5 ms·

HackerNews Front-Page Funnel Analysis

33 points by tansey 16y ago

petenixey 16y ago
When the "Dry erase HOPA" slideshow was revealed as a myth, the subsequent interviews with the Resig brothers revealed how they always got two waves of traffic from such hoaxes. The first is the hoax traffic, the second wave is the wave of traffic about the hoax story. I'm interested to see the increasing frequency of the "HN stats followup article" describing the numbers from the traffic to the first article. Despite the frequency this may still be an under-utilised property of HN. Why not follow the recursion down the rabbit hole - and tell us how many signups come from the followup article too? Perhaps bonus points for publishing a chart of signups v. derivative index of article? I jest but honestly only slightly.
- tansey 16y ago
  > Why not follow the recursion down the rabbit hole The thought did cross my mind. ;) Seriously though, we plan on focusing the blog on analyzing press releases, political speeches, and other text currently making headlines. We just were so excited by the HN response that we thought this would make a great "Hello World" post. I also think it's the only funnel analysis follow-up for a HN submission; most usually focus just on page views and unique visitors.
vessenes 16y ago
This is fascinating to me, thanks for the data, and congrats! One interesting comparison I have is the day I got my first web company's first slashdot post, roughly 2000, maybe late 1999. I don't still have the logs, but we had significantly more clickthroughs, on the order of 50 to 100k if I recall correctly. Given the HN 'pageview per day' numbers I've been reading, I find the clickthrough rate surprisingly low; especially because the way I read HN is to click through to almost everything on the homepage. Other thoughts on this? It seems like a techcrunch or (even? slashdot) article would be worth more, which surprises me. Well, techcrunch doesn't surprise me. I guess I'm wondering what size / demographic community we're all in here.
- tansey 16y ago
  Thanks! Glad you like the article! My advisor from grad school had a blog post on the front page as well, with similar vote counts. He tracked about 5,000 UV total for the 48 hour period it was popular (24 hours on the front page). So the views seemed in line with my expectations. I think the value of HackerNews is not in the quantity but the quality of users. I only had two spam/garbage posts on my blog, and no spam signups that I noticed. At its core, this community is a set of really passionate, generally positive early-adopters.
- Swizec 16y ago
  A lot of people come to HN for the commentary, not so much for the post. What I've noticed is that I will usually read only the comments and only click through to the page if it seems exceptionally interesting. That is to say, the opinion of the HN crowd on a topic, matters more (and/or is more useful) than the topic itself.
  - vessenes 16y ago
    Yes, I agree with this use case for HN completely, and it ties to my experience. Compared with say slashdot in 2000, interactivity is easier and generally more interesting here. At the time, there were very few social aggregators, so the use cases were different.
leenewton 16y ago
Nice analysis... sidenote: If you're measuring the emotional influence of content on people, shouldn't it be "affectcheck"?
w1ntermute 16y ago
> In descending order, the top browsers used by readers were: Chrome (52.97%) Firefox or Mozilla Compatible Agent (25.18%) Safari (23.71%) Internet Explorer (3.89%) Opera (2.79%) Wow, that's pretty crazy - I didn't know Chrome had such a high market share. Is this an accurate reflection of the browser market share on HN in general?
- tansey 16y ago
  I just realized looking at these numbers that I used the # pageviews per browser / total UVs to calculate those percentages. So if you're wondering why they add up to > 100%, that's why. Just understand that the actual numbers are roughly proportional to those I provided in the article.
- acangiano 16y ago
  Chrome is so high because, arguably, nothing comes even close as far as browsers go. I tried to use FF after using Chrome for months, and I simply couldn't go back. I'm willing to bet most HN folks are in the same boat.
zyfo 16y ago
Good write-up. The unaccounted percentage can be chalked up to simply “other.” What does minus 8.54% other web browsers mean?
- tansey 16y ago
  Sorry about that, as I noted in another comment, I used page views in the numerator and unique visitors in the denominator. I've updated the article with the appropriate percentages.
Chrono 16y ago
Nice write-up! +1 for Monty Python reference. :)

33 points by tansey 16y ago

When the "Dry erase HOPA" slideshow was revealed as a myth, the subsequent interviews with the Resig brothers revealed how they always got two waves of traffic from such hoaxes. The first is the hoax traffic, the second wave is the wave of traffic about the hoax story. I'm interested to see the increasing frequency of the "HN stats followup article" describing the numbers from the traffic to the first article. Despite the frequency this may still be an under-utilised property of HN. Why not follow the recursion down the rabbit hole - and tell us how many signups come from the followup article too? Perhaps bonus points for publishing a chart of signups v. derivative index of article? I jest but honestly only slightly.

tansey 16y ago
> Why not follow the recursion down the rabbit hole The thought did cross my mind. ;) Seriously though, we plan on focusing the blog on analyzing press releases, political speeches, and other text currently making headlines. We just were so excited by the HN response that we thought this would make a great "Hello World" post. I also think it's the only funnel analysis follow-up for a HN submission; most usually focus just on page views and unique visitors.

vessenes 16y ago

This is fascinating to me, thanks for the data, and congrats! One interesting comparison I have is the day I got my first web company's first slashdot post, roughly 2000, maybe late 1999. I don't still have the logs, but we had significantly more clickthroughs, on the order of 50 to 100k if I recall correctly. Given the HN 'pageview per day' numbers I've been reading, I find the clickthrough rate surprisingly low; especially because the way I read HN is to click through to almost everything on the homepage. Other thoughts on this? It seems like a techcrunch or (even? slashdot) article would be worth more, which surprises me. Well, techcrunch doesn't surprise me. I guess I'm wondering what size / demographic community we're all in here.

tansey 16y ago
Thanks! Glad you like the article! My advisor from grad school had a blog post on the front page as well, with similar vote counts. He tracked about 5,000 UV total for the 48 hour period it was popular (24 hours on the front page). So the views seemed in line with my expectations. I think the value of HackerNews is not in the quantity but the quality of users. I only had two spam/garbage posts on my blog, and no spam signups that I noticed. At its core, this community is a set of really passionate, generally positive early-adopters.
Swizec 16y ago
A lot of people come to HN for the commentary, not so much for the post. What I've noticed is that I will usually read only the comments and only click through to the page if it seems exceptionally interesting. That is to say, the opinion of the HN crowd on a topic, matters more (and/or is more useful) than the topic itself.
- vessenes 16y ago
  Yes, I agree with this use case for HN completely, and it ties to my experience. Compared with say slashdot in 2000, interactivity is easier and generally more interesting here. At the time, there were very few social aggregators, so the use cases were different.

leenewton 16y ago

Nice analysis... sidenote: If you're measuring the emotional influence of content on people, shouldn't it be "affectcheck"?

w1ntermute 16y ago

> In descending order, the top browsers used by readers were: Chrome (52.97%) Firefox or Mozilla Compatible Agent (25.18%) Safari (23.71%) Internet Explorer (3.89%) Opera (2.79%) Wow, that's pretty crazy - I didn't know Chrome had such a high market share. Is this an accurate reflection of the browser market share on HN in general?

tansey 16y ago
I just realized looking at these numbers that I used the # pageviews per browser / total UVs to calculate those percentages. So if you're wondering why they add up to > 100%, that's why. Just understand that the actual numbers are roughly proportional to those I provided in the article.
acangiano 16y ago
Chrome is so high because, arguably, nothing comes even close as far as browsers go. I tried to use FF after using Chrome for months, and I simply couldn't go back. I'm willing to bet most HN folks are in the same boat.

zyfo 16y ago

Good write-up. The unaccounted percentage can be chalked up to simply “other.” What does minus 8.54% other web browsers mean?

tansey 16y ago
Sorry about that, as I noted in another comment, I used page views in the numerator and unique visitors in the denominator. I've updated the article with the appropriate percentages.

Chrono 16y ago

Nice write-up! +1 for Monty Python reference. :)