Post-mortem
Check backlinks for a page before the redirect map is final
A redirect map is a list of promises about where link value will go. This post-mortem of an invented migration shows which promise broke and what a valuer would have checked in advance.
The case file (invented, for illustration)
Picture a small shop moving from an old PHP site to a new framework. The shop, its URLs and every number here are made up.
The old site had 400 URLs and the team wrote redirects for the 120 they thought mattered, then pointed a catch-all at the homepage for the rest. Launch was a Tuesday in late October 2026, and by the following week impressions had slid on pages nobody had thought about.
Nothing was wrong with the redirects they wrote. The loss sat in the 280 they did not.
A hypothetical like this is no proof of anything. It is a rehearsal, and the Search Console coverage report and the free appraisal routine are what you would use on the real thing.
The week-by-week timeline
- 1
Day 0: launch
Catch-all sends every unmapped URL to the homepage. A search crawler treats many of these as soft 404s, and the pages lose their own identity.
- 2
Day 3: the sitemap
The new sitemap lists only new URLs, and lastmod is set to the deploy date for every row, so it no longer says which pages really changed.
- 3
Day 9: canonicals
The new templates canonicalise product variants to the parent. Two old URLs that held links now point at a page that never received them.
- 4
Day 21: the audit
Someone finally runs toppages on the old domain. Six URLs hold most of the referring domains, and four were in the catch-all.
What an appraiser would have priced beforehand
| Old URL | Referring domains (example) | Redirect on launch day | Priced loss risk |
|---|---|---|---|
| /guides/sizing | 46 | Catch-all to homepage | High: the page is the asset |
| /blog/2019-review | 31 | Mapped, one hop | Low |
| /p/12-red-boot | 9 | Canonical to parent | Medium: parent inherits, variant is lost |
| /about.php | 4 | Catch-all to homepage | Low |
What the evidence supports, and what it does not
A direct 301 is the baseline. Google has said since 2016 that 3xx redirects no longer lose PageRank, so a clean one-hop map is not where value leaks.
The leak is the mismatch: a redirect that points at an unrelated page can be treated as a soft 404, and a canonical that disagrees with the redirect sends mixed signals. Nobody outside Google can measure the loss precisely, which is why you count linking domains and treat the count as the exposure rather than as a forecast.
Before the move, run toppages for the old domain (2 credits) and refdomains for the five biggest URLs. After the move, repeat both once the monthly refresh lands and compare. Our index may need a cycle to notice. Stay calm. Do not panic on day 10, and keep the first report as your before photo.
One practical caution. A redirect map that looks tidy in a spreadsheet can still fail on the server, so test the top 20 with a crawler such as Screaming Frog and confirm each returns a single 301 to a live page.
Sort old URLs by referring domains, not by traffic. Map the top 20 by hand, one hop each, to the closest equivalent. Only then let a pattern handle the long tail, and never point the long tail at the homepage by default.