Every platform article I write ends up at the same sentence, so I may as well put it on its own page: the only question I ask about a tool is whether it will let me do the job for the client.

Not whether it is popular. Not whether it is the one I already know. Not whether it makes me look clever at a conference. Whether it lets the work get done properly, for this business, and keeps getting done for years afterward.

That sounds obvious. Almost nobody actually chooses that way.

Two ways to get this wrong

The first is the guy who uses one tool for everything, because it is the one tool he has. Every client gets the same platform whether it fits or not, and the parts that do not fit get bent until they mostly work. You can spot these sites immediately: they are all doing something slightly awkward that nobody would have chosen on purpose.

The second is the shop that writes everything from scratch. Custom checkout. Custom content system. Custom image handling. It sounds thorough, and it is a catastrophe on a five-year horizon — every line of it is a thing that breaks, a thing that needs patching, a thing only they understand, and a reason you can never leave. Some of them know exactly what they are doing when they build it that way.

Both failures have the same shape. The choice was made for the convenience of the vendor, not the outcome of the client.

The rule: build what makes you different, buy everything else

There is a line running through every project, and knowing where it sits is most of the job.

On one side is the part that makes you not interchangeable. Your design. The way your business is actually structured and explained. The specific thing you do that the competitor down the road does not. That part gets built, deliberately, by a person, for you.

On the other side is everything a thousand companies have already solved better than we ever will in a client's budget. Payment processing. Checkout and fraud. Email delivery. Maps. Authentication. Content delivery. Image optimisation. All of it bought, integrated, and never thought about again.

Nobody in the history of commerce has chosen a business because it wrote its own checkout. But people choose businesses every day because the website made them feel like they were dealing with professionals.

So we spend the budget where it changes the outcome, and we refuse to spend it re-solving problems that are already solved. That is not laziness. Rebuilding a solved problem on a client's money is the expensive kind of showing off.

Design is the part that has to be built

If only one thing gets custom attention, it is this one.

Design is the differentiator, and it is the one thing a tool cannot hand you. A template is designed to be used many times — that is its entire function — so by definition it cannot be the thing that makes you distinct. Whatever else is true about a platform, if it will not let us design freely, it is out. That single requirement decides more platform arguments than everything else combined.

It is also the requirement most often quietly traded away, because designing inside the lines of a template is fast and nobody bills less for it.

The back of house has to be excellent too

The other half of the decision is the part nobody demos.

You are going to open the admin every day this business is open. If the tools do not talk to each other, someone on your staff becomes the integration — typing an order into a second system, reconciling two lists that disagree, re-entering a customer who already exists. That is a permanent tax, paid in hours, that never appears on any quote.

So back-of-house tools get judged on two things: whether they are genuinely good at their own job, and whether they integrate with everything else without a person in the middle. A tool that is excellent in isolation and refuses to connect to anything is not a good tool. It is a future problem with a nice interface.

The benchmark that gets quoted at me

If you go looking, you will find data showing hosted builders beating WordPress on Core Web Vitals. A 2024 comparison drawn from Chrome's real-world user data put Squarespace around 58% of visits passing and Wix around 52%, against WordPress at roughly 38%.

Those numbers are real. They also do not mean what people quote them to mean.

That is a platform average, not a comparison of competent builds. A hosted builder enforces a floor: you cannot construct a catastrophically slow Wix site, because you are never allowed near the parts you would have to break to make one. WordPress has no floor at all. The 38% is measuring the entire population — millions of sites assembled by people with no professional help, running twenty-six plugins that all load on every page, last updated in 2019.

I will say the uncomfortable part plainly, as somebody who builds on it: WordPress in amateur hands is more dangerous than any builder on the market. The failure modes are worse, they compound quietly, and there is nobody watching. If the choice is genuinely between a hosted builder and an unmaintained WordPress install, take the builder. It is not close.

But a floor is not the only thing a platform can give you, and the builders buy theirs by taking the ceiling away too. Which is the same conclusion as everything else on this page: the platform is not the variable. The operator is.

So the real choice was never fast against slow. It is fast and looking like everybody else, or unique, optimised, and fast. A builder hands you the first by removing the second from the table. We would rather do the work and keep both.

What that looks like in practice

The stack is not a religion, it is a set of answers that currently win:

  • WordPress where a business genuinely needs deep content, complicated editorial workflows, and total freedom in how things are structured.
  • Static builds at the edge where speed and design matter most. This site is Astro with a Git-backed CMS on Cloudflare — the design is entirely ours, the pages are pre-built, and there is no database to be slow at the worst moment.
  • Shopify for real product businesses, because payments, fraud, inventory and the Shop app should belong to a company with a compliance department — and with its Storefront API we can still build the storefront ourselves rather than accept a theme.
  • Real application development with our partners in Columbus, on the rare occasion the thing being asked for is genuinely software rather than a website.

If something better arrives next year, we will use that instead. Being loyal to a tool is a way of putting your comfort above the client's outcome, and it is how agencies end up defending decisions they made in 2016.

The four questions

You can use these on us, and I would rather you did:

  • Will it let you build the part that makes me different? If the design has to live inside somebody's template, the answer is no.
  • Does it handle its own job without becoming my problem? Payments, security, uptime. If the answer involves me doing maintenance, it is the wrong tool.
  • Does it connect to the rest of what I use? If a person has to sit between two systems retyping things, that person is the integration and you are paying for it forever.
  • When something unusual comes up, can you get at it? Every business eventually needs something the brochure did not cover. That day arrives for everybody.

Why this is the whole argument

I have written about Wix, Squarespace, and Shopify against WooCommerce. Those read like three separate arguments. They are one argument.

For most of our clients we are the outside IT and marketing department, not a vendor who delivered a website once and went quiet. That is the job. And you cannot do that job on a platform that hands nobody the keys, or with a stack that makes the client the glue between four systems, or by dropping a business into a template and calling the difference a design.

Pick tools that let you do the work properly. Build the part that separates the client from everyone else. Buy the rest from people who are better at it than you are.

That is the entire method. Everything else is just applying it — including when it points back at a tool we used ourselves for years.