site and spend the money on local SEO instead.
