RankPilot All articles
E-Commerce SEO

Twelve Gates Before Wheels Up: A Domain Migration Checklist That Protects Revenue

RankPilot

A domain migration is the highest-stakes maneuver in search. Everything the site has earned—links, crawl trust, brand queries, category authority—travels through a single window of time, and the margin for error is thin. For a US retailer, the difference between a clean cutover and a sloppy one is not measured in rankings positions. It is measured in Cyber Monday revenue that either arrives or does not.

Most migrations do not fail because of exotic technical problems. They fail because nobody agreed in advance what would happen, who would decide, and what evidence would justify turning back. The twelve gates below are designed to remove that ambiguity.

Phase One: Preflight (Gates 1–4)

Gate 1 — Take a baseline you can defend. Before a single URL changes, capture 90 days of organic sessions, impressions, clicks, and revenue by landing page. Export the top 500 URLs by revenue, not by traffic. That ranking matters later, because a page with modest traffic and high conversion value is not interchangeable with a popular blog post.

Gate 2 — Build a complete URL inventory. Combine the XML sitemap, server logs, internal link crawl, and analytics landing-page report. Reconcile the four lists until discrepancies are explained. In practice, the analytics report will surface URLs no sitemap ever contained—legacy product pages, filter combinations, campaign landing pages that still earn links.

Gate 3 — Classify every URL by fate. Each entry gets one of four labels: keep one-to-one, consolidate, retire with a redirect to a close topical match, or remove with a 410. Anything labeled "decide later" is a gate failure. Ambiguity at this stage becomes a soft 404 at cutover.

Gate 4 — Confirm the redirect map is specific. Wildcard rules are a convenience, not a strategy. A blanket pattern that sends every old product URL to the homepage teaches search engines that the old structure had no meaningful equivalent. Map product to product, category to category, and guide to guide. Redirect chains longer than one hop should be flattened before launch.

Phase Two: Rehearsal (Gates 5–8)

Gate 5 — Replicate the full stack on staging. Robots directives, canonical tags, sitemap generation, structured data, pagination, and internal search parameters all need to behave on staging exactly as they will in production. A staging environment that blocks crawlers is fine; a staging environment that silently drops canonical tags is not.

Gate 6 — Run a crawl diff, not a spot check. Crawl the staging host and compare it against the current site. The output should show expected changes and nothing else. Every unexplained difference—a missing canonical, a new noindex, a broken pagination sequence—is a defect to close before Gate 9.

Gate 7 — Pressure-test performance and error handling. Measure time to first byte and Core Web Vitals on the new host under realistic load. Confirm that 404 and 500 pages behave correctly and that the redirect layer does not add latency to every request.

Gate 8 — Write the rollback plan and name the decision-maker. The rollback plan is a document, not an instinct. It specifies who can authorize a reversal, how DNS or server configuration is restored, and how long restoration takes. One person owns the call. Committees do not roll back migrations at 2 a.m.

Phase Three: Cutover (Gates 9–10)

Gate 9 — Execute in the lowest-traffic window you have. For most US retailers that is a Sunday night or a Tuesday before dawn, adjusted for the West Coast. Avoid the two weeks before major promotional periods. If the calendar is already tight, delay the migration rather than compress the rehearsal.

Gate 10 — Verify the live signals immediately. Within hours of cutover, confirm: redirects resolve in one hop, the new sitemap is submitted, robots directives are correct, canonicals point to the new host, internal links use the new domain, analytics and Search Console properties are receiving data, and paid campaign destination URLs have been updated so media spend is not funding redirects.

Phase Four: Landing (Gates 11–12)

Gate 11 — Monitor against the baseline, not against feelings. Daily for the first two weeks, then weekly through week eight, track indexed page count, impressions by directory, organic sessions to the top 500 revenue URLs, and conversion rate by landing page. A small dip is normal. A structural decline is not.

Gate 12 — Close out and decommission deliberately. Keep redirects live for a minimum of twelve months—longer for pages with substantial external links. Retire the old sitemaps, update brand assets and marketplace listings, and archive the redirect map with an owner's name attached so it can be maintained after the project team disperses.

Defining Rollback Triggers in Advance

A rollback threshold agreed upon during rehearsal is a decision. A threshold invented during a crisis is an argument. Reasonable triggers include: a sustained decline in organic sessions to the top 100 revenue URLs exceeding 25 percent over seven days with no alternative explanation; a collapse in indexed pages that persists past fourteen days; cascading server errors or redirect loops affecting crawl budget; or a measurable drop in conversion rate on migrated templates that survives a two-week window.

Any one trigger does not automatically mandate reversal. It mandates a review with the named decision-maker, a documented diagnosis, and a deadline for the next review. What it forbids is silence.

Stakeholder Checkpoints That Prevent Surprises

Run three formal checkpoints: a go/no-go review after Gate 8, a post-cutover verification call within 24 hours of Gate 10, and a 30-day performance review. Invite engineering, merchandising, paid media, and finance. Merchandising cares about category pages that vanished from navigation. Paid media cares about destination URLs. Finance cares about the revenue line that funds the next quarter. Each group will catch a failure mode the SEO team is not positioned to see.

The Landing Is the Point

A migration is not complete when the DNS propagates. It is complete when the new domain is outperforming the old baseline on the metrics that pay for the project. Treat the twelve gates as a checklist with signatures, not a set of suggestions, and the maneuver becomes what it should be: a controlled change of address rather than a bet on the weather.

All Articles

Related Articles

Who Gets the Lifeboat? Scoring Every Page Before You Purge Content

Who Gets the Lifeboat? Scoring Every Page Before You Purge Content

Ground School to Cross-Country: A New Online Store's First 90 Days of Search

Ground School to Cross-Country: A New Online Store's First 90 Days of Search

The Citation Scoreboard: Auditing Who AI Answer Engines Actually Quote

The Citation Scoreboard: Auditing Who AI Answer Engines Actually Quote