Google Penalty Recovery – Watch community website (+$31860/mo, 386% ROI)

Introduction

One of the largest community websites in the “Watch” industry, with ~85000 organic visitors daliy was hit by google manual actions.

Story

The first note we got was at the beginning of March; afterward, the organic traffic on this site dropped to 25%. The letter from Google had some sample URLs, and we quickly figured out the problem (outgoing spam links to adult content). Taken a few hours to investigate and a few more hours for devs to fix it. At the end of March, we sent an appeal, which was approved.

In 2 months (in May), we got the same manual action again, but there were no sample URLs in the letter. So we did colossal work to save the website, cleaned up all wrong outgoing links, found and cleaned spam topics, blocked hundreds of users, and implemented anti-spam technologies backed by Machine learning and Artificial Intelligence for natural language processing.

We crawled more than 5 million URLs to complete this work. After cleaning up everything, we sent an appeal again (end of May). This time it took about 2 months for Google to reply, and we only got a successful message in July. 

Challenge

Some challenges we had:

• crawling such a considerable amount of pages and analyzing each piece of content (in total approx 100,000,000) generated by users for being non-spammy,
• studying external links to get an idea are those links sending users to adult/spammy websites or a legit one (in total approx 3,000,000 of external links analyzed). 
• challenge to crawl external websites due to their WAF (web application firewall) or Cloudflare or any other antibot systems restrictions, which we overcame with the help of using proxies and a javascript-enabled browser.

Solution

• we created an algorithm for analyzing external links to get an idea (automated) is site adult/non-adult, spammy (affiliate)/non-affiliate, and implemented it.
• created an algorithm for analyzing user behavior and messages to flag a user as a good community member, suspicious, or evil. Which finally led to the analysis of “evil” and “suspicious” users for their content. Some users were blocked, and some were identified as hacked (and we changed their passwords).
• addiitonally to the user and external links analysis – we also implemented a simple but effective text analysis tool by using the most popular anchor texts that spammers use, which also flags messages with this “black word list” as suspicious and sent to the moderator queue for a manual check-up.

Result

All these actions led to a successful recovery from google’s penalty and recovery in traffic. On average, we recovered lost 25% of traffic and generated even more (28% total increase), which led us to 18000 new organic traffic sessions per day.

ROI

As the site monetizes using ads, selling premium subscriptions, being affiliated with some watch shops and some additional on-site services, the approximate ARPU (average revenue per user) is $0.059

Where 18000 organic sessions * $0.059 = $1062 in daily revenue.

After deducting our hourly and developer costs, it led to approx ROI for the first month of 386%, and after the first month, no more work was needed, and the site got its revenue of $31860/monthly recovered.