Score every page as migrate, rewrite, or kill before you build a single thing, and you do it from a spreadsheet of real signals: traffic, inbound links, conversions, duplication, and whether a machine can actually read the page. Migrate the pages that earn and still work. Rewrite the ones that earn but have rotted. Kill the ones nobody visits, nobody links to, and nobody would miss. The call is a data decision, not a taste decision, and it happens on paper, not in a sandbox.
Picture the night before a house move. Every room is now piles on the floor. There's the load the truck pile, the this needs fixing before it goes anywhere pile, and the pile by the curb for whoever drives past first. You are not negotiating with the truck about square footage yet. You're deciding what deserves a seat.
A content migration is the same floor, except the stuff is pages and the house is your domain. And the single most expensive mistake (I'll say it now, before someone emails me) is renting the truck first and sorting later.
Why you sort before you build, not during
Here's the thing about deciding mid-build: every page you haven't scored becomes a tiny argument. Someone on the team feels the careers page matters. Someone else remembers a blog post that did well in 2019. The template is half-built, the clock is running, and now you're relitigating a 4,000-page site one hunch at a time. That's how a 2,000-page site becomes a 2,000-page site on the new platform, debt and all, because nobody had the nerve to curb anything.
Sort on paper and the arguments happen once, early, cheaply. A page either clears the bar or it doesn't. The spreadsheet doesn't have feelings about the careers page.
The signals that actually decide the pile
You need enough data per URL to make the call without opening the page and squinting. In practice that's a handful of numbers from analytics, your backlink tool, and a crawl. Put them side by side and most pages sort themselves.
| Signal | Where it comes from | What it tells you |
|---|---|---|
| Organic sessions (12 months) | Analytics / Search Console | Does anyone arrive here on purpose? |
| Referring domains | Backlink tool | Does killing this page snap someone else's link? |
| Conversions / assisted conversions | Analytics | Does it do a job even with low traffic? |
| Duplication / near-duplication | Crawl | Is this the tenth version of the same thing? |
| Last meaningful update | CMS / crawl | Is the content stale or just old? |
| Machine-readability | Crawl / render check | Can Google and answer engines read it at all? |
That last row is the one teams forget. A page can have traffic and links and still be a problem if its content only exists after a pile of JavaScript renders, or if it's a thin wrapper around a PDF. That page isn't a clean migrate. It's a rewrite wearing a migrate costume.
Reading the piles
Once the numbers are in front of you, the three piles fall out roughly like this.
Load the truck. Decent traffic, real inbound links, converts or supports conversions, reads cleanly. This page works. Move it as-is, keep its URL, map its redirect, and don't touch a word. Most teams have fewer of these than they assume, which is itself a useful thing to learn early.
Fix before it goes. The page earns, or has links pointing at it, but it's stale, duplicated, or unreadable by machines. The value is real; the execution has rotted. These are your rewrite pile, and they're where your writers' hours should actually go, instead of being spread evenly like margarine across every page in the site.
Curb it. No traffic, no links, no conversions, no unique reason to exist. Kill it. Not everything gets a redirect either: a dead page with zero inbound links and zero traffic can return a clean 410 and nobody will ever notice, which is the whole point. (Resist the urge to 301 everything to the homepage. That's not forwarding the mail, that's throwing it in a drawer.)
The grey zone is pages with links but no traffic. Those links are equity you'd be setting on fire. Usually the answer is redirect the URL to the closest surviving page, even if the old content dies. The link lives; the dead weight doesn't.
Make the score a number, then stop arguing
The trick that makes this survive a real project is turning the signals into one score per URL, with a threshold everyone agreed to before they saw which of their favorite pages fell below it. Agree on the rules while it's abstract. Apply them when it's personal. A pre-migration scan can hand you this whole sheet, scored and sorted, so the build team inherits a decision instead of a debate. If you want the broader workflow it sits inside, our migration guide walks the steps around it.
Do this and the build becomes boring, which is the highest compliment a migration can earn. The developers get a finite list of pages that definitely ship. The writers get a short list that definitely needs work. Nobody discovers in week nine that the truck is full and the good furniture is still in the hallway.
So should you score on paper before you touch a template? Yes, flatly, every time, and I'll die on this hill. The build is where good content goes to get cloned; the spreadsheet is where bad content goes to die quietly. Do the dying first. Your future self, standing in a half-empty new house with only the furniture that earned its spot, will thank you for leaving the broken lamp at the curb.