Score a page before you migrate by putting it through three questions: does it still earn traffic or citations, is the information on it actually correct, and would you be embarrassed if it launched word-for-word on the shiny new site? Pages that pass all three you migrate. Pages with good bones and stale flesh you rewrite. Everything else you kill. And you do all of this in a spreadsheet, before anyone so much as opens a template.
Here is the thing nobody warns you about. When you replatform, the temptation is to treat the old site as furniture and just haul all of it into the new house. Every page comes along, boxes unlabeled, because sorting felt like a pre-launch problem and launch felt far away.
Then launch is not far away, and you have moved your mess into a nicer building.
Picture it as a hiring round
Imagine the new site is a smaller office. Fewer desks, better chairs, and a manager who actually reads the org chart. Every single page from the old place has to walk in and reapply for its job. Nobody gets grandfathered in because they were there first. Twelve-year-old blog post that ranks for one dead keyword? Sit down, we will call you.
Most of your pages, if you are honest, are that one colleague who has been "here forever" and whose actual responsibilities no one can name. They are not bad. They are just not doing anything. And a migration is the one moment you are allowed to ask out loud.
The three verdicts map cleanly onto how you would run that review. You keep the people doing real work. You retrain the ones with potential but bad habits. And you let go of the ones who have quietly been collecting a paycheck since the last redesign.
The scorecard, before you touch anything
You want this to be boring and repeatable, because you are going to do it a thousand times. Here is what each verdict actually means and the signals that push a page there.
| Verdict | What it means | Signals that put it here |
|---|---|---|
| Migrate | Move as-is, keep the URL, keep the value | Gets real traffic or is cited by AI answers; still factually true; links point to it; you would defend it in a meeting |
| Rewrite | Good idea, wrong execution; move the URL, redo the content | Ranks or earns links but the content is dated, thin, or duplicated three times over; product names changed; the intent is right |
| Kill | Do not carry it over; redirect or retire it | Zero traffic in a year; nobody links to it; it duplicates a better page; the thing it described no longer exists |
None of these verdicts require a working new site. That is the whole point. You are grading on paper, in the same place your grandmother filed her recipes, not in a sandbox that costs a developer a day to spin up.
Why "migrate everything" quietly loses
Every page you carry over is a page you have to redirect, QA, and re-earn crawl budget for. Move ten thousand when three thousand were pulling their weight, and you have handed your team seven thousand extra chances for a 404, a redirect chain, or a duplicate that splits your own ranking signals down the middle.
Killing a page is not vandalism. A clean 301 to the nearest surviving page passes most of the equity along and tells search engines and answer engines exactly where the good stuff went. (Losing a page with no redirect is the vandalism. That is the version where the mail just stops.)
And here is the part people forget: rewrite is the expensive verdict, not kill. A rewrite means someone with a keyboard and a brain has to actually sit down and do it. So the honest question on every borderline page is not "is this worth keeping" but "is this worth someone's afternoon." If the answer is no, it was never a rewrite. It was a kill wearing a nicer coat.
How to actually run it
Pull an inventory of every URL. Bolt on the numbers that make a verdict defensible: organic sessions over the last twelve months, inbound internal and external links, last-modified date, and a duplication flag so near-identical pages get judged as a group instead of one at a time. Sort by traffic, and watch the shape appear. A small band of pages does almost all the work. A long, quiet tail does almost none.
Then go top to bottom and write a verdict in one column. Migrate, rewrite, or kill. No abstentions, no "maybe." Maybe is how ten thousand pages survive.
This is roughly the plan a pre-migration crawl produces automatically, scoring duplication, orphaned assets, and redirect risk so the verdict column mostly fills itself and you argue over the edge cases. If you would rather see how that scoring shakes out across a real site, we walk through it at replatformradar.com/migrate. But the tool is optional. The discipline is not.
What I actually do
I kill more than feels comfortable, and I have never once regretted it after launch. My rule is simple: a page reapplies for its job, and if I cannot say in one sentence what job it does, it does not get a desk. Not out of cruelty. Out of respect for the pages that are actually working, which deserve a fast, clean site that Google and the answer engines can read without wading through a decade of ghosts.
Do the sorting before the move. Because the alternative is unpacking every box in the new place at midnight, holding up a page nobody remembers writing, and asking the room the question you should have asked in the spreadsheet: what does this even do?