Score a page before you rebuild by grading it on three honest questions: does anyone actually want it (traffic, links, revenue), is it accurate and unique, and how much would it cost to make it right? Pages that earn on all three migrate. Pages people still want but that are stale, thin, or duplicated get a rewrite. Pages nobody visits, nobody links to, and nobody would miss get killed and redirected. You can do the entire pass in a spreadsheet, weeks before anyone opens a template.
Here's why I'm even bringing it up. I keep meeting teams who lifted-and-shifted every last URL because deleting things felt dangerous, then watched traffic sag anyway because they'd faithfully rebuilt 4,000 pages that were already dead weight. They moved the whole pantry, including the jar of something from 2016 nobody will admit to buying.
Think of it like reprinting the menu
A good kitchen doesn't reprint the menu by photocopying the old one. Someone stands in the walk-in and makes decisions. This dish sells every night and takes eight minutes: it stays. This one people order and love, but the recipe's drifted and the description is a lie: it gets reworked. And this one? Three orders last quarter, two of them sent back. That gets 86'd.
Your site is the menu. Migration is the reprint. Every page is a dish waiting to find out whether it survives the edit, and (this is the part people skip) the reprint is the one moment where cutting something is free. On the new platform, every page you carry over costs you a template that has to hold it, a redirect that has to point at it, and a reviewer who has to sign off on it.
The three questions that do the sorting
You don't need a data science team. You need three columns and the discipline to fill them in honestly.
The first is demand. Pull twelve months of organic sessions, inbound links, and whatever conversion or revenue signal you can attach. A page can earn its keep on any one of these. The About page converts nothing and ranks for nothing, but seven journalists linked to it, so it's not going anywhere.
The second is quality and uniqueness. Is the page accurate today? Is it saying something no other page on the site already says? This is where duplication shows up as the villain it is. If forty pages all say roughly the same thing, you don't have forty assets. You have one asset and thirty-nine dilutions of it, and machines reading your site can't tell which one you actually meant.
The third is effort. Not "can it be saved" (almost anything can) but "is it worth the hours." A page with real demand and terrible content is a rewrite. A page with real demand and good content is a straight migrate. A page with no demand and terrible content is not a rescue project. It's a jar from 2016.
| Signal | What it tells you | Leans toward |
|---|---|---|
| Strong traffic + links + accurate content | It earns its seat and it's ready to go | Migrate |
| Traffic or links, but stale or thin | People want it; the content lets them down | Rewrite |
| Near-duplicate of a better page | You're competing with yourself | Rewrite (or consolidate) |
| No traffic, no links, no conversions | Nobody would notice it gone | Kill |
| Orphaned asset nothing links to | A PDF in the attic | Kill |
Where the word rewrite hides the real work
Rewrite is the pile that lies to you. It looks like a decision, but it's really a promise, and it's the promise nobody staffs. Everyone happily migrates the good stuff and gleefully kills the junk. Then the rewrite column sits there with 900 rows in it, and on launch day those 900 pages ship exactly as they were, because the writers were never given the hours.
So be brutal with that column. If you can't name who is rewriting a page and roughly when, it isn't a rewrite. It's a migrate you feel guilty about, or a kill you're afraid to make. Pick one.
Killing a page is not deleting it
The word kill scares people because they hear "leave a 404 for Google to trip over." That's not it. Killing a page means it doesn't get rebuilt, and its URL gets a decision: a 301 to the closest living relative, or a clean 410 if there genuinely is no relative and you want the search engines to stop asking. What you never do is let it quietly vanish and hope nobody had it bookmarked.
This is the whole game, really. Scoring on paper is what lets your redirect map write itself, because every killed URL already has a note next to it saying where it should point. Do the scoring after you've built the new site and you're doing archaeology under deadline, which is exactly how the tough redirect calls get made at 11pm by whoever's still awake.
The tedious part is getting honest numbers for thousands of URLs at once: the traffic, the inbound links, the duplication, the orphaned files nobody remembers. That inventory is the thing a pre-migration crawl exists to hand you, so the argument you're left having is the human one about which dishes deserve a spot, not the one about where the data lives.
So what would I actually do?
Score everything before the platform is chosen, not after. I'd rather walk into a migration with a spreadsheet that already says migrate, rewrite, or kill next to every URL and defend those calls in a room, than carry 12,000 pages across because deleting felt rude. The reprint is the only free edit you'll ever get. Use it. Otherwise you're just paying a moving company to relocate a jar you were always going to throw out.