E-commerce redesign without losing what already sells
In short
Most redesign damage is avoidable. Inventory the pages, products and queries that earn money, freeze checkout and product pages while you rebuild around them, map every old URL to a new one, and ship in slices with a rollback plan. Measure revenue per session before and after, not just how the new design looks.
A redesign feels like a fresh start, but your store is not starting fresh. Years of links, rankings, habits and muscle memory point at the current pages. The job is to change the skin without stopping the heart: keep what earns, rebuild the rest, and prove with numbers that the new version sells at least as well.
Inventory what earns before touching anything
List the pages that bring revenue, not just traffic. Export the top product pages, category pages and blog posts by orders and revenue, then note which search queries send them visitors and which outside sites link to them.
That list is your protected set. Everything on it gets special treatment: same intent, same content depth, same internal links, and a redirect if its address changes. Pages outside the list can be merged, rewritten or removed.
Take screenshots and save the raw numbers. You will compare against them for weeks after launch, and memory is a poor baseline.
- Top 20 pages by revenue, with their queries and backlinks.
- Checkout, cart and account flows, step by step, as they behave today.
- Site speed on the money pages, measured on a real phone.
- Forms, pop-ups and apps, with what each one earns or costs.
Freeze the money paths
Checkout is where redesigns go to die. A prettier cart that converts half a point worse can erase the gains of the whole project. Rebuild around checkout first, and touch it last, with its own testing window.
The same applies to the product page template and the mobile menu. These are high-traffic, high-stakes surfaces where small regressions hide. Change one thing at a time on them, and keep the old version one click away until the new one proves itself.
Map every redirect, then check them twice
Every URL that earned a link, a ranking or a bookmark needs a destination. Export the full URL list, decide keep, merge or remove for each, and write the redirect map before the build starts, not the night before launch.
Redirect to the closest equivalent, never everything to the homepage. A category goes to its new category, a product to its replacement or its parent category, a post to its updated version.
- One-to-one map: old URL, new URL, reason.
- No redirect chains: old pages point straight at the final address.
- Internal links updated to the new addresses, not left pointing at redirects.
- A post-launch crawl within 24 hours to catch 404s and chains.
Rebuild in slices, not in one big reveal
Ship the new homepage first, then categories, then product templates, then content. Each slice gets a week of data against the old baseline before the next one ships.
Slices make problems small and blame obvious. When conversion dips the week the new product template ships, you know exactly where to look. With a big-bang launch, everything changed at once and nothing can be isolated.
Keep a rollback path for every slice. A redesign you cannot undo is a redesign you cannot test honestly.
Run a launch-week protocol
Launch early in the week, never on a Friday, and never during your peak season. Watch revenue per session, checkout completion and site speed daily, not vanity traffic.
Expect a wobble. Rankings shuffle when addresses change, and regulars need a few visits to learn new paths. Judge the redesign after three to four weeks of data, unless revenue per session falls off a cliff, in which case roll back the last slice first.
- Launch Tuesday to Thursday, with the team available the next two days.
- Daily check: revenue per session, checkout completion, speed, 404 rate.
- Fix redirects and broken internal links within 24 hours.
- Write down every lesson while it is fresh; the next redesign starts there.
Keep them wherever possible. A kept URL needs no redirect and keeps its history. Change addresses only when the structure genuinely improves, and redirect the rest one-to-one.
The inventory and redirect map take longer than most teams expect, often a third of the project. Rushing that phase is what makes launches slip and rankings drop.
Avoid it. Freeze everything except fixes during your busiest weeks, and schedule the launch for a quiet period with the team available to react.
Redesign when the platform and data are sound. Rebuild when the platform itself is the problem: slow, rigid, or impossible to measure. Most stores need the first, not the second.