How long should we wait after launch before judging whether the redesign worked?

The first seven days after a launch measure your launch, not your website. Three things move the numbers at once and none of them are the design.

Returning visitors are relearning where things are. Someone who found your pricing page in two clicks for the last three years now hunts for it. In the launches we have watched, that shows up as a short dip in pages per session and a rise in time to first click, and it fades as the returning share of your audience cycles through once. If most of your visitors come back monthly, that takes about a month.

Caches and indexes are churning. Your CDN is cold, your old URLs are still in Google’s index, and analytics may be double-counting or under-counting until the new tags settle. A traffic chart drawn from that week is measuring infrastructure.

And nobody accounts for the calendar. We have seen a team panic over a 30% traffic drop in a week that turned out to be the week their industry’s biggest conference ran, which had done the same thing the previous two years. Launch weeks are usually chosen for internal convenience, which means they land wherever they land in your seasonal cycle.

So the honest answer to “how did it go?” on day three is: we do not know yet, and here is the date we will.

Different numbers become readable at different speeds. Read them in this order, because reading a slow metric early is how teams talk themselves into reversing a good decision.

These windows are what we typically see across launches, not a law. A site with 200 visitors a week needs longer for everything below, because small numbers move on their own.

1. Days 1 to 3: technical health. Broken pages, failed forms, missing redirects, Core Web Vitals in the lab. These are readable immediately and should be fixed immediately. They are not verdicts on the design.
2. Week 1: page speed and error rates in the field. Real-user timings stabilise once the cache warms. If your largest pages are still slow after seven days, that is real.
3. Weeks 2 to 4: behaviour on individual pages. Scroll depth, click-through on primary buttons, form completion rate. New visitors are the clean sample here, so segment them out and ignore the returning cohort until week four.
4. Weeks 4 to 8: conversion rate. It needs enough completed conversions to mean anything. If you get 40 leads a month, four weeks is one sample and eight weeks is two. Do not compare a two-week window to a two-week window.
5. Weeks 8 to 12: organic search. Google needs to recrawl, drop the old URLs and settle rankings on the new ones. Movement inside eight weeks is normal churn, and we treat it as information only if it is still there in week ten.
6. One quarter, year over year: revenue and pipeline quality. Compare the quarter after launch to the same quarter last year, not to the quarter before it.

Waiting is the default, not a rule. Five findings justify changing or reverting something inside the first week, and all five are failures rather than disappointing numbers.

Everything else waits. “Bounce rate is up” is not on this list. Neither is “the CEO does not like the hero”.

• A conversion path is broken. Form submissions or checkouts have gone to near zero, not down 20%. Check the inbox, not the dashboard, because a form can appear to submit and send nothing.
• Money has stopped. Payments failing, a pricing page 404ing, a booking calendar not loading on mobile.
• A whole device class is unusable. Something looks fine on the designer's laptop and is broken on iPhone Safari or a 1366px sales laptop, and that segment is a third of your traffic.
• Redirects are missing at scale. Old URLs returning 404 instead of a 301 to their new home. Fix these in days, not weeks, because the cost compounds in search.
• Accessibility has regressed to the point of exclusion. Keyboard navigation trapped, contrast that makes body text unreadable, a form with no labels.

Before you blame the design, rule out the two cheaper explanations. Run the checks in this order because tracking is free to verify and seasonality costs one query.

First, verify tracking. Compare a hard number you own, actual leads in your CRM, signed-up accounts, orders in the payment dashboard, against what analytics reports for the same days. If the CRM says 22 leads and analytics says 9 conversions, you have a measurement problem, not a design problem. This is the single most common cause of a fake post-launch crisis we see, and it usually comes from a tag that was never re-added to the new thank-you page.

Second, rule out the calendar. Pull the same weeks from last year and the year before. If traffic fell 18% in the two weeks after launch and it fell 15% in the same two weeks last year, you have found your answer.

Third, split new visitors from returning ones. If new visitors convert at or above the old rate and returning visitors have dropped, that is relearning, and it recovers. If new visitors have dropped too, the page is genuinely doing worse and it is worth changing.

Fourth, look for a single page rather than a sitewide effect. Real design problems usually concentrate. A drop spread evenly across every page and channel is almost always tracking, seasonality or an indexing issue.

At eight weeks you will have a list of pages that underperform their old versions. Change one thing at a time, starting with the cheapest change that could explain the gap.

Start with copy, not layout. Headline, subhead and button label are the fastest edits your team can make in the CMS and they carry most of the persuasion. In our projects, teams reach for “move the section up” when the actual problem is that the headline describes the product instead of the reader’s problem.

Then restore what you removed. Redesigns quietly delete things: a comparison table, a customer logo row, a short FAQ under the pricing. If a page lost conversions and lost a section, put the section back before you redesign anything.

Then change the layout, and only on the pages where the numbers justify it. Give each change two to four weeks before you judge it, for the same reason you gave the launch eight.

And write down what you learn per page. The pages that dropped tell you more about your audience than the pages that rose.

Put four dates in the calendar before launch day and tell whoever is asking for a report that those are the dates.

Day 3: technical check only. Report “no broken paths” or the list of what is broken. No conclusions about performance.

Week 2: behaviour on the top five pages, new visitors only. Note observations, make no changes beyond fixes.

Week 8: the first real verdict. Conversion rate against baseline, page by page. This is the meeting where you decide what to change.

Week 13: search and revenue, compared to the same quarter last year. This is where you find out whether the redesign worked.

All of this assumes you wrote down what the old site did before you replaced it. If you have not launched yet, that is the one job worth doing this week, because a redesign with no baseline can never be proven to have worked. If you have already launched without one, use last year’s same-quarter data as your baseline and accept it is rougher.