What eight weeks of Crit Pick data actually show

Eight weeks in, Crit Pick has 44 clicks and 5,050 impressions. Four weeks ago it had 18 and 1,650. So the honest one-line version is this: the numbers roughly tripled, they are still tiny, and the most interesting thing is not the total, it is which pages moved and which did not. One evergreen guide quietly carried the whole site. The single biggest bet I made has not earned a single click.

Crit Pick is an SEO case study I run in the open for SeeIndie, and also a games project I had wanted to build for years. The experiment is real because the site is. This is the second checkpoint, and unlike a snapshot, four more weeks is just enough data to start seeing slope instead of noise.

Data through 29 July 2026 · Google Search Console

The number that matters is the slope

A single month of traffic on a new site tells you almost nothing. Two months lets you draw a line between two dots, and a line is the first thing worth talking about.

Here is the line. At the four-week mark the site had 1,650 impressions and 18 clicks, sitting at an average position of 11.1. Four weeks later it is 5,050 impressions and 44 clicks. Impressions roughly tripled, clicks a little more than doubled. For a brand-new domain doing nothing but publishing low-competition pages and linking them together, that is the shape you want: not big, but bending upward.

Here is the same thing from a different angle, which makes the direction harder to argue with. Look at just the most recent 28 days on their own: 3,541 impressions and 27 clicks. That single most-recent month did more than double the impressions of the entire first month, and it holds more than half of every click the site has ever earned. The line is not only pointing up, it is getting steeper.

Now the part most write-ups would quietly skip. Over the same stretch the average position drifted from 11.1 to 12.1, and click-through fell from 1.1% to 0.9%. If you only read the averages, you would think the site got slightly worse while I was not looking. It did not. What happened is that a whole new cluster of pages, mostly Palworld, got indexed during the window and entered ranking around positions 20 to 45. Averaging a pile of fresh page-three pages into the mix pulls the mean down even while the pages that were already ranking climbed. The total went up. The average went down. Both are real, and if I showed you one without the other I would be selling you something.

And this was not an accident I am explaining away. It was a choice. I am deliberately building out the Palworld side, and my rule is that every single Pal gets its own page, because that is how the search actually splits: people look up one creature at a time. The catch is that a space like the Paldex is already thoroughly covered by big, established sites, and my brand-new pages have no authority to lean on yet, so they land on page three and sit there for now. In the short term that is page after page dragging my average down. In the longer term it is the only way to cover the topic properly. I made peace with the trade.

What compounded: one evergreen guide

If you want a single picture of why the method is worth continuing, it is the RimWorld animal tier list. In month two it had 174 impressions and exactly 1 click. Today it has 992 impressions, 16 clicks, and an average position of 9.6, which is the top of page one. That one page is now more than a third of the whole site's clicks.

What faded: the timely posts

The opposite pattern is just as clear. The demo review that led the leaderboard in month two, a timely piece riding fresh interest in an upcoming release, has gone almost flat: 7 clicks then, 7 clicks now. Impressions crept up as it kept appearing, but the click count did not move, because the moment it was built for has passed.

This is exactly what I predicted last time, so I will take the small win: news content spikes early and then sits there as a nice one-time bump, while evergreen content starts slow and keeps climbing. Eight weeks of data now shows both curves on the same site at the same time. If you are a solo dev deciding where to spend a weekend, that contrast is the whole argument for writing the guide over chasing the news.

The bet that has not landed

Here is the one that stings. In month two I made a point of the GW2 event timer as the cluster's front door, the one high-volume utility page worth fighting for at roughly 12,100 searches a month. The idea: a useful tool page pulls in traffic that a niche guide never could, and funnels it into the rest of the cluster.

Eight weeks later that page has 62 impressions, zero clicks, and an average position of 17.8. Page two. It only started showing up in search at all around 20 July, so it was crawled and ranked late, and it has a long way to climb before anyone sees it.

I want to be careful about the lesson, because the tempting takeaway is “tools do not work” and that is not what the data says. The lightweight GW2 class quiz, also a tool rather than a guide, ranks fine: position 8.8, a 2.9% click-through, the best click rate on the site. So it is not tools versus guides. The real reason is simpler and less flattering to my month-two optimism: that search is already owned. There are established, well-linked event timers players have used for years, and a two-month-old domain has no business outranking them yet. A page like that is not won by existing, it is won by first building enough authority in the surrounding cluster that Google trusts you on the topic at all. I treated it like a page I could just publish and rank. It is a page I have to earn my way up to. The front-door idea might still be right. The order of operations was wrong.

The title problem is still a title problem

Last checkpoint I flagged the RimWorld lore page as the clearest to-do on the site: lots of impressions, almost no clicks, which is a title-and-snippet problem rather than a ranking problem. It had 538 impressions and 2 clicks.

It now has 956 impressions and 3 clicks. A 0.3% click-through rate, which is worse than it was. More people are seeing it, almost none are clicking, and I did not fix the title the way I said I would. The honest reason is not strategic: I was in the middle of a house move, and a title rewrite on a page that already ranks kept losing to real life. That is the unglamorous truth of running a project like this alongside everything else, and pretending the gap was some clever wait-and-see would be exactly the kind of tidy story this whole experiment is meant to avoid. The job did not get done, the data noticed, and now it is on the record until it is fixed.

Internal linking: the deep dives lead, the hubs lag

One structural thing you can only see once there is enough of a site to compare. The deep-dive guides are on page one. The hub pages they all link up to are stuck on page two and three. That is the reverse of what people assume happens.

So the next checkpoint has a clear thing to watch: do the hubs finally start climbing, or does the theory need a rethink?

The honest verdict, so far

Directionally, the method is working. Roughly three times the impressions and more than double the clicks in four extra weeks, all of it from difficulty-10-to-30 long-tail pages, with one evergreen guide proving that the compounding is real and not just hopeful. The misses are just as clear, and each has a plain reason.

None of that is failure at eight weeks, but all of it is the kind of thing a case study is supposed to say out loud instead of cropping out of the screenshot. Forty-four clicks is still nothing. This is a slope, not a result. The reason to keep reading is that you get to watch it either keep bending up or flatten out, with the method laid bare enough that you can judge it for yourself.

One last honest thing, because it colours all of the above. Crit Pick is a games project I actually care about, not just a test rig for SeeIndie, so not every page here is an optimization play. Some exist purely because I wanted them to, and a page made out of real interest is never fully optimized. That is almost what makes it real. I think that is a feature, not a leak. A site that only ever ships the keyword-perfect page reads like one, and its numbers are the easy kind to stage. This one has pages that are just mine, made because they mattered to me and not because a tool told me to.

I would rather hand you honest data off a project I love than clean data off a machine.

What is next

Three concrete jobs before the next checkpoint.

I will publish the next set of numbers here the same way, what climbed, what stalled, and what I got wrong.

Common questions

Sources: Google Search Console (clicks, impressions, CTR, position) for critpick.me, eight-week view and the last-28-days export; the earlier checkpoint post for the four-week comparison; own keyword research for the event-timer search volume.