Data Solutions
The short version
Role Players Guild is a tabletop RPG reference site. Its article corpus landed in Google's index on June 17, 2026. By August 19 it was serving 1,318 pageviews and 658 unique visitors a day, entirely from organic search. No advertising. No social distribution. No newsletter.
That is 63 days, or about nine weeks, from indexed to peak.
I am writing this up because the mechanism is more transferable than the number. The interesting part is not that a site grew. It is that the Search Console data disagrees with how most people pick keywords, and it disagrees consistently enough to act on.
There is also a postscript, at the bottom, about what I did wrong. It matters more than the growth does.
The timeline, from Search Console
| Date | What happened | Evidence |
|---|---|---|
| May 20 | Domain earns its first impression | 1 impression, essentially no article content |
| June 17 | Article corpus lands in the index | Impressions on /articles/ URLs jump from 70 to 264 in a single day |
| June 26 | First double-digit click day | 10 clicks |
| August 2 | Crosses 100 clicks a day | 105 clicks |
| August 15 | Search Console peak | 211 clicks, 11,319 impressions, average position 8.9 |
| August 19 | PostHog peak | 1,318 pageviews, 658 visitors, 756 sessions |
The gap between the two peak figures is worth explaining, because conflating them is how traffic claims become dishonest. Search Console counts clicks from Google. PostHog counts every visit from every source. On the same days, Google was supplying a bit under a third of the total, with the rest arriving from Bing, DuckDuckGo, Brave, Ecosia and direct navigation. Both numbers are real. They measure different things.
What actually drove it
Narrow queries beat broad ones, by an order of magnitude
This is the finding I would keep if I could only keep one.
| Query | Clicks | Impressions | CTR | Position |
|---|---|---|---|---|
pf2e necromancer guide | 40 | 161 | 24.8% | 4.7 |
pathfinder 2e necromancer guide | 17 | 50 | 34.0% | 5.5 |
feats 5e | 3 | 1,287 | 0.23% | 9.7 |
The third row is the one to sit with. feats 5e delivered eight times the impressions of the other two combined and produced 3 clicks. It ranks on page one. It simply does not convert, because a person typing two generic words has not decided what they want yet, and page one is full of wikis and official sources that will satisfy them faster than a new site will.
pf2e necromancer guide is a different kind of phrase. It names a game system and a specific class, and the word "guide" states the format the searcher expects. There is exactly one document that answers it. Build that document and the click-through rate goes to a quarter of everyone who sees it.
The practical rule: a query that tells you what page to build is worth more than a query with a bigger number next to it.
Hub pages outperformed the guides they linked to
The two highest-traffic URLs on the site were not articles. They were indexes.
| Page | Clicks | Impressions | Position |
|---|---|---|---|
/articles/dnd-5.5/class-guides | 216 | 7,902 | 9.4 |
/articles/dnd-5e-2014/feat-guides | 203 | 9,874 | 10.5 |
/articles/pf2e/classes/necromancer | 162 | 1,848 | 6.9 |
/articles/tales-of-the-valiant/class-guides | 63 | 646 | 6.6 |
/articles/pf2e/feat-guides/skill-feats | 61 | 3,415 | 7.7 |
Three of the top five are category hubs. The reason shows up in the query data: people search the plural. "Class guides", "feat guides", "classes ranked", "tier list". A list page answers that phrasing better than any individual guide can, and it is the only page on the site that every guide in the category links back to, so it accumulates internal links no leaf page has.
Note the necromancer guide sitting third with a fraction of the impressions. It converts far better than the hubs do because its queries are narrower, which is the previous section restated. The hubs win on volume, the leaves win on rate, and you want both.
Completeness is what makes a hub work
A hub page that lists eight of twelve classes is a worse page than one listing all twelve, and not by a third. It is worse in kind. Somebody arriving from "dnd 2024 classes ranked" wants the full ranking. Four gaps means they leave and open the wiki instead.
This is the argument for publishing a category as a set rather than drip-feeding it. It is also, as the postscript explains, the argument that I took too far.
What this does not prove
I want to be exact about the limits here, because case studies that skip this section are marketing.
The niche was soft. Tabletop RPG character options are a subject with passionate readers and very few people building careful reference pages. Several of these queries had no good result before this site existed. A commercial niche with funded competitors does not behave this way, and nothing in this data suggests it would.
Nine weeks is not a template. It depended on publishing a large corpus at once into an uncontested space. A site adding pages steadily into a competitive market should expect quarters, not weeks.
Volume was doing real work. Roughly 1,300 articles is a lot of surface area, and some of the growth is simply the arithmetic of having more pages that can rank. Pretending the strategy alone did it would be false.
The postscript, which is the actual lesson
On August 22, three days after the peak, Google traffic to the site fell 96 percent. Impressions went from 14,203 to 828. Average position went from 8.9 to 49.7. It has stayed there.
I did that. The publishing rate that built the corpus kept going after it should have stopped, and a site adding pages that fast eventually looks like exactly the thing Google's scaled content policies exist to catch. The growth strategy and the failure have the same cause, which is why this article would be dishonest if it ended one section earlier.
Two things are worth saying about it.
The first is that the traffic did not disappear. Bing, DuckDuckGo, Brave and Ecosia grew across the same window, from about 275 to about 410 pageviews a day, and the site still serves roughly 450 organic visits a day with Google essentially absent. A site that loses its largest channel overnight and keeps most of its audience was built less fragile than it looks. That was not deliberate, and I am not going to claim it was.
The second is that I am working on the recovery now, and I will publish what happens either way, including if it does not work. That is the follow-up to this article and it will be more useful than this one.
What to take from this
If you are building a content site, the three findings that transfer are: target queries specific enough to name the page you have to write, build the hub before the leaves and complete the category, and judge keywords by click-through rate at position rather than by search volume.
The fourth finding, which cost more to learn, is that a publishing rate is a constraint and not a target. Decide what yours is before the growth curve talks you out of it.
All figures in this article come from Google Search Console for theroleplayersguild.com and from PostHog, covering May 20 to September 4, 2026. Traffic figures describe mid-August 2026.