Case study · 2026
COCA
A search content program that learned which pages bring new readers, and stopped the ones that didn't.
April to September 2026. Sources: Google Analytics (visitors who accepted analytics cookies), Google Search Console, RankFuel publishing records and DataForSEO.
RankFuel's two formats on COCA's blog, April to September. News stopped in July.
Fewer, more relevant appearances: impressions fell from about 230,000 a month to 29,000 as news stopped.
ChatGPT, Perplexity and similar; visits that began on a blog post. Small, and more than doubling.
About COCA
COCA is a fintech card and wallet app, with a loyalty programme and a refer-a-friend scheme. Its priority markets include Germany, Italy, Portugal and Thailand.
The brief
Most of COCA's blog readers already knew COCA: they arrived from Telegram, the app and messages. The job was to reach people who don't know COCA yet, through Google and AI assistants, and to learn which pages actually move them.
What we did
Salo ran COCA's search content with RankFuel, Myosin's SEO and AI-search engine, which has written COCA's guides and news since February. Then he measured what each kind of post really did.
- Kept what works. RankFuel's 107 card guides kept readers for 25 seconds and earned 3.2 Google clicks per post, more than four times the news rate. The best of them is the page Google shows for “coca card fees”.
- Stopped what didn't. News posts drew volume, not readers: 10 seconds on average, 0.7 Google clicks per post, and 70% earned none. They stopped in July, and the share of people who clicked from Google rose from 0.2% to 1.0%.
- Found the next wins. People search for COCA's bank, token, rewards and card fees with no dedicated page to land on. Those answers are the cheapest search wins available, and the Q4 plan writes them.
- Closed the tracking gaps. The report found that the app-store click event had stopped recording in mid-May, and that Analytics kept detailed data for only two months, with the fixes for both.
From here, a monthly scorecard measures every new page at 28 and 56 days, and checks how AI assistants answer ten questions people ask when choosing a card.
Where people stay in charge
COCA's own posts about the product stay COCA's: they're the ones that send people to the app. The engine writes the search guides that reach people earlier.