Most website projects end at launch. The agency hands over, the team celebrates, and attention moves to the next thing. Which is a shame, because launch is the first day the site produces real evidence about whether it works. Everything before that was an educated guess.
The research on this is old and consistent. Nielsen Norman Group measured usability improvements of about 38% per design iteration, with a median improvement of 165% from a first version to a final one across their case studies. The gains come from iterating on what you have learned, not from starting again. This is how to set that up.
Iterate, optimise, redesign: what the words mean
Iteration is a deliberate repeat of a design step to improve it: you observe a problem, change one thing, and check whether the change fixed it. Optimisation is iteration aimed at a number, usually conversion or speed. A redesign replaces the structure. The first two compound. The third resets the clock, and we have written before about how to tell when it is actually warranted. Post-launch work should live almost entirely in the first two.
Week one: check that it works
Nothing here is about improving the site. It is about making sure the instruments are switched on, because a month of missing data cannot be recovered.
- Analytics is firing on every page, and the two or three actions that matter, an enquiry sent, a sign-up completed, a call booked, are recorded as events. Test each one yourself and watch it arrive.
- Search Console is verified, the sitemap is submitted, and the home page has been requested for indexing. Check that old URLs redirect to their new homes rather than to a 404.
- Forms deliver. Submit each one and confirm the email lands with the right person, not in a spam folder or a departed colleague's inbox.
- Speed is measured in the field, not just in a lab. Open PageSpeed Insights and note whether real-user data exists yet; for a new domain it takes a few weeks to appear.
- An accessibility scan is clean of the six common errors. Launch week is the cheapest moment to fix them.
Month one: decide what you are measuring
Pick three business metrics and write them down with their starting values. Not pageviews. The things the site exists to produce: qualified enquiries, trial sign-ups, booked demos, applications. If you cannot name three, the site does not have a job yet, and that is the first problem to fix.
Alongside those, track three health metrics Google publishes for every site with enough traffic: the Core Web Vitals. Largest Contentful Paint should be under 2.5 seconds, Interaction to Next Paint under 200 milliseconds, and Cumulative Layout Shift under 0.1, each measured at the 75th percentile of real visits over a rolling 28 days. That last clause is the important one. A site that feels instant on your laptop can fail for the quarter of visitors on mid-range phones, and it is their experience that counts.
Passing is worth the effort. In Vodafone Italy's published A/B test, two otherwise identical pages differed only in loading speed; a 31% improvement in Largest Contentful Paint produced 8% more sales. Speed is a conversion variable, not a technical footnote.
A launch is a hypothesis with a budget behind it. The month after is when you find out whether it was right.
Month two onward: the iteration loop
- 1Find where people fail. Look at the funnel between landing and the action you care about. Where do most visitors stop? Session recordings and search queries in Search Console will tell you what they came for and where they gave up.
- 2Watch five people. A five-person usability test on the live site takes a day and explains the analytics. Numbers tell you where; people tell you why.
- 3Keep a queue, ordered by evidence. Every proposed change goes in with the problem it solves and how you will know it worked. The top of the list is whatever removes the most friction for the least effort.
- 4Ship one change at a time. Change the headline and the form and the pricing in the same week and you will never know which one moved the number. Small, sequential, measured.
- 5Wait for the window. Four weeks for Core Web Vitals to reflect a fix. Enough visits for a conversion change to be more than noise. Formal A/B testing needs traffic most sites do not have; a before-and-after over matched periods is honest enough for the rest.
- 6Write down what you learned. Each result, good or bad, becomes the brief for the next change. After six months you have something no redesign can buy: a record of what your audience actually responds to.
What not to do
- Do not redesign at ninety days because the numbers are flat. Flat numbers are a diagnosis problem. Find the failing step first.
- Do not chase a Lighthouse score. It is a lab number from a single simulated run. Google ranks on field data from real visitors, and the two often disagree.
- Do not add tools to fix a lack of insight. A heatmap, a chat widget and two more analytics scripts will slow the site and tell you the same thing five sessions with real users would have.
- Do not treat search visibility as somebody else's job after launch. Titles, descriptions and page structure are design decisions, and the first three months of Search Console data show exactly which ones landed.
We stay on after launch for exactly this reason. A site is not finished when it goes live; it is finished when it has been measured, adjusted and measured again, and nobody can find the next thing worth changing. If your site launched and then went quiet, tell us what it was meant to do. Working out why it is not doing it is usually a short job.
Common questions
How long after launch should we wait before judging results?
Four weeks minimum for anything measured from real visitors. Google’s Core Web Vitals data is a rolling 28-day window, seasonal traffic patterns need a full cycle to show, and small sites need time to collect enough visits for a number to mean anything. Check that things work in week one. Judge how well they work in month two.
How often should a website be updated after launch?
Continuously, in small increments, rather than in one big overhaul every few years. Ship one change at a time, measure it, and keep a queue ordered by evidence. Sites that iterate monthly tend to compound gains; sites that redesign every three years reset to zero and lose what they learned.
What metrics should we track after a website launch?
Three business metrics tied to what the site is for, such as enquiries, sign-ups or qualified demo requests, plus the three Core Web Vitals for health, plus Search Console impressions and clicks for visibility. Anything beyond that is optional. Pageviews, bounce rate and session duration are not goals; they are diagnostics.