Migration risk becomes actionable when you express it as a single weighted score per URL, built from three measurable inputs: redirect risk (does this URL earn links or traffic that a broken redirect would waste), duplicate content (does near-identical text dilute or cannibalize what you migrate), and orphaned files (are assets referenced but unlinked, or linked but missing). Score each factor 0–5, weight redirect risk highest, and you get a ranked triage list that tells you what to migrate carefully, what to consolidate, and what to leave behind. This article gives you the formula and the thresholds.
Why one score beats three separate reports
Most pre-migration audits produce three disconnected spreadsheets: a redirect map, a duplicate-content list, and an orphaned-asset dump. Each is useful, but none tells you which page to fix first when you have 40,000 URLs and six weeks. A composite score forces a shared scale, so a high-authority page that is also duplicated and links to a missing PDF rises to the top — where it belongs — instead of hiding in the middle of three separate lists.
The goal is not precision to two decimals. It is defensible ranking: when a stakeholder asks why you migrated one page and killed another, the score is your answer.
The three inputs, defined
1. Redirect risk (weight 0.5)
This is the cost of losing the URL. Score it by combining external link equity, organic entrances, and conversion role. A page with referring domains and steady organic traffic scores 5; a parameterized URL with zero inbound signals scores 0. High scores mean the redirect must be one-to-one and tested; low scores can be collapsed or dropped.
2. Duplicate content (weight 0.3)
Score how much this URL overlaps others by shingled text similarity. Near-exact copies (95%+) score 5 and signal a consolidation decision. Templated boilerplate that inflates similarity but carries unique core content scores lower. High duplicate scores paired with high redirect risk are the trickiest: you want the equity, but you cannot migrate five copies of it.
3. Orphaned files (weight 0.2)
Two failure modes live here. Orphaned assets are files (PDFs, images, downloads) that exist and get direct traffic but sit outside the crawlable link graph — easy to miss in a content-model export. Broken references are links pointing to files that no longer resolve. Score by traffic to the asset plus the count of pages that depend on it. A gated whitepaper with backlinks scores 5; a stray thumbnail scores 1.
The formula
RiskScore = (Redirect x 0.5) + (Duplicate x 0.3) + (Orphan x 0.2)
Range: 0.0 to 5.0Weight redirect risk highest because a lost redirect is the failure users and search engines feel immediately and permanently. Duplicate content is a quality problem you can fix on either side of the migration. Orphaned files are the quiet killer — cheap to catch pre-migration, expensive to discover from a 404 report three weeks after go-live.
Reading the score
| Score | Band | Action |
|---|---|---|
| 3.5–5.0 | Critical | Manual review, verified 1:1 redirect, regression-test after cutover |
| 2.0–3.4 | Elevated | Consolidate duplicates, confirm asset links, batch-validate redirects |
| 0.8–1.9 | Routine | Migrate in bulk, spot-check |
| 0.0–0.7 | Low | Candidate to kill or 410; skip if no signals |
Worked example
Take a product-comparison page: 22 referring domains and 4,000 monthly organic entrances (Redirect = 5), 88% text overlap with two regional variants (Duplicate = 4), and it links to a spec sheet PDF that returns 404 (Orphan = 4).
(5 x 0.5) + (4 x 0.3) + (4 x 0.2) = 2.5 + 1.2 + 0.8 = 4.5Score 4.5, Critical band. The plan writes itself: keep this URL, pick one canonical version and consolidate the regional variants into it, restore or replace the PDF before cutover, and put the redirect on the must-test list.
How to gather the inputs
- Redirect risk: join your crawl to Search Console (queries, clicks), Google Analytics (entrances, conversions), and a backlink source (referring domains). Normalize each to 0–5.
- Duplicate content: run shingled or MinHash similarity across rendered body text, not raw HTML, so navigation and footers do not inflate scores.
- Orphaned files: diff the link graph against the asset inventory in both directions — files with no inbound links, and links with no resolving file.
This cross-referencing is exactly what a pre-migration content-intelligence pass automates: it crawls the site, joins the traffic and link signals, and returns a per-URL migrate / rewrite / kill call instead of three disconnected exports. If you are scoping this by hand, our migration planning guide walks through the inventory join step by step.
Common mistakes that skew the score
- Counting boilerplate as duplicate content. Strip shared templates first, or your product pages will all look like copies of each other.
- Ignoring assets in the content model. Many exports move pages but not the files those pages depend on. Orphan scoring catches this before the links break.
- Treating all traffic equally. A page with entrances but zero conversions and no links may still be safe to consolidate. Blend the signals; do not lean on one.
- Scoring once. Re-score after consolidation. Killing duplicates changes which URL should inherit the redirect.
The score does not make the decision for you. It makes the decision rankable, explainable, and repeatable — which is what turns a chaotic replatform into a defensible plan.
Where to start
Run the three inputs on your top 500 URLs by traffic first. That subset usually holds most of the redirect risk and the duplicate clusters, and it lets you calibrate your weights before you scale to the full site. Once the bands feel right, apply the same formula across everything and let the Critical and Elevated rows set your migration sequence.