You can put a real number on how badly a replatform is likely to hurt you, and it comes down to three measurable things: how many valuable URLs are about to change without a mapped redirect, how many pages are near-duplicates fighting each other, and how many files and pages have no internal links pointing at them. Weight those, sum them, and you get a migration risk score. The catch is that the number is only as honest as the person who explains what went into it.
Here is the moment that made me care. A team I know got their migration greenlit because a readiness dashboard glowed a confident "94%." Everyone relaxed. They cut over. Traffic fell off a cliff about ten days later, and when they went looking for the 6% that was apparently fine to ignore, it turned out to be the URLs that earned most of their organic traffic. The score wasn't wrong so much as it was uninterpretable. Nobody could tell you what the 94 was made of.
A migration risk score is a credit score for your site's move
Think about how a credit score actually works. It's a single three-digit number, and everyone treats it like gospel, but underneath it's a weighted blend: payment history matters most, amounts owed matters a lot, length of history and new credit matter less. The single number is useful precisely because you can pull it apart and see which part is dragging you down. A migration risk score should behave the same way. One headline figure for the exec who wants a go/no-go, and three sub-scores underneath so the people doing the work know where the trouble lives.
The three components that carry almost all the real danger are redirect risk, duplicate content, and orphaned files. Everything else (metadata drift, image alt text, schema) is real but recoverable after launch. These three are the ones that quietly bleed you before anyone notices.
How I weight the three
Weights are opinions, so here are mine, and here's the reasoning. Redirect risk gets the most weight because a URL that changes path with no 301 in place doesn't degrade gracefully. It returns a 404, the ranking signals a 301 would have passed forward evaporate, and the fix after the fact is a slow crawl back. Duplicate content comes second because it dilutes and confuses rather than deletes. Orphaned files come third because they're mostly a discoverability tax, not an outright loss (still a tax, though).
| Signal | What it measures | Weight |
|---|---|---|
| Redirect risk | Share of traffic-earning URLs changing path with no mapped 301 | 45% |
| Duplicate content | Clusters of near-identical pages competing for the same query | 30% |
| Orphaned files | Pages and assets with zero inbound internal links | 25% |
Weight by value, not by page count
This is the mistake the 94% dashboard made. If you score every URL equally, a 50,000-page site with 49,000 tidy tag archives will drown out the 300 pages that pay the bills. Don't do that. Multiply each page's raw risk by its value, where value is real traffic, real revenue, or real backlinks pointing in. A high-risk page that nobody visits and nobody links to is a rounding error. A low-risk page that earns a fifth of your organic sessions is not.
So the page-level formula I actually use looks like this: page_risk = (redirect_factor * 0.45 + duplicate_factor * 0.30 + orphan_factor * 0.25) * value_weight. Roll those up, normalize against the worst plausible case, and your top-line score means something you can defend in a meeting.
Scoring redirect risk
For each URL that earns traffic or holds links, ask one question: does it survive the move at the same address, or is there a mapped 301 to its true new equivalent? Not to the homepage. Not to a category page that sort of relates. Redirecting everything to the homepage is the digital equivalent of forwarding all your mail to "the building." Google treats those soft, irrelevant redirects as effectively broken. A URL with a clean one-to-one 301 scores near zero. A URL changing path with no target scores at the top. Chains (a redirect pointing at another redirect) sit in between and add latency you'll regret.
Scoring duplicate content
Cluster your pages by near-identical text and by the query they target. Every cluster with more than one member is a small internal argument about which page should rank. Score each page by how many siblings it's competing with and whether a canonical tag already picks a winner. A migration is the best chance you'll ever get to consolidate these, because you're touching every URL anyway. Merge, canonicalize, or kill.
Scoring orphaned files
An orphan is any page or asset with no internal link pointing to it. Crawlers find pages by following links, so an orphaned page relies entirely on old external links and a sitemap that may not survive the cutover. The classic casualties are PDFs, gated whitepapers, and old landing pages that were only ever reached from an email blast three years ago. They still rank, they still convert, and your new CMS has no idea they exist. Score by whether the file is reachable from the new site's link graph at all.
What I'd actually do
Build the score. Genuinely, build it, because a defensible number stops migrations from being greenlit on vibes. But never, ever ship the single top-line figure without its three sub-scores next to it, and never weight by page count when you could weight by value. If I had to trust just one component, it's redirect risk, every time, because it's the only one that turns your best pages into 404s overnight. The other two you can chip away at for months. That's the sub-score I'd stare at before signing off, and it's the first thing a tool like ours surfaces when it crawls your pre-migration inventory.
That team with the 94%? Their redirect sub-score, had anyone computed it, would have been a screaming red. The number wasn't lying. It just hadn't been asked the right question. Ask it now, while the old site is still standing.