The SiteRoot team7 min read
There are 42 businesses running on SiteRoot in production today. Thirteen of them are trades, which makes trades the largest single group by a clear margin. The rest spread across health, food, professional services and a long tail of one of a kind businesses that do not fit a category at all.
Forty two is not a big number. It is, however, enough to have been wrong about several things in public, so here is what we actually learned.
Nobody wants a blank canvas
This was the first assumption to go.
The received wisdom in website software is that owners want control: a canvas, a toolbar, freedom. What we saw instead was that a blank canvas is where a project goes to die. The owner opens it, stares at it, decides they will do it properly on the weekend, and the tab is still open six weeks later.
What works is being asked questions. What do you do, who do you do it for, where do you work, what do you want people to ring you about. Five questions about your own business, which you can answer without thinking, and then something finished to react to. Reacting to a real page is easy. Starting one is not.
Almost everything we have changed about the product since has been a version of moving work off the owner and onto the questions.
Speed matters less than we assumed, and more than we assumed
The median build takes 13.1 minutes from the last answer to a finished site. That is the number we watch.
We expected that to be the headline. It is not, quite. Nobody has ever told us their website arrived too slowly. What thirteen minutes actually buys is something less obvious: it is short enough that the owner is still there. They have not made a coffee and lost the thread. They see the site while the conversation about their business is still fresh, which is when they can tell whether it sounds like them.
The flip side is that the tail matters more than the median. A build that occasionally takes far longer is worse than one that always takes twenty minutes, because the owner has no idea whether to keep waiting. Predictable beats fast, and we did not believe that until we had watched it happen.
The photos are the hardest part, by a distance
Not the writing. The writing is difficult and we spend a lot of effort on it, but the photos are worse.
A real photo of your own van, your own shopfront, your own finished work is worth more than any stock image in existence, because the only question in a visitor's head is whether you are a real business near them. Stock photography answers that question badly. A stock tradesman in a stock kitchen reads as a business that could be anywhere or nowhere.
But most owners do not have usable photos. They have twelve blurry job shots in a camera roll with the flash on, taken to show a customer a fitting, not to sell anything. So the honest state of things is this: we pick images that suit your industry and your services, and you can swap any of them, and the single best hour you can spend on your own website is walking around your work with a phone in daylight.
We have not solved that one. We are not going to pretend otherwise.
Owners read their own site in a completely different order
We watch where people look first when their site appears. It is never the hero heading a designer would agonise over.
It is their phone number. Then their business name, checking the spelling and the capitals. Then their suburb. Then, usually, whether the list of services is missing the one that actually pays the bills.
Those four checks are the whole first impression. Everything else is second. It changed how we build: the facts a business is identified by have to be exactly right before anything else is even worth showing, because if the phone number is wrong nothing else on the page is being read.
What we got wrong
A few worth naming.
We built a good version of something and then did not use it ourselves. Every page of our own marketing site was serving the same page title and the same description as the home page, for months. We do that properly for customers on every site we build. We had simply never looked at our own. If you want a lesson in why you should open your own website on a phone tonight, that is it.
We assumed people would ask for changes in product language. They do not. Nobody asks to update a heading. They say "can you make the top bit say we do emergency call outs". The whole editing side of the product got rebuilt around plain requests, because the version that expected the right words was a version only we could use.
We overestimated how much anyone wants to look at options. An early version offered a long gallery of designs. It measurably made things worse: more scrolling, more second guessing, no better outcome. A small number of directions chosen for your industry beats a hundred chosen for nobody.
What is still hard
In the interest of not writing a victory lap.
Making copy sound like the owner rather than like a brochure is the central difficulty of this product, and the failure mode is always the same: confident, pleasant sentences that could be about any business in the country. We have gates that catch the worst of it and it is still the thing we work on most.
Businesses that do not fit a shape are hard. A one person operation that does three unrelated things, or a trade with a retail counter, or anything seasonal. The more usual a business is, the better the first build.
And nothing about any of this is a growth service. We are self serve software: you build the site, you edit it, it stays live, and we watch that it stays up and that its certificate is current. Nobody here is doing your marketing, and a website is not a substitute for one.
The part that has held up
One thing has been true from the first build to the forty second.
The gap for a small business is almost never the website looking good enough. It is the website being current. Prices from two years ago, hours that changed after Easter, a service you stopped offering, a mobile number from your last phone. Every one of those is a small lie your business is telling a stranger at 9pm, and the reason they stay on the page is that changing them used to mean emailing someone and paying for an hour.
Making the change cost nothing and take a minute turns out to matter more than anything we do at build time. That is the thing we would keep if we had to throw the rest away.
