What Goes into Building Your Website

I published my prices. They’re on the pricing page, in real dollars, next to what the other guys charge. Do that and one question comes back fast: why does a custom site cost $1,495 to $7,500 when Wix says a website costs $30 a month?

Fair question. The $30 buys you software that lets you assemble a page. The build fee pays for everything else, and everything else is the part that decides whether anyone ever finds you. Here is what that work looks like on a real build, start to finish.

Before you ever write a line of code

The first week of a build has no code in it at all. It has research. Skip this part and you get a pretty site that ranks nowhere, which is what most template sites are.

What the site is actually for

Before a page gets drawn, the question is what the site is for. Booking appointments, capturing leads, selling a product, running a directory, or educating prospects until they are ready to call. Each one needs a different structure. A booking site lives or dies by the calendar flow. A directory needs filterable search and a unique page per entry. A service business needs trust signals and a short path from the home page to “book a call.” If your developer never asks, you get a generic five-page brochure that does none of it.

Competitor research

This summer I built a site for a family lawyer in Toronto. The first thing I did was search the way his clients search: divorce lawyer Toronto, child support lawyer, separation agreement Ontario. Then I went through the firms on page 1. What pages do they have? What do they publish? Where are they thin? You can’t outrank a field you haven’t studied.

Keyword planning

A keyword is a person with a problem. Someone typing “how is child support calculated in Ontario” is days away from calling a lawyer. I build a list of those searches, sort them by intent, and assign each one to a page. Every page on the site exists to answer a specific search. No page, no ranking, no call.

What are the people on page 1 doing to get there?

Page 1 is not luck. The firms at the top had depth: real reviews, guides, and working tools. So the plan for the family-law site included a child support calculator built on the official 2025 Federal Child Support Tables. I parsed the government’s own PDF into 1,341 rows of data and verified the results against the published figures. A drag-and-drop template will never give you that, and that is exactly the kind of page that earns links and repeat visits.

The SEO plan and the AI plan

There are two audiences now. Search engine crawlers, and AI assistants like ChatGPT and Perplexity that people ask “who’s a good family lawyer near me?” Ranking in one doesn’t guarantee the other. So the plan covers both: structured data that machines can read, an llms.txt file that tells AI models what the site is, clean question-shaped headings, and answers stated plainly enough to be quoted.

Design, colour, vibe

The lawyer already had a brand at his main firm. I pulled his navy and blue directly from that site, matched the serif headings, and carried the identity over so a client moving between his sites never wonders if they’re in the right place. Design isn’t decoration. It’s the ten-second trust decision every visitor makes before reading a word.

Now it’s time to code

Research done, plan on paper. Now the part people picture when they think “web developer.”

Planning the pages

Every page gets a job before it gets a design. The home page converts. Service pages rank. Tool pages earn links. The contact page removes friction. If a page has no job, it doesn’t get built.

Architecture beats layout

The shape of a site matters more than its colour palette: where the sitemap branches, what links to what, how deep a page sits from the home page, and whether the same content has one canonical address or seventeen accidental duplicates. Get it wrong and Google indexes the slow page, the wrong page, or none of them. Get it right and a small site outranks a sprawling one. Drag-and-drop builders produce some of the worst architecture on the internet: flat, duplicate-heavy, and often serving the actual content from JavaScript that Google has to render before it can read a word.

Curating content and generating images

For the family-law build I pulled the lawyer’s real Google reviews into the site as trust signals. Real words from real clients beat stock photos of handshakes every time. Images get generated where needed, then compressed and sized so they load fast on a phone in a parking lot, because that’s where your clients actually are.

Writing with SEO and AI in mind

Headings phrased as the questions people type. The answer in the first paragraph, not the fifth. Schema markup underneath so machines understand what the page is about. This is the difference between content that reads well and content that reads well and gets found.

Copy that is persuasive, professional, and built to rank

For a regulated professional there’s a third constraint: the regulator. Law societies have advertising rules, and a website is a regulated communication. I spent 33 years as an Ontario paralegal, so the copy gets written to persuade, to rank, and to survive a compliance read. Most developers only manage the first one.

Running content through review and grading

Every page runs through my audit system and gets a score out of 100. The family-law site first graded at 87. The client asked what it would take to go higher. I fixed what the audit flagged and it launched at 90. You can’t improve what you don’t measure, so I measure.

Testing the website

Making sure it works

Click every link. Submit every form, then submit them wrong on purpose. Test on a phone, on a tablet, on a slow connection. A contact form that silently fails is a client you’ll never know you lost.

Showing it to the client

The client walks the site and sends a punch list. The Toronto lawyer sent his changes on a Sunday; they were live Monday morning. Revision rounds are part of the build, priced in from the start.

Getting ready to launch

Final pass: meta titles and descriptions on every page, social share images, redirects mapped from any old site, a proper 404 page. Boring work. Boring is good. Boring stays up.

Deploying it

DNS records

The domain gets registered where you can see it, with you named on it, and the records pointed at the new server. This matters more than people think, and it’s why my FAQ tells you to check who the registrant is on your own domain.

Email that actually arrives

A website without working email is a billboard without a phone number. Most clients belong on Google Workspace at about $6 per user per month, or Microsoft 365 at about $7 if they already live in Outlook. Either way, SPF and DKIM get configured at DNS level so your outgoing mail is not flagged as spam. Self-hosting mail is a maintenance trap: SPF, DKIM, DMARC and reverse DNS all have to be perfect, IP reputation has to be earned, and one mistake puts your business email in spam permanently. Those records get set at the same time the site is deployed. If your developer never asks which mail platform you are on, that is a flag.

The server

No shared hosting with a thousand strangers. The site runs on a dedicated server with its own resources, monitored and backed up nightly. When a client asked me where his site lived and what it cost to keep it there, I sent him the details the same day. That’s your answer test for any provider.

The SSL certificate

The padlock. Browsers shame sites without it and search engines demote them. The certificate gets installed and set to renew automatically, so it never lapses at 2 a.m. on a long weekend.

SEO

The site is live. Search engines don’t know that yet, and they don’t owe you a visit.

Registering the site with Google, Bing and Yandex

I register the site with Google Search Console, Bing Webmaster Tools, and Yandex on launch day, and submit the sitemap to each. AI assistants lean heavily on Bing’s index, so skipping Bing means skipping a growing share of how people find you.

A proper sitemap

A sitemap with honest last-modified dates on every page. When the dates are truthful, crawlers learn to trust the sitemap and come back when something actually changes. When they’re faked, crawlers learn to ignore it.

Connecting analytics

You should know which pages pull people in and which page they were on when they decided to call. Analytics goes in at launch, not as an afterthought, so the data starts on day one.

Running the site audit

A full audit at launch, then monthly. The family-law site had 56 pages in Bing’s index within the first month, and the audit landed in the client’s inbox so he could see it himself. You shouldn’t have to take your developer’s word for anything.

Social preview cards

When someone shares your URL on Facebook, LinkedIn, Slack, iMessage or X, the preview card comes from your Open Graph tags. Missing or generic tags produce a blank or wrong-looking card, and the link earns a fraction of the clicks it should. Every page gets og:title, og:description, an og:image at 1200x630, a twitter:card, and JSON-LD underneath, with a custom image per article. Look at the portfolio: every card on that page is pulled from each site’s real og:image tag, because I set one on every site.

Why the lowest bidder costs more

You already know the cheapest quote is not writing the best copy or doing the keyword research. What is less obvious is what happens after they ship. They disappear when something breaks. They keep you on their hosting so you cannot leave. They build on a proprietary builder so the site cannot be ported anywhere else. They never add to it, never update it, never measure anything. Six months later you are back at the start with a slightly newer flyer.

That is not a one-time sunk cost, it is recurring damage. Every month a bad site stays up is a month of prospects searching for what you do, finding someone else, and never learning you existed.

It’s now launched

Launch day is the start line. Rankings compound from what happens after: publishing, updates, and the monthly grind that turns a new site into a page-1 site. That’s the ongoing work my care plans cover, and it’s a story about money, which deserves its own article: What does your website really cost?

Already have a website? Request a free audit and I’ll grade it the same way I grade my own work, score out of 100, no charge, no obligation.

Common questions

Why does a custom website cost more than Wix or Squarespace?

The $30 a month buys software that lets you assemble a page. The build fee pays for everything that decides whether anyone finds the page: competitor research, a keyword plan, the content, the architecture, testing, deployment and search engine registration. You are paying for the findability, not the pixels.

How long does it take to build a website?

For a custom small-business site, four to eight weeks is realistic, and the first week has no code in it at all. That week is research: competitors, keywords, and what each page is for. Rushing that part is how you get a good-looking site that ranks nowhere.

How is a website tested before it goes live?

On real devices and real connections, not just a desktop browser window. Every form submitted, every link followed, every page checked on a phone over cellular. Then the technical pass: page speed, layout stability, accessibility, and whether the content is in the HTML where crawlers can actually read it.

What does deploying a website actually involve?

Putting the code on a server you control, pointing the domain at it, installing a certificate so the site loads over HTTPS, configuring the web server, and setting up backups before anyone visits. Done properly it is invisible. Done badly it is why a site is slow, insecure, or occasionally offline.

Is a website finished when it launches?

No. Launch is when a new site starts earning its position, and new sites start with none. Rankings arrive over months as pages get indexed, links accumulate and content gets published. A site that launches and is never touched again peaks in its first year and declines from there.