Last fall, a customer shared the following with me:
"I know we have great content in our forum. A lot of it ranks really well in Google. But most of our users are drive-bys: They come to us through Google Search, look at a page, and leave. Is there anything we can do to increase engagement and hopefully also boost ad revenue?"
Later that week, our team came up with an idea: What if we could give forum users a list of suggested searches based not just on popular keywords, but the most meaningful topics in a discussion? What would happen to engagement and ad revenue?
That same day, we built a prototype of what we called More Like This (MLT) using natural language processing (NLP). "NLP is a way for computers to analyze, understand, and derive meaning from human language in a smart and useful way." (Algorithmia) In our case, we used NLP to create a Threadloom-generated post inside a discussion thread, extract key entities from the discussion, and suggest them to the user as searches. Here's what it looked like:
The following week, we launched an A/B Optimizely experiment the following week. Based on three weeks of data with MLT enabled on half of web traffic, here's what happened when people clicked on MLT:
In short, users doubled their engagement, eCPM went up over 400%, and ad revenue on a per-session basis went up over 700%.As you can imagine, we're pretty excited about this! But there's still work to do. CTR on MLT was <5% – not bad for a few days work, but we think we can improve that with more time on the user interface. So over the next several months, we'll be tweaking MLT and running a few more experiments. Stay tuned!