A redesign is the only common business project that routinely makes things worse and is nearly always judged on whether it looks better.
The pattern is familiar enough to be a genre. The old site looked dated. The new one is beautiful, everybody internally is pleased, and three months later leads are down, search traffic is down, and nobody can say which change caused it — because everything changed at once.
The cause is not bad design. It is that the old site was doing things nobody had written down, and the redesign removed them without knowing they existed.
A page that ranked for a phrase you never targeted. A paragraph that answered the objection your best customers had. A URL that three years of external links point at. An unglamorous comparison table that quietly did more selling than the homepage. None of that is visible in a design review. All of it is discoverable in an afternoon of measurement.
Do the afternoon first.
The six things to write down before anyone opens a design tool
1. Which pages actually bring people in
Pull the last twelve months of traffic by landing page. Not pageviews — entry pages, the ones people arrive on.
The result is almost always a surprise, and it is almost always long-tailed: a handful of pages you expected, and then a set of specific, unglamorous pages doing real work. An old blog post. A page about one narrow service. Something you forgot existed.
Those pages are assets. Mark every one of them as protected before the project starts.
2. Which pages produce enquiries
Different list, and the more important one. Traffic is not the business. Which pages did people last see before they contacted you?
If you cannot answer this, that is the first finding of the project, and the fix is not a redesign — it is getting analytics and form tracking in place and waiting a month. A redesign launched without this measurement cannot be evaluated afterwards, which means you will never know whether it worked.
3. What people search for to find you
Search Console shows the queries that surface your pages, including the ones you never targeted. This list tells you what your market believes you do, in their words rather than yours.
It also tells you which phrases you are accidentally ranking for and must not lose. A redesign that rewrites every page in shinier language routinely deletes the exact wording that was earning the traffic.
4. Where people leave
The pages with the highest exit rate on the path to your goal. A step where half your visitors leave is the highest-value thing on your list, and it is a specific, fixable problem rather than a reason to rebuild everything.
Very often the answer here makes the redesign unnecessary. If the drop is at a checkout step, a form, or a page that takes nine seconds to load on a phone, you have a defect, not a design problem.
5. Your current speed, measured properly
Record today's numbers before you change anything, on a phone on a mid-range connection, on your top three entry pages.
You need this for two reasons. It is the baseline that tells you afterwards whether the new site is faster or slower — and new sites are very often slower, because a redesign is where heavy imagery, video backgrounds and extra fonts get added by people who are reviewing them on a fast laptop. Speed is a revenue variable, not a technical one, and we make that case in full in speed is a pricing decision.
6. Every URL you have
Export the full list — from your sitemap, your analytics, and Search Console. This is the document that prevents the single most damaging redesign failure, which is not design at all.
The two rules that protect everything else
When we relaunched Finkonkurs, a bilingual crypto data portal whose brand dates back to 2010, the entire migration ran under two rules written down before any work started:
No URL changes. Nothing de-indexed.
Those two sentences are the most valuable thing in this article, and they are worth more than any design decision in a relaunch.
Years of indexed URLs and external links are an asset accumulated slowly and destroyed instantly. The relaunch had to protect them, so a middleware layer serves a snapshot of the old site at its original URLs and 301-redirects everything else. The mirror ended up covering all 1,047 pages of the snapshot — up from 468 when it was keyed only to the old sitemap, because the sitemap turned out not to know about every page that existed. It can never shadow a live route.
That gap between 468 and 1,047 is the lesson. Your sitemap is not a complete list of your URLs. Half the pages worth protecting were ones the site itself had forgotten.
If URLs genuinely must change — sometimes they must, for structure or language reasons — then every old URL needs a permanent redirect to its closest new equivalent, one by one, and "redirect everything to the homepage" is the same as deleting them. On the same project the bilingual routing was hand-tuned rather than generated: English unprefixed, Russian at /ru/ but with native Russian slugs, alternates redirecting to a single canonical, and flat vanity URLs resolving to the right detail page. That is deliberate work. It is also the difference between a relaunch and an outage with a new logo.
Ask whether you need a redesign at all
Once you have the six measurements, ask the question honestly, because four different problems all get described as "our site needs a redesign".
"It looks dated." Sometimes true and sometimes worth fixing on its own — a site that looks abandoned raises a question about whether the business is. But this is a visual refresh: type, color, spacing, imagery. It does not require new URLs, new content or new structure, and it should not get them.
"It does not work on phones." That is a defect with a fix, and it is usually cheaper than a redesign.
"We cannot update it ourselves." That is a CMS problem. See the case against your page builder for why this one often gets solved in the most expensive available way.
"It does not bring in customers." This is the only one that might genuinely need a rebuild, and it is also the one least likely to be solved by appearance. Look at measurements 2 and 4 first. If people arrive and leave at the price question, the problem is that you do not publish prices, not that the typography is tired.
We say on our about page that we will not praise a weak idea to win a signature, and this is where that commitment costs us most often. A refresh plus three fixed pages is a smaller invoice than a rebuild, and it is frequently the right answer.
If you do redesign, change one thing at a time
The most common reason redesign outcomes cannot be explained is that they were all shipped simultaneously: new design, new content, new structure, new URLs, new platform.
When results move, there are five candidate causes and no way to separate them. When results move down, there is no way to roll back the one that did it.
Sequence it instead:
- Fix defects first — speed, mobile, broken forms. Measure.
- New design, same content, same URLs. Measure.
- New content, page by page, starting with the ones that produce enquiries. Measure.
- Structure and URL changes last, if still needed, with the redirect map written before launch.
This is slower in calendar terms and considerably faster in learning. It also means that at every step you can tell whether things got better — which is the entire point of doing it.
And keep the old site reachable until you are sure. Not live, but archived: the pages, the copy, the URL list. Nobody regrets having it.
The launch-day checklist
Whatever you change, do not launch without these.
- Every old URL either still works or 301s to its closest equivalent — checked from an exported list, not from the sitemap
- Form submissions tested end to end on the live domain, and the messages confirmed as arriving
- Analytics installed on the new site and recording, with the goal or conversion event working
- Speed measured on a phone against your recorded baseline
- Search Console updated, new sitemap submitted, old URLs monitored for crawl errors for a month
- The pages you marked protected in step 1 spot-checked individually
Then watch for four weeks. A redesign is not finished on launch day; it is finished when you know whether it worked.
Tell us what you need. You get a fixed price and a delivery date back — free, and with no obligation to proceed.
Frequently asked questions
How long should I measure before starting?
If analytics are already running, a day of analysis is enough — you are reading twelve months of existing data. If they are not, install them and wait at least a month before committing to a direction. Launching blind and then being unable to evaluate the result is the more expensive option.
Our traffic is tiny. Does this still apply?
The measurement takes less time, and the URL rule matters just as much — small sites often have a few old pages carrying a disproportionate share of enquiries, and losing one of them is proportionally worse. What you can skip is the statistical caution; with low traffic, judge changes on whether enquiries arrive, not on rate differences.
Is it ever right to change every URL?
Occasionally: a genuine restructure, a language strategy change, a move from a platform that imposed ugly paths. When it is right, it is right with a complete one-to-one redirect map produced before launch. What is never right is changing URLs because the new system defaulted to a different pattern and nobody checked.
Should I keep the old content?
Keep it, then improve it in place. Rewriting from scratch throws away wording that was already earning search visibility. The safe order is: keep the URL, keep the phrases that rank, improve the rest.
What if we do not have analytics at all?
Then that is the project. Install analytics, add form tracking, publish nothing else for a month. It costs almost nothing and it converts every subsequent decision from a matter of opinion into a matter of evidence.
