Knowing how many readers a blog gets is pretty important for any blog writer. We like to pretend it doesn’t even matter, put on a face of “Oh well, I’m just writing this for myself, you know, to get the word out, I don’t really care if anyone reads it”

What a bunch of hipsters!

Paragraph funnel

tl;dr –> Only ~30% of readers actually read a whole blog post! 16% if it’s particularly long, but 95%+ make it from one paragraph to the next on average.

Yeah right dear blogger, that’s why you keep posting links to your blog in your twitter stream, that’s why you submit it to various link sharing sites and that is surely why you are writing all of your stuff in public as well. Writing for yourself just wouldn’t feel the same if it was done privately in a random notebook would it?

Sometimes I manage to get a couple of thousand people to come upon a blogpost I have written and … then what?

Problem is that the usual analytics tools only count hits. That is, how many people came onto your blog. Recently Google Analytics started changing that a little bit by also counting how many users perform some sort of social action. This is a pretty good measure of how good people think your post was, or even simply how interesting it seemed.

However, that’s only a small part of what I care about. I want to learn how good a writer I am, how well do I captivate a reader and keep them reading? Surely the most important mark of a good post is how many people actually read it!

Enter Mixpanel, stage left.

With this in mind I set up a simple javascript that sends an event to Mixpanel every time somebody scrolls past a paragraph of text. To this day I have never actually taken the time to look at this data, but at the beginning of this month I wrote a very lengthy blogpost.

1700-ish words, 58 paragraphs long.

That post received 3445 hits, 5 G+, 38 tweets and 3 facebook likes.

EDIT: The post was “I went through YC as an intern, here’s what I learned

Right, not a total throwaway post. People actually liked it! Yay!

But here’s where it gets interesting. Out of all those people only 574 made it to the end. Somewhere in all that text I lost over 80% of my readers!

Wow, that can’t mean anything good for my writing skills. 1700 words is about five pages in a standard book format. Now imagine if I was writing an actual book … nobody would make it to the end!

Bleeding readers like a boss

Looking at the data in more detail it looks like only 80% of the users read the first ten paragraphs, and 70% read the first 20 paragraphs. Not too bad at all.

Then something strange happened in this particular post, only 41% of the readers made it to the 22nd paragraph. This happens to be right next to the photo of some drawings on a shower wall and the title of this section is “Work hard, party harder” … I guess people don’t like partying and showers?

After that it settles down, takes another 20 paragraphs to lose another 10% of readers. And another jump, plomps down to just 23% readers making it to the 40th paragraph. This is around the spot where I say that being only an employee in a startup is much easier than being a founder. I guess people don’t like hearing that either?

There are no more significant jumps after that, but just over 15% of the readers made it all the way to the end and for that, I thank you awesome 574 people who decided my blog was worth reading through to completion.

Although interestingly enough, outside those jumps (yes I checked, other posts ending at those paragraphs doesn’t account for the jumps) on most paragraphs more than 96% of users make it from one paragraph to the next.

Had another hit on the 3rd of August. The data there is much more level, with 33% of readers making it to the end of a 30 paragraph post. But interestingly enough, only 61% made it from first to second paragraph!

I think I need to start experimenting with this in a more organized manner. Got my baseline, about a third of my readers make it to the end of a normal length post. Now it’s time to figure out how to increase that.

PS: if you’re interested in the JavaScript that makes this magic happen, just check out the source.

Enhanced by Zemanta
  • http://twitter.com/WD_Katie Katie Watkinson

    I like writing blogs! And I do read other peoples blogs……sometimes.

  • David T Macknet

    I think that attributing drop-offs like this may not be quite accurate.  I mean, yes, there will be people who quit reading when they read something they don’t like, but people also quit reading because they’re reading on a screen which displays a painful-to-read font (e.g., whichever font this comment box uses), or simply determine that your article doesn’t match up with the attention-grabbing title, or have no terribly great interest in the issue and find that they’ve been reading for too long, or readers who see the length of text-gone-by and where the scroll bar is and don’t remember that comments figure into the length of the page.  All manner of things may lead people to leave.

    You’ll even have (as I just did) users who do their reading in one browser (with the security locked down) and only pull open a javascript-enabled browser if they want to comment.  So, I’m counted twice, perhaps, and I “read” every single paragraph in your article in less than 1 second when I visited the second time.

  • http://profiles.google.com/whalebot.helmsman helmsman helmsman

    For me it is not a moment decision to drop reading something.  Before this I lose focus and scroll down automatically. So reason to drop is not located in the most “bleeding” paragraph, it is located earlier

  • theChess92

    I scrolled to the bottom of this page just because you might have some javascript tripper.

  • Dustin Whitney

    just the first paragraph for me

  • justin

    how many times did you count me when I had you open in a browser tab for a month and have a redneck OS which demands daily reboots, meanwhile browser smart enough to restore my scroll points in 100+ tabs on daily browser reopening

  • Cymen Vig

    What is wrong with your fonts/css? http://imgur.com/m6Nt0

  • Eli Huebert

    My guess is that people got to “Work hard, party harder” where the end of a normal article would be and realized they were only halfway through. I’ve done that quite a bit when I only really care enough about a topic to give it  5-10 minutes of my life but it turns out the author has churned out seemingly unlimited reams of  text.

  • Tim

    This is a sweet use of Mixpanel.  Thanks for using the product =)

  • Anonymous

    Nice hack! Would be good to see how it works on varied types of text. (Usually after two scrolls I start fast-scanning, to see if the author highlighted their intended takeaway at the conclusion, or if they just wanted us to travel awhile with them… ;)

    From your uptop summary it sounds like each paragraph causes an average 5% audience loss, but I’m not sure this is supported by the bodytext… can you confirm/deny?

    tx, jd/adobe

  • Alex

    Where did you get this from “tl;dr –> Only ~30% of readers actually read a whole blog post! 16% if it’s particularly long, but 95%+ make it from one paragraph to the next on average.”

  • http://twitter.com/crisfole Chris Pfohl

    I love waypoints!  Nice move.

  • Foljs

    He summed his findings –those that he explains in toto in the following 10 paragraphs…

  • http://swizec.com Swizec

    It’s a sweet product. Wouldn’t be able to do this without the lifetime account though :)

  • http://swizec.com Swizec

    That’s probably what happened. I think the two drop-offs correlate to two kinds of readers who had different expectations/tolerances for length.

    But this requires further testing to make sure.

  • http://swizec.com Swizec

    The font looks unreadable when it isn’t anti-aliased and a lot of (mostly Windows users) people were complaining that they can’t read. So I added a text-shadow as a quick fix.

    Would love to hear a better suggestion.

  • http://swizec.com Swizec

    To be honest, I have no idea. But for this particular analysis I only looked at a single specific day of data.

  • http://swizec.com Swizec

    Yep, there are a whole lot of variables. That’s why I said at the end of the post that this metric needs some solid verifiable testing (one variable at a time).
    Right now I just got a general feeling for what’s going on.

  • http://swizec.com Swizec

    It is in fact supported by the body text you did not read :P

  • http://swizec.com Swizec

    Yep, it’s an awesome awesome library!

  • Anonymous

    I hope Alex was being funny. Otherwise…

  • Anonymous

    I’d be willing to pay (say, $0.50 / article) for these types of statistics about each of my writings.  Thanks for the code (I might very well try it out), but you might consider (reconsider?) evaluating this as a product.

    My articles tend to be fairly technical (http://www.hizook.com/), which creates a problem: how to write for the non-technical casual reader while still appealing to the highly-technical target audience.  These sorts of metrics would really help me tailor the articles’ scope, placement of summaries vs details, etc.

  • http://twitter.com/zidarsk8 Miha Zidar

    you should also time, how long it took someone to scroll down. Say someone scroll over 1700 words in 5 seconds, I would pretty much doubt he actually read anything. Thats me mostly cause I am usually too lazy to read walls of text.

  • Pingback: Make Your Blog Post Work For You « The Writers Help YOU with Editing and Ease Your Life

  • http://mybyeworld.wordpress.com/ Imogen Shepard

    And here I thought I was doing OK with my new Blog with the amount of hits I have been getting.!!