How tiny changes in typography have dramatically increased time on the page

almost 2 years ago from , front-end developer @ flawlessapp

Hi, I am a front-end developer and I started diving into web typography and found out that by changing font styles on a webpage you could achieve massive results. For example, by decreasing letter spacing for titles and increasing them in the regular text the Reader becomes less aware of size content and could actually start reading it.

Also, I realized that justified alignment is not the best case for your sections, because browsers are not handling the alignment very well. So, simply use text-align: left; and stop thinking about random white space between the words.

Experiment continued for a month and the traffic was equally the same. The results are pretty interesting:

  • old version — avg. time on page before typography improvement = 1:38min
  • current version — avg. time on page after improvement = 3:37min

Why did that happen? What do you think?


  • Duke CavinskiDuke Cavinski, over 1 year ago

    My theory is that this whole post is a clever ruse to promote your product. ;)

    21 points
    • Thomas Michael SemmlerThomas Michael Semmler, over 1 year ago

      Considering that their "old version / current version" - Links are completely misleading, I would categorise this as spam tbh.

      -1 points
    • , over 1 year ago

      O_o sorry if my question made you feel uncomfortable! The product is not public yet, so nothing to promote there... and when it will be available it should be in free beta (as far as I know).

      My whole idea is to learn more around my initial area, which is web\front-end dev. So I want to do small experiments by myself and check results.

      0 points
    • Vl Berezovsky, over 1 year ago

      hey! As I mentioned to Thomas already, I didn't have some bad intentions.

      I read Rhythm in web typography and decided to improve the page.

      The product is in development and I have finished my part with the web version. While I have some free time, I want to do more self-development. Web design is interesting to me. Tweaking typography is a fast and interesting task. My team is very supportive, so I can experiment with it.

      We have some traffic, so why not to learn in practice?

      0 points
  • Clarissa H., over 1 year ago

    very surprised there was such a difference between the 2 versions. what was your sample size?

    9 points
    • , over 1 year ago

      it was 20 work days experiment:

      ~1000 uniques page visitors in the first 10 days

      ~800 unique page visitors in the next 10 days

      I didn't make A/B testing as traffic split might be small. But I compared the results without changes and with. Nothing except typography was changed.

      0 points
  • vadim mikhnovvadim mikhnov, almost 2 years ago

    the Reader becomes less aware of size content and could actually start reading it

    That's just a theory though. i.e. have you considered that maybe you just significantly decreased readability and now people spend more time processing your copy?

    8 points
    • Jim SilvermanJim Silverman, over 1 year ago

      there's not a significant visual difference; seems more likely that there's just outliers skewing the averages. i'd be curious to see the number of users involved in this experiment.

      0 points
      • , over 1 year ago

        it was 20 work days experiment:

        ~1000 uniques page visitors in the first 10 days

        ~800 unique page visitors in the next 10 days

        no specific ads, or promo. It was the same organic traffic

        0 points
    • , over 1 year ago

      ooo... I haven't thought about it. My initial idea was that I got a good improvement because people spend more time on the page!

      0 points
  • Adam Fisher-CoxAdam Fisher-Cox, over 1 year ago

    Honestly, I don't buy that kind of a jump from these changes - what was your sample size? Have you kicked the tires on this and tried to figure out all the ways this could possibly be bad data?

    5 points
  • Alex JohnsonAlex Johnson, over 1 year ago

    Averages are not great at determining time-based info as a single outlier can totally skew the data.

    If one person spent significantly longer in the "current version" group (e.g. if they spent 30 minutes reading this page) then the avg will be way different than a tighter distribution with no outliers.

    Even with similar traffic, low and high outliers can lead to false insights when looking at averages.

    Median is usually more insightful than avg in this type of data set. (but I like your data-informed thinking here. Just needs some tweaking and caution.)

    4 points
  • Jan SemlerJan Semler, almost 2 years ago

    It is hard to read. You cannot see words seperately. It is like one big word per block. Also Vadim might right, they take longer because it is hard to read. So is the conversion up or down?

    Also didn‘t understand why some designers using spacings in their designs so massivly. Fonts are created in a way so you shouldn‘t change letter spacing at all. if you do it because it might look better, then you doing something wrong. It is another case if you have a weak font and need to make adjustments for readabilty. But to use for visual esthetics or for performance improvements i doubt that it has any effect.

    1 point
  • Alex HoffmanAlex Hoffman, almost 2 years ago

    How is more time spent a good thing? At one point Facebook actually changed their algorithm to deliver more important updates in your feed to the top, the result was people spent less time checking their feed. Even though the user benefited from this, advertising did not so they changed it. For you, I would compare time spent in relation to whether or not they signed up for early access. Check out something like mixpanel for more helpful metrics

    1 point
  • Jared CJared C, over 1 year ago

    This is interesting, definitely a unique but important thing to try tweaking. I think stripe does something like this with their typographic treatments.

    1 point
  • Fırat Demirel, almost 2 years ago

    I absolutely believe that text size and font type is very critical on reading, but we need a better data-backed story to reach your results. The pages that you shared are not text heavy pages btw. I guess you made some mistakes in the first version of your page...

    0 points
    • , almost 2 years ago

      it was 2800 unique visitors:

      ~1000 uniques page visitors in the first 10 days (old typography)

      ~800 unique page visitors in the next 10 days (new typography)

      Do you think it's not enough data?

      0 points
  • S LloydS Lloyd, over 1 year ago

    Well, you have applied wrong shadow on the last CTA button :) But the current version is surely easy to read.

    0 points