Why traffic drops after a redesign

The new site goes live, everyone is pleased, and three weeks later search traffic is noticeably down. That drop almost never comes from the design. It comes from what got dropped during the move, and what gets dropped is usually the part nobody looks at.

Addresses changed and nothing redirects

This is the most common cause and the most preventable. The new site builds addresses differently: /services/web-design becomes /web-design, the trailing slash disappears, a category shifts a level.

Whatever an old address had accumulated in search is attached to that address. Without a permanent redirect, the new address starts from nothing and the old one returns a 404.

There is a measurable check for this: take the old site’s sitemap, request every address against the new site, and leave no 404 in the list. That list can run to hundreds of rows, and it is the one list that has to run before launch.

The copy got shorter

The second cause is a side effect of design decisions. The new design is airier, has less text, leans on images. The word count on a page falls to a third of what it was.

Most of what a search engine knows about a page comes from that text. When phrases that used to match queries move into an image or a heading, the page stops answering the same questions.

This is not an argument against writing less. It is an argument for knowing which phrases were bringing traffic before you cut. A cut made without that list is a blind one.

Titles and descriptions got templated

The third cause moves quietly. On the new site, page titles come from a single template and every page carries the same pattern.

The old site had titles written by hand that matched the way people search. The template flattens them, and the text shown in search results becomes generic. Click-through falls, and ranking follows it.

Templates are not the problem; the assumption that a template fits every page is. The pages that carry your traffic should keep hand-written titles.

Internal links broke

The fourth cause is the least noticed. On the old site, pages linked to each other from inside the body copy; on the new one, the links moved into the navigation and a related-posts component.

Internal links build both the visitor’s path and a search engine’s sense of how pages relate. Pulled out of the body, the connection between pages weakens and anything deep in the site gets harder to reach.

Don’t migrate without measuring

All four causes come from one omission: no measurement was taken before the move.

Three things need recording before launch. The list of the hundred pages that bring traffic. The queries each of them arrives on. And the current address of each.

With those three lists, the migration turns into a checklist: what is this page’s new address, is the redirect in place, was the copy kept, is the title still written by hand.

Without the lists it takes three weeks to notice the drop, and by then there is no “before” to compare against.

A drop is not always a mistake

There is an exception and it needs stating honestly. Some redesigns cut traffic deliberately: thin pages are merged, visitors arriving on irrelevant queries are filtered out.

In that case the number to watch is not visitors but conversions. If traffic is down twenty per cent while enquiries are up, the migration worked. An assessment that does not read both numbers together makes a good decision look like a mistake.