You Just Launched a Website. Here's What to Do in the First 90 Days

A new website is a strange thing to own. You have spent weeks on the wording, argued about the photographs, approved it, and then it goes live and almost nothing happens. Eleven visitors on day one, most of them your mum and the person who built it. Four on day two. By the end of the first fortnight you are refreshing a dashboard you do not fully trust and wondering whether the whole thing was a mistake.

That quiet stretch is where the bad decisions get made: a redesign in week three, a hurried ad spend in week five, or the more common outcome, never looking at the site again until something breaks. What follows is a plan for the first 90 days that takes roughly an hour a week. The order matters more than the speed. Measurement comes first, because every decision after it depends on knowing whether the last thing worked.

Before anything else, confirm that your analytics are working on every page, not just the home page. Analytics usually run on a tracking snippet — a short piece of code your provider gives you that sits on each page and records a visit. If that code is missing from your contact page, your booking page, or your blog posts, those visits simply do not exist as far as your numbers are concerned, and you will spend three months drawing conclusions from a partial picture.

The test takes ten minutes. Turn the wi-fi off on your phone so you arrive as an ordinary visitor on mobile data, then walk through your own site the way a customer would: home page, services, a blog post, the contact form, and the thank-you page people see after submitting. Keep a live view of your traffic open on a laptop while you do it. Every one of those pages should appear. Our guide to verifying an install runs through this properly, including what to check if a page is missing.

The commonest failure is a site where most pages were built from one template and two or three pages — often the newest ones, or ones added by a different person — were built another way and never got the code. The second commonest is your own traffic drowning everything else. If you visit your site eight times a day to check it, exclude your own connection so the numbers reflect customers rather than you. Finish the week by setting that exclusion up and then leaving the dashboard alone for a few days.

Visitor counts on their own are close to useless in month one, because the number is small and it bounces around for reasons you cannot control. What you need is a definition of success specific to your business, recorded automatically, so that in week twelve you can say "we had 47 enquiries" rather than "traffic feels better, I think".

Pick one to three things, no more. For a solicitor it might be a completed enquiry form and a tap on the phone number. For a bakery, a tap on the phone number, a click on the directions link, and a download of the celebration cake price list. For a freelance designer, a form submission and a click on the email address. These are called goals: a specific action you want visitors to take, counted every time it happens. Setting up goals is a five-minute job and it is the single highest-value thing you will do all quarter.

The ones people forget are the off-site actions. A phone tap, a WhatsApp click, a map link, a PDF menu — these are often the real conversions for a local business and they leave no trace unless you ask for them to be recorded. Those are custom events: a thing you nominate as worth counting. A plumber whose customers almost all ring rather than fill in a form will otherwise conclude, wrongly, that the website does nothing.

A new site has no reputation, which in practice means search engines have no particular reason to show it to anyone. The fastest fix is not clever writing, it is being referenced from places that already exist. Start with a Google Business Profile if you serve a local area — for a bakery or a plumber this will often produce more enquiries in the first quarter than the website itself does.

Then connect Search Console, Google's free tool that tells you which search terms showed your pages and how often anyone clicked. Linking it to your analytics means the search data sits alongside the visit data instead of in a separate tab you forget about; here is how the Search Console connection works. Expect it to show almost nothing for the first month. That is not a fault, it is a new site.

After that, make a short list of places that already mention your business or should: your professional body, a supplier whose products you fit, the local chamber, the trade association directory, the venue you work with, a former client's site. Email them and ask for a link to your new address. Six genuine mentions from relevant places are worth more than sixty entries in directories nobody reads. Be honest with yourself about which is which — if you have never clicked a directory as a customer, your customers probably have not either.

This is the month that does the most long-term work. Open your phone, your inbox, and your memory of the last twenty enquiries, and write down the questions people genuinely asked before they bought. Not what you wish they asked. The real ones: how long does a boiler replacement take, do you do gluten-free, what happens if the other side refuses to respond, can you work from my rough sketches, do you need a deposit.

Give each question its own page with the question as the heading, answered plainly, in four hundred to seven hundred words. No preamble about how we are passionate about quality. Answer it in the first two sentences and then explain the exceptions. Publish one a week for four weeks. These pages tend to be the ones that quietly accumulate search traffic over six and twelve months, and they are also what AI assistants read when someone asks a question in your area and the assistant goes looking for a straight answer.

There is a secondary benefit that shows up quickly. When somebody rings and asks the same question for the ninth time, you have a page to send them. More than once I have seen owners realise, writing these, that their services page never actually said what they charge for or where they travel to.

By now you have eight or ten weeks of data and a handful of new pages, and the site has had enough time to develop the usual small faults: a link to a page that was renamed, an image that makes one page take six seconds to load on a phone, a page nobody can reach because nothing links to it. None of these are dramatic on their own and all of them cost you enquiries.

Run a scan of the whole site rather than clicking around hoping to notice. The free report card checks for broken links, slow pages, missing descriptions and pages with nothing pointing at them, and gives you a ranked list rather than a wall of warnings. Fix the top three. Ignore anything that would take a developer a day and save a visitor half a second.

Then use your goal data to answer one question: of the people who reached your contact page, how many submitted the form? If 140 people reached it and 9 submitted, that is a form worth examining — it may be asking for a postcode and a budget before it has earned the right to. If 140 reached it and 38 submitted, your form is fine and the problem is getting more of the right people to that page in the first place. Those two conclusions lead to completely different work, which is why this step comes after the measurement and not before.

Do this once a month, on a fixed day, rather than daily. Daily numbers on a new site are noise. Look at four things: total visitors this month against last, goals completed this month against last, which three pages got the most visits, and where those visitors came from — search, social, a link from somewhere, or typed directly.

Write down what you changed that month, in the tool itself if you can. Annotations are just dated notes on your traffic chart — "started posting on Instagram", "supplier added our link", "site was down Tuesday morning" — and they turn a chart you cannot interpret into a record you can. In nine months, when you are trying to remember whether that spike in March was the local paper or the ad you tried, the note will be there.

If a weekly nudge suits you better than a monthly diary entry, a weekly report from the AI advisor will read the data for you and send a short ranked list of what changed and what to do about it. The point of either approach is the same: a small, regular look, rather than a panicked deep dive every six months.

Do not redesign. Three months is not long enough to know whether the design is the problem, and a redesign resets whatever small foothold you have in search results. Do not judge the site by bounce rate — the proportion of people who look at one page and leave — because on a site where the home page answers the question, a high figure can be entirely healthy.

Do not buy traffic before the goals are recording properly. Advertising into a site where you cannot see what converts is the most reliable way to waste money in this trade. And do not install five tools. One set of numbers you check monthly beats four dashboards you have logins for and never open.

Set your expectations where the evidence is. A brand new site for a small local business commonly spends the first month in single or low double-digit daily visits, mostly direct and mostly people you told. Search traffic generally starts to show up somewhere between weeks six and twelve, and the question pages you wrote in month two usually do not do their best work until month five or six. This is a rough range from general experience, not a promise — a plumber in a city of 400,000 and a solicitor in a market town will see very different curves.

What you should have at day 90 is not a busy site. It is a site where tracking is verified on every page, two or three goals are recording, you know which pages people actually read, you have a handful of genuine mentions elsewhere on the web, four pages answering real questions, and a monthly habit that takes twenty minutes. That is a working foundation, and almost nobody has one.

If you want to set this up as you go, you can start a free trial, install the snippet in about ten minutes and have the first week's numbers verified before the weekend. Doing it in the first 90 days is considerably easier than reconstructing the year afterwards.