What actually happens to organic traffic

Strip the anecdotes and the distribution looks like this. It's wider — in both directions — than most operators assume.

• 01
  The full distribution from the 892-migration dataset: average recovery 523 days, fastest recoveries 19–23 days, 17% never recovered after 1,000 days. Industry syntheses put 60–80% of migrations at some measurable traffic loss, and only about 1 in 10 improving rankings. Nearly the entire spread is explained by execution, not platform choice.
• 02
  The normal, healthy pattern: a 10–30% organic dip that stabilizes within 2–6 weeks, with recovery to 95–100% of baseline by weeks 8–12 for well-executed migrations. Ecommerce-specific cases cluster around a 15% dip in month one, recovered by month three. Enterprise sites past 50,000 pages run longer; 3–6 months is typical.
• 03
  The unhealthy pattern has a name in agency practice: the migration hangover. A drop of 50% or more with no recovery after four weeks. It marks a structural failure (missing redirects, a shipped noindex tag, broken canonicals), not normal recrawl volatility, and it requires diagnosis rather than patience.
• 04
  The downside cases are expensive and specific. One large retailer lost roughly $5M in a single month after its IT consultants rejected URL redirect recommendations during a redesign. Another documented case dropped 63% overnight after redirecting HTML pages but forgetting linked assets — images, PDFs, videos — then spent 11 days fixing redirects and four months recovering.
• 05
  The upside cases are just as documented. An ecommerce migration handled as an SEO project produced a 126% organic traffic increase; another recovered rankings a botched earlier migration had destroyed and added 139,000 monthly organic sessions plus 22,000 new keywords. Migration done well is an optimization event, not a survival event.
• 06
  Ahrefs adds the number that should shape your caution: 60% of pages that lose rankings in a migration never fully recover them. Traffic recovers in aggregate before it recovers page by page, which is why the protection work in this guide happens per-URL, ranked by value, and why the ECM migration checklist treats the crawl as Phase 1, not a launch-week task.

Where the equity actually leaks

Google doesn't penalize platform changes. Traffic leaks through a short list of specific, preventable mechanisms — and knowing them by name is most of the defense.

The redirect-layer leaks

• →
  Incomplete maps. The classic version in ecommerce: platform URL structures change wholesale (Shopify's /products/ and /collections/ paths never match Magento's or WooCommerce's — the route-specific traps are in the ECM Magento and Shopify migration guides), and any URL without an explicit 301 to its closest equivalent becomes a 404. Bulk-redirecting everything to the homepage doesn't count — Google treats it as a soft 404 and the equity evaporates anyway.
• →
  Redirect chains and 302s. Chains leak link equity and slow crawling; temporary redirects tell Google the move isn't permanent. The standard is one clean server-side 301 per URL: not JavaScript redirects, not meta-refresh, no intermediate hops.
• →
  Forgotten non-HTML assets. The 63%-overnight-drop case above came from exactly this: images, PDFs, and downloadable resources carry backlinks and rankings too. A two-year-old whitepaper with 40 backlinks ranking #3 for a valuable keyword loses all of it if deleted without a redirect.

The on-page and technical leaks

• →
  Metadata that doesn't survive the transfer. Custom title tags and meta descriptions (often stored in platform-specific attributes) arrive lost, truncated, or replaced with defaults. Migration tools treat SEO fields as premium add-ons and silently skip them on basic plans. Verify top pages against the pre-migration export.
• →
  Launch-day technical landmines: the staging noindex tag that ships to production, canonicals still pointing at the staging hostname, and an outdated XML sitemap. Each one is a five-minute check in staging and a multi-week recovery in production.
• →
  Core Web Vitals regression. If the new platform loads slower than the old one, rankings decline even with perfect redirects — one documented case tied a 44% organic decrease to a rushed launch. Post-launch targets: LCP under 2.5 seconds, INP under 200ms, CLS under 0.1. Speed is part of the SEO migration scope, not a separate project — and worth a tech stack audit line item before the theme build starts.
• →
  Content changed mid-migration. Redesigns that quietly rewrite headers, thin out body copy, or remove optimized content stack a second variable on top of the platform change — and make the post-launch diagnosis ambiguous. Migrate first, optimize second, or accept that you won't know which change caused what.

The second scoreboard: AI citation equity

A migration in 2026 is judged on two boards. Rankings retained is the old one. Citations retained — across AI Overviews, Perplexity, and ChatGPT — is the new one, and most migration checklists don't mention it.

• 01
  A botched migration can remove a site from AI answers as fast as from rankings. AI answer engines assemble citations from crawlable, structured, consistently-resolving pages. Broken redirects and lost schema don't just drop positions; they drop the site out of the answer layer entirely, and the operator often doesn't notice because nobody was tracking it.
• 02
  Schema preservation moved from nice-to-have to launch-day non-negotiable. Product, Review, FAQ, and Breadcrumb markup must render — server-side — on the new platform from day one. Well-structured product pages increasingly surface in Google AI Overviews with no extra optimization, which means the markup is now a traffic asset with a measurable value, not a checkbox.
• 03
  Add an AI-citation KPI to the migration baseline. Before launch, record where the site appears in AI Overviews and answer-engine citations for the top 20–30 revenue queries alongside classic rank tracking. It's a second before-and-after that catches losses classic rank trackers miss, and it belongs next to the conversion and traffic baselines in your core KPI set.
• 04
  The upside runs through the same pipe. Platforms with strong default structured data give migrating merchants an AI-visibility bump they never explicitly built — one of the quieter reasons replatformed stores sometimes exceed their old organic baseline instead of merely recovering it. Platform-by-platform schema defaults are covered across the ECM platform reviews.

The protection work, in priority order

Pre-migration work determines 60–70% of the outcome before anything goes live. The playbook below is ordered by traffic value, because engineering hours are finite and URL counts are not.

• 01
  Export a 12-month baseline you'll defend later: top pages by clicks and impressions, top queries, average positions, and organic sessions from Search Console and analytics. Without this snapshot, the post-launch argument about whether the migration worked is unwinnable in either direction.
• 02
  Crawl everything, then map by value, not by count. The working rule from migration practice: the top 20% of pages drive roughly 80% of organic traffic, so score URLs by traffic and backlink equity and build the redirect map in that priority order. Every URL still gets a 301, but the pages that pay the bills get mapped, reviewed, and tested first.
• 03
  Validate in staging like it's the launch. Crawl the redirect map against the staging site, confirm the highest-value pages resolve with single clean 301s, check for the noindex tag and staging-hostname canonicals, and verify schema renders server-side. Errors in URL mapping compound irreversibly after launch; in staging they cost minutes.
• 04
  Launch with the monitoring already running: real-time crawl monitoring at cutover, the new sitemap submitted, the old sitemap briefly accessible so crawlers process the redirects, and Search Console coverage watched daily for two weeks. A 404 spike caught on day two costs hours; found in week six it costs a quarter.
• 05
  Keep redirects alive for the long haul: 12 months minimum, longer for domain moves. Redirect maintenance is the most commonly abandoned piece of migration hygiene, and pruning redirects early is how recovered equity leaks a second time. If organic drives meaningful revenue, a migration-experienced SEO specialist owning this playbook end-to-end is the cheapest insurance available against the 523-day outcome.

Reading your recovery: dip, hangover, or win

The post-launch weeks are a diagnosis exercise. The data gives you clean thresholds for which situation you're in and what each one demands.

• 01
  Weeks 1–2: volatility is the plan working. Expect 5–15% fluctuation in the first 30 days as the index recrawls; a 10–30% dip is within normal range. The checks that matter now are mechanical (coverage reports, 404 lists, redirect spot-checks on top pages), not rank obsessing.
• 02
  Weeks 2–6: the fork. Stabilizing toward baseline means you're on the healthy curve; hold your nerve and change nothing. A sustained drop past 30% points to a real technical problem worth diagnosing now. A drop of 50%+ with no recovery after four weeks is the hangover — audit for the structural failures (missing 301s, noindex, canonicals) immediately, because waiting doesn't fix structural causes.
• 03
  Weeks 8–12: the verdict window. Well-executed migrations sit at 95–100% of baseline here, matching Mueller's 4–12 week stabilization guidance. Run the review against the pre-migration snapshot (traffic, positions on top queries, AI citations, and Core Web Vitals) and score it honestly.
• 04
  If you're in the bad tail, act like it. The 523-day average exists because teams treated structural failures as volatility and waited. The documented fast fixes (the 63%-drop case recovered in four months once redirects were completed in 11 days) share one trait: someone diagnosed the mechanism instead of watching the graph. ECM tracks how these platform moves keep playing out in weekly Insights.
• 05
  And if you're at baseline by week 12, you're not done. You're at the starting line. The 1-in-10 migrations that improve SEO treat recovery as the floor: the new platform's speed, schema, and structure become the optimization budget the old stack never allowed. That's the actual return on the whole exercise.