A redesign is a gift to new users and a tax on existing ones. The whole craft is in keeping the tax small.
When people already use a product daily, they are not looking at your interface — they are executing muscle memory. Move the thing they reach for and you have not improved their day, you have interrupted it.
Decide what is load-bearing before you touch anything
Some things in a product are structural: the primary navigation order, the location of the main action, the vocabulary. Change one of those and every returning user has to re-learn the product.
We listed those first and put them behind a rule — nothing on the list changes unless there is a measured reason, written down, with someone's name on it.
Ship in phases that each stand alone
Big-bang rewrites stall halfway and leave a product that is half two things. Every phase we shipped was complete on its own and improved something a user could name.
If a phase cannot be described as a benefit to a user in one sentence, it is not a phase, it is a milestone in your project plan.
Change appearance and behaviour on separate days
The visual layer went out first, with behaviour untouched. Everything looked new and everything still worked exactly where it had. The interaction changes followed once the surprise had worn off.
Support volume in the week after launch was the number we watched, and it barely moved — which was the entire objective.
What we deliberately did not change
The navigation order, the words on the primary action, and the keyboard shortcuts. All three were arguably improvable. None of them were worth the cost of asking a working audience to re-learn a habit for a marginal gain.
Want this run on your site?
A fixed-fee performance audit, delivered in a week.