To decide whether a page should migrate (come over untouched), get a rewrite (keep the address, fix what's inside), or be killed (redirect it or let it 404 and never speak of it again), you score four things that exist long before any template does: how many humans and machines actually reach it, whether anyone links to it, whether it earns money or leads, and whether it's just a worse copy of a page you already have. Everything green means migrate. Real value trapped under bad writing means rewrite. Nobody home means kill.
Picture yourself standing at the door of the new site. It's launch night. There's a velvet rope, a clipboard, and a line of pages stretching back around the block, every one of them convinced it belongs inside. Your job is not to be nice. Your job is to decide who gets in.
And here's the part everyone gets backwards: this happens before you argue about the CMS, before the design system, before a single component gets built. You do not rebuild 4,000 pages and then wonder which ones mattered. You work the door first, hand out three stamps, and only then start building for the guests who made the list.
The three stamps you're handing out
Migrate is the easy one. The page pulls its weight, the content still holds up, and the sensible move is to carry it across with its URL intact and get out of the way. Don't gold-plate it. Don't "improve" it because you're feeling ambitious at 11pm. Bring it in and pour it a drink.
Rewrite is the page that clearly matters but shows up looking rough. People still land on it. Other sites still link to it. It converts. But the copy is five years stale, or it's thin, or it reads like it was written to hit a keyword density target from 2014. The address is worth keeping. The words are not. So you keep the URL and you fix the human bit.
Kill is the one people can't say out loud. It's the page nobody has visited since a router firmware update three summers ago. It has no links, no traffic, no conversions, and it duplicates something better three folders over. You are not deleting a memory. You're declining to move a broken lamp into a house you just cleaned.
The four signals that decide the stamp
You can score every one of these without touching a template, because they all come from data you already have or can pull in an afternoon.
| Signal | Where it comes from | What it's telling you |
|---|---|---|
| Reach | Analytics organic sessions, plus server logs for crawler and AI-bot hits over 12 months | Does anyone, carbon-based or otherwise, actually arrive here? |
| Authority | Referring domains and inbound links to the exact URL | Has the wider web decided this address is worth pointing at? |
| Value | Assisted conversions, revenue, form fills, downloads tied to the page | Does it do a job for the business, or just take up shelf space? |
| Uniqueness | Content similarity against the rest of the site | Is this the real thing, or a near-duplicate of a better page? |
Score each one, and the stamp mostly picks itself. High reach, high authority, still fresh: migrate. High authority and value but the content's a duplicate or stale: rewrite, because the links are worth keeping even when the words aren't. Zero across the board: you already know.
The trap is scoring on reach alone. A page with barely any traffic but forty referring domains is not a kill, it's a rewrite you'd be foolish to lose. And a page with decent traffic that's a straight duplicate of your best guide isn't a migrate, it's a consolidation. Read the row, not the one column that flatters your gut.
The pages that lie to you at the door
Some pages are very good at looking important. A high-traffic page that's actually your 404 template getting hammered by dead links. A PDF from 2019 with more inbound links than your homepage that no human has opened since the quarterly report it announced. An orphaned page with real backlinks and no internal path to reach it, quietly leaking authority into a room with no doors.
These are the ones a spreadsheet of pageviews will wave straight through, and they're exactly the ones that come back to haunt a launch. Working the whole line by hand across thousands of URLs is where the plan quietly dies, which is the boring problem a pre-migration crawl exists to solve: it scores every page and asset on those four signals at once so the door moves faster than one clipboard ever could.
Do the scoring before the CMS conversation, not after
Here's why the order matters more than the criteria. Every hour you spend on the door is an hour you don't spend rebuilding pages that were never going to get in. A migration where you kill even a fifth of the inventory up front is a migration with a fifth less to build, test, redirect, and re-test. The kill list is not the sad part of the project. It's the part that makes the deadline survivable.
So do it first. Not because it's tidy, but because building templates for pages you're about to delete is the most expensive way to procrastinate ever invented.
What I'd actually do
I'd run the door before I let anyone open a design file. Score reach, authority, value, and uniqueness for every URL, stamp each one migrate, rewrite, or kill, and treat that list as the scope of the whole project rather than a chore to get to later. When it's genuinely close, I lean toward rewrite over kill for anything with real inbound links, because you can always trim a guest list later but you can't get those links back once you've slammed the door on them.
The pages don't get a vote. That's the whole point of standing at the door.