One article,
taken apart.
This is a real post shape, not a paragraph blob with a headline on top. Tap any part of the list to see where it sits in the page.
How To Rank In 2026: A Four-Month Case Study
We published three posts a week for four months and tracked every one. Organic traffic went up 312%, but not evenly, and not for the reason we expected.
What follows is the whole experiment, including the two points where the data made us change our minds. If you are weighing whether cadence beats craft, this is the closest thing to a controlled test we could run on our own site.
What actually moved the needle
Our assumption going in was that length would win. Long, exhaustive guides rank and short posts do not. That is the received wisdom, and on our site it turned out to be mostly wrong.
What compounded instead was structure. The posts that climbed were the ones wired back into our pillar page on content operations. Every new article added one internal link, and the pillar absorbed the authority. Eight inbound links over four months moved that single page from position 34 to position 6. None of the posts feeding it cracked the top twenty on their own.
Pages carrying five or more internal links from newer posts pulled 6.4× the organic sessions of pages with none.
Source: margot.co analytics, Jan–Oct 2026
That ratio held across every cluster we tried, including the two we expected to fail. It is the single number that changed how we plan a quarter.
The three-post rhythm
Three a week is not a magic number. It is simply the slowest cadence at which our clusters closed fast enough to matter. At two a week, a ten-post cluster takes five weeks to finish, and by the time the last post lands the first one already reads as dated.
The rhythm does something less obvious, too. It forces you to write shorter. When you owe three posts, you stop treating each one as a monument. Our average length fell from 2,900 words to 1,850, and the shorter pieces outperformed, not because search engines prefer them, but because we actually finished them.
What it costs to keep the rhythm
The honest constraint is not ideas. It is production. Three posts a week is twelve a month, and twelve a month is either a full-time hire or a retainer that eats the budget the content was supposed to earn back. We ran a retainer for two months before doing that arithmetic properly.
| Channel | Cost / article | Articles / mo |
|---|---|---|
| Agency retainer | $420 | 4 |
| Freelance pool | $180 | 9 |
| Ghost.blog | $0.80–1.00 | 28 |
The traffic curve lagged the publishing curve by about eleven weeks. Nothing happened for the first two months, which is the part nobody warns you about and the reason most teams quit in March.
The two months where nothing happens
Weeks one through eight produced almost no measurable change. Sessions drifted within their usual range, a handful of posts landed on page four, and one of them briefly outranked our homepage for a term nobody searches. It felt exactly like failure feels.
The temptation in that window is to intervene: rewrite the titles, chase a different keyword set, add a hundred internal links in an afternoon. We did the first two in week five and it cost us a fortnight of clean data, because we could no longer tell whether a movement came from the cadence or from the edits.
If you take one operational rule from this, take that one: decide the plan, then leave it alone long enough to read the result. Eight weeks of discipline is cheaper than eight weeks of guessing.
Google's own helpful content guidance says much the same thing in fewer words: write for people, publish consistently, and show first-hand expertise. That advice has barely changed since 2022. What has changed is how quickly thin pages get filtered out before they ever reach page one.
What we would do differently
Two things. First, we would build the pillar page before the cluster rather than after. We spent six weeks pointing links at a page that was not ready to receive them, and those six weeks bought us nothing.
Second, we would leave old posts alone for the first quarter. Every hour spent refreshing a 2024 article was an hour not spent closing a cluster, and no refresh moved a single ranking until the cluster around it was complete. Refreshes compound later; they do not start anything.
If you are starting from zero, pick one topic you can plausibly own, write the pillar first, then publish around it until the cluster closes. The order matters more than the volume.
The short version
Cadence beat craft, but only because cadence let us finish clusters, and finished clusters are what the internal links were pointing at. A faster schedule with no structure underneath it would have produced twelve orphan posts a month and nothing else.
So the order is: pick the topic, build the pillar, then publish around it often enough that the cluster closes inside a quarter. Hold the plan for eight weeks before you judge it. Keep the posts shorter than feels comfortable. And measure the pillar, not the posts. The posts are the scaffolding; the pillar is the building.
Frequently asked
Eleven weeks for us, measured from the first post to the first sustained position change. Expect nothing for two months, and resist adjusting the plan inside that window, because the noise is louder than the signal.
It works, it is just slower. One post a week closes a ten-post cluster in ten weeks instead of three. If nobody in your category is publishing at all, weekly is plenty.
Cadence is not the risk; indifference is. Every post here answered a question a real customer had asked us. If you cannot name the person who asked, do not write the post.
Opt-in features are switches in your brand profile. Flip one on and every article after it carries the change.
