What title length is effective?
I’ve been working with Social Actions data for a while, and I think I may finally have something small and perhaps useful. Social Actions data is comprised of volunteer and donation opportunities (actions) that people around the web are searching for, and, occasionally, clicking on. I ran an analysis of the hits vs. the length (word count) of the titles.
Or so I thought! Turns out that the first time I posted this all I was doing was counting the number of hits *total* for social actions of a particular title length. What really matters is hits per action. If you ever post something vaguely statistics related and then go to bed, thinking about someone like Nate Silver reading what you wrote should be enough to make you get out of bed to fix it.
Here’s the incorrect first one (using total hits):

Compiled from Social Actions data
And the second: (average hits)

Average Hits vs. Title Length
Well, the first one is certainly prettier. The first one seems to indicate that 3-4 words is the best, which is the wrong conclusion to draw. The second one is messier and seems to show that between 10-20 word titles, results are a bit better than outside that range. The data starts getting a little crazy after 25+, but I think at that point outliers start to matter a lot more. If you only have 3 actions in the system with title length of 27, say, and one of them has a lot of hits…
After doing the second one I realized that there is potentially a second problem: actions that have been in the database for a long time are bound to collect more total hits. At the moment, we don’t know when the hits occurred, so it’s difficult to gauge a hit rate without filtering the actions by the duration of their posting. And even if we do that, we might filter groups of long-running actions like donation campaigns.
Did we learn anything here? Well, I learned that deceptively simple statistics are difficult to get right. We might have learned that titles between 10-20 words long in social actions posts are a little more likely to get clicked on. From all the above we should conclude only that more study is necessary!
By the way, my method was to use the PHP functions strip_tags(), then str_word_count() to determine the number of words in the title.












