Recently we did an experiment to see if decreasing the font size on our blog would increase the time people spent reading our blog. This idea that by decreasing your font size you can get careful reading instead of scanning stemmed from a eye tracking study article.
From Direct Creative:
5. Smaller type promotes closer reading. This makes sense because smaller type is harder to read. So, to read it, you have to really focus. Larger type promotes scanning rather than reading.
Be careful with this one. No one is suggesting you shrink your web type to make it barely legible. I think the takeaway is to avoid making your type too big if you want close reading and avoid making it too small if you want to communicate rapidly.
So on March 21, we changed our font size from 19px to 16px.
I compared visitor activity on our blog from Apr. 5 - Mar. 22nd (“after group”) and from Mar. 7th - Mar. 21st (“before group”). To try to control the experiment, I excluded new posts and returning visitors. Each sample group consisted of visitor data collected over 15 days.
We didn’t have any increase in time spent on page but was able to reduce bounce rate.
The average time on page actually decreased but you look carefully you can see there’s an anomaly at March 24. There the “before” time was vastly more than the “after”. Since the “average time on page” figure averages the time across all visits and is not weighted by post, a hit post with a lot of traffic one day can significantly skew the data.
It’d be better methodology if the average time on site was averaged per blog post and then averaged again to get the overall change. This would reduce the effects of outliers.
If you look at the graph after that anomaly, it seems the “after” time is roughly on par with or slightly above the “before”. I don’t believe there is any real significance here worth further testing though.
Bounce rate measures how engaging your content is and if someone leaves your site without looking at another page. Since I don’t have good reason to believe decreasing the font size will decrease bounce rate, I’ll use a two-tailed test. Running the 2-tailed t-test on the numbers showed a p-value of 0.101 which isn’t very significant. Since the p-value (0.101) is greater than the significance level (0.05), we accept the null hypothesis that the two proportions are equal.
In conclusion, going from font size 19px to 16px in the Mercury typeface didn’t produce any real benefits.
Taige Zhang is a Product Marketer at Kera. You can follow him on Twitter at @taigeair.