The Experiment I Ran by Accident

Between the 22nd of June and the 6th of July, the only thing I changed on one client’s website was twenty-one links. Not twenty-one pages, not twenty-one rewrites — twenty-one in-content links between pages that already existed, wired in across three batches with wording the client had approved. A label printer in Suffolk, forty-five years in the trade, sixty-six pages of WordPress. The content queue was deliberately parked that fortnight, so nothing else touched the site. No new pages, no title changes, no directory submissions. Just the links.

I didn’t set out to run an experiment. But that’s what I ended up with, because scientists have a name for what happens when the world accidentally arranges itself into a clean test — a natural experiment. One closed set of changes, one defined window, and a ready-made control group: all the pages the links didn’t point at.

Then the recrawl came through, and I pulled the full position data across the thirty-seven keywords I track for this client. The pages the links pointed into moved. A security term searched 1,300 times a month went from position 19 to position 9 — page two to page one. A supplier term went 19 to 9 as well. The niche flagship went back to number 1. Elsewhere: 12 to 5, 11 to 6, 26 to 8, and an industrial term from 19 to 8, onto page one. Seventeen improvements in all, against eight small dips — and the dips were scattered across pages the links never touched, which is exactly the noise you’d expect from Google on any given fortnight.

That control group is the part I keep coming back to. If everything had risen, I’d tell you Google simply liked the site more that month and the links were a coincidence. But the untouched pages barely moved while the linked-into pages jumped. The change and the result line up too cleanly to wave away.

Here’s why I think this matters for you, whatever you sell. Internal linking is the lever that comes free with every website you’ve ever paid for, and it mostly sits unpulled. The industry will sell you new content at hundreds of pounds a page, and backlinks at worse, because those are the levers that come with an invoice attached. The lever with no invoice attached rarely comes up. Yet the hard part — pages that rank at all, pages with something worth saying — you’ve already done. The links just tell Google which of your own pages matter most, in Google’s favourite language: a link inside real sentences, from one relevant page to another.

It’s the same argument I made in the power of retrofitting — that making what you already own work harder beats building something new — except this time it comes with a receipt. And if you’re wondering how I know what moved rather than guessing, that’s the boring discipline of tracking rankings against real search data week after week, so that when something shifts, I can say what shifted and when.

So here’s the version you can do yourself this week. Pick the three pages on your site that earn you money — the ones a customer lands on just before they call or buy. Then find the pages you’ve already written that talk about the same subject: the blog post, the FAQ, and the case study you’d forgotten you had. Link from those pages into your money pages, inside a proper sentence, with wording a human would actually click — “our custom labels for medical suppliers”, not “click here”. Ten or fifteen links like that is an afternoon’s work, it costs nothing, and on the evidence of this accidental experiment, it can move a page-two ranking onto page one inside a fortnight.

One correction before I go. Last week I told you I’d be away from the 2nd of August to the 18th of September — six and a half weeks off, which even I thought was generous when I re-read it. The dates got mangled at my end: the real break is a fortnight, Saturday the 5th of September to Saturday the 19th. The promise underneath it doesn’t change — the monitoring runs every day whether I’m at the desk or in the sea off the north of Scotland.