CoinStats fixed a Core Web Vitals bug, tripled search impressions

DebugBear, a Core Web Vitals monitoring vendor, publishes a small library of named customer case studies. One of them covers CoinStats, a cryptocurrency price and portfolio tracking site, and walks through a single root cause, a single fix, and a reported before-and-after change in search performance. Unlike the Rakuten and Vodafone A/B tests Google itself ran, this one is a vendor-published customer story rather than a controlled experiment, and it is worth reading with that distinction in mind.

What was actually broken on CoinStats’ pages#

DebugBear’s write-up says CoinStats “was doing poorly on Google’s Core Web Vitals” and had a high bounce rate, and that the recurring issue across pages was a slow Largest Contentful Paint. Using DebugBear’s tooling, the team traced the cause to how images were served: CoinStats “used to include a lot of Base64 images in the HTML document” instead of linking to them through a normal img src URL. Base64-encoding an image inflates its size by roughly a third and forces the browser to download it as part of the HTML response itself, before the page can start rendering, rather than as a separate, cacheable request.

What CoinStats changed, and what DebugBear reports it did#

The fix was narrow: replace inline Base64 images with conventional img src URLs across the site. DebugBear’s case study quotes CoinStats’ Senior SEO Specialist, Rafayel Begoyan, saying “we’ve updated around 15,000 pages and saw some nice gains from it,” and that the change “cut in half” the HTML document’s download size. DebugBear reports the outcome in two figures: the share of “Good URLs,” a Search Console Core Web Vitals metric, and search impressions on the affected pages both “increased 3x since implementation.”

Why this is a directional data point, not an audited rate#

DebugBear’s page does not publish a start date, an end date, the raw impression counts behind the 3x figure, or the actual before-and-after Core Web Vitals scores for the pages involved. It is a vendor case study built to sell a monitoring product, told from the customer’s side, with one named quote and no independently reproducible numbers. That does not make the underlying claim false, embedding fewer Base64 images in the HTML is a well-understood way to speed up Largest Contentful Paint, but it does mean the “3x” figure should be cited as CoinStats’ own reported comparison, not as a rate any site removing Base64 images should expect to match.