What happens in the 90 days after a website launches

Launch day feels like the finish line. It is closer to the day a restaurant opens: the build is done, and now you find out what you actually built.

Most of the value in a website project is created or lost in this window, and it is the window nobody plans. The agency moves on, the owner goes back to running the business, and the site sits exactly as launched for three years.

A website is equipment. This is the commissioning period, and equipment that never gets commissioned runs at whatever setting it left the factory on.

What should happen in the first week?

Find the breakage. There is always breakage.

Submit every form on the site from a phone, not a laptop, and confirm the message arrives at the address you expect. Then check the spam folder, because new sending domains land there constantly.

Then find out how fast that address gets read, because a form that works into an inbox nobody watches is a form that does not work. There is real evidence on what the delay costs. Harvard Business Review published an audit of 2,241 US companies in March 2011 in which 23% never responded to a lead at all, and the average response among those who answered within thirty days was 42 hours. The same piece reported a separate analysis of 1.25 million leads across 42 companies: firms that tried to make contact within an hour were nearly seven times as likely to qualify the lead as those that tried an hour later, and more than sixty times as likely as those who waited a day.

Two caveats have to travel with that. Co-author David Elkington founded InsideSales.com, the company whose platform generated the data, and HBR is an editorially reviewed magazine rather than a peer-reviewed journal. Treat it as high quality vendor research published under a strong masthead. The direction is not in dispute even if the multiples are.

Tap the phone number on every page. Tap the address and confirm it opens maps correctly.

Click through the main menu on a phone and look for anything cut off, overlapping, or unreadable. Desktop testing hides an entire class of problem. The mobile conversion problems desktop testing never finds are worth a full pass on a real phone, on cellular, away from your office wifi.

Load ten old addresses from your previous site and confirm each one lands somewhere sensible.

Confirm the site is not blocked from search. A staging block shipped to production is invisible to visitors and fatal to rankings, and it is the single most expensive week-one mistake.

Write every defect down in one list. Scattered feedback across texts and calls is how items get missed.

What should happen in weeks two to four?

Verification and patience, in that order.

Submit the sitemap in Search Console and watch pages get indexed. Some will take longer than others. That is normal.

Confirm conversion tracking fires. Submit a test inquiry and check that it registers as a conversion, not just as an email. Tracking that silently stops working is the most common reason a business cannot tell whether a site helped.

Take a speed reading against a threshold rather than a score. Google rates Largest Contentful Paint as good at 2.5 seconds or below, Interaction to Next Paint at 200 milliseconds or below, and Cumulative Layout Shift at 0.1 or below, each assessed at the 75th percentile of visits, and all three have to pass. A screenshot of a number out of 100 tells you almost nothing. Three pass or fail readings at the 75th percentile tell you what a slow quarter of your visitors experience.

Then leave the design alone. This is the hard part. Rankings commonly move around after a launch and settle over several weeks, and owners who start reversing decisions in week three end up with a site that is half old and half new.

If there are real defects, handle them as a batch through whatever revision process you agreed. Feedback delivered as one clear list gets fixed quickly, and using revision rounds well is the difference between a tidy punch list and a project that never closes.

One thing you can safely ignore in this window is what competitors are doing. If one of them is outranking you with a business name stuffed full of keywords, that is a separate and reportable problem, and what to do about a competitor with a keyword stuffed business name has nothing to do with your launch.

What should happen in month two?

Publish the things that did not make launch.

Almost every project launches with a few pages missing, usually the service pages nobody had time to write. Month two is when they get finished, and it is the difference between a site that grows and a site that was a one-time event.

Add photos of work you have done since launch. Add the questions customers have asked in the last month as content. This is the cheapest form of ongoing improvement available and it requires no vendor.

Fix the small visual items you noticed but did not want to hold the launch for. If the type is a little small or the button color is not doing its job, this is the moment, and the fixes are usually mechanical rather than creative. Choosing colors and typography without a design background is mostly a set of rules, and applying them after launch is perfectly reasonable.

Also confirm your listings still match. Any address, phone number, or hours that changed during the project need to be correct everywhere, not just on the new site.

What does the month three read actually look like?

How should I read a ranking change in month three?

Carefully, and never by converting it into a traffic forecast. This is where the most confident-sounding number in local SEO turns out to be fiction.

Somebody will offer you a table showing what each local pack position is worth in clicks. Position one gets some percentage, position two a bit less, position three less again. Those tables do not survive checking.

Follow the trail. Google exposes no local pack position dimension anywhere. The Google Business Profile performance report publishes Views, Calls, Website clicks, Direction requests, Messages and Bookings, and no rank at all. Search Console has no local pack dimension either. So no one can compute click-through rate by local pack position from platform data, because the platforms do not emit the position.

What is actually circulating is worse than an estimate. One widely quoted figure set is attributed to a BrightLocal local pack CTR study whose URL returns a not-found error and which does not appear in BrightLocal's research index. A second set is attributed to a Moz local CTR study that also does not exist, published by the same aggregator. A third comes from an SEO agency that describes its own method as a meta-analysis with no sample size, citing a list of blogs, one of which now cites it back.

The one traceable study of this shape is a BrightLocal click test from October 2018 that put 5,500 Mechanical Turk testers in front of five San Francisco service-area search results on desktop and recorded the whole local pack taking 32.3% of clicks with no ads present. That is stated preference on a search results layout that no longer exists, from a company that sells local SEO software. It is not a forecasting tool.

So measure actions, not modelled clicks. Take your own profile's calls, website clicks and direction requests for the 90 days before and the 90 days after, and report the delta. Pair it with a rank grid so the ranking movement is documented separately. Two series, correlated, described as correlated. Any vendor who hands you a click forecast from a position change is quoting a number that does not exist.

Then notice which promises from the project were kept. Response times, handover of files and logins, what was included versus billed. That is worth knowing before your next engagement with anyone, and it is why the questions to ask a web designer before you sign are the ones about the boring parts rather than the design.

Where does this plan break down?

Two places worth naming.

When the sample is too small to say anything. At 20 inquiries a quarter, a move from 39 to 45 is not a result. It is a normal quarter. Treat month three as a direction check and a defect hunt, not as a verdict, and give a genuinely uncertain launch two quarters.

When too much changed at once. A new site, a new domain, a new phone number and a new ad campaign in the same month produce a number you cannot attribute to anything. If you must change several things, stagger them by a few weeks so the record has edges in it.

What to do this week

If you are pre-launch, put four calendar entries in now: a week one defect sweep, a week three verification check, a month two publishing session, and a month three review. Book them like appointments, because nothing here happens spontaneously.

If you launched recently, do the week one sweep today even if it is late. Forms, phone links, old addresses, indexing status.

If you launched months ago and none of this happened, start with the two checks that find money: submit your own contact form and confirm it arrives, and confirm your conversion tracking has been recording. Those two go wrong more often than anything else on a website.

Questions about your launch plan?

Email me at eric@seod.com with your launch date, whether it has happened yet or not. I will send back the 90 day checklist with the specific dates filled in for your project and the three checks I would not skip in your situation. It takes me a few minutes and it is the part almost every project leaves out.

More on running a project in the web design and branding library.