Growth September 29, 2020

I tracked & analyzed 13,159 Hacker News posts over the last two weeks. Here's what I learned. (Interactive Charts)

Dan Ankle @ankleio
  1. 4

    I forgot to mention, the feature length article with interactive charts can be found over at my blog here:

    Otherwise, the tweet storm is a more succinct summary.

  2. 3

    I've done a similar analysis. It's interesting to learn that most submissions that reach the front page are put there by mods, not by algorithms - you can tell by noting that at the time of reaching the top spot most of them have no upvotes and no comments, and come from posters with barely any karma.

    Given that I think that data such as when the top posts were posted only tells you when top posters are active, not what is the best time to do it. Sometimes a post will just stay there, inactive, for a few hours, and then suddenly reach the top when the mods wake up. Moreover sometimes mods will put a thread to the front multiple (three) times. I think that the only thing the algorithm does is decide which of the top positions a thread will take.

    In general there are two types of articles: bumped ones and non-bumped ones. The bumped ones get to and you get a dofollow link. non-bumped ones just stay in and are nofollow. Around 10% posts become bumped.

    It seems like Hacker News is more like a moderated newsletter than a community social news site.

    1. 1

      "It's interesting to learn that most submissions that reach the front page are put there by mods, not by algorithms" - I don't think this is necessarily true. It probably happens somewhat but I am not sure it's "most".

      Unfortunately, it seems getting to the top is pretty gameable via upvoting through friends, but staying there and not getting flagged by a mod is also a part of the challenge.

      70% of posts were unique posts over the two weeks (meaning one account only posted one submission), so I don't think top posts only tells you when "top posters are active". However, I do agree in that it tells you when posters are active in general, which is probably indicative of traffic on the website (although I can't be sure).

      Perhaps I could make a chart which shows the most popular days/hours for voting (to see if it's different).

      1. 1

        Your refresh time is 2 minutes, maybe there are some front page upvotes before you have a chance to fetch the data, thus your high average. Also see how many submissions with no upvotes reach one of the 9 or 10 top pages. Around 8% reach the top1 as you found, but around 12% reach the top 10 - why?

        Also track the position of an article over time, especially one with no upvotes - take an article from /newest, and as often as possible check which position it's on. Often you'll see something like 8 hours at -1, then suddently for example 62, then gradually falling to the bottom, then -1 again. Maybe the algorithm randomly picks some articles from /newest and gives them a chance? But then how does it pick which 10% of the articles get bumped? I'm sure it's not the number of upvotes.

        I believe upvotes and comments only influence how long an article stays on top, up to a total of 3 days, in rare cases 4. Also too many comments in relation to upvotes will cause it to be flagged for flame war. I'm happy to be proven wrong, but until you can show a recreatable method how to reach the front page then none of this data is useful - it's just a fun project.

  3. 2

    Very cool analysis Dan, keep up the good work. I also enjoyed reading your post "I re-wrote this headline 9 times."

  4. 2

    Well done Dan.

    How many votes does a post need to reach the front page of Hacker News in 2020?
    5.22 is the average number of votes a post needed within an average time of 27.1 minutes to reach the front page for the first time.

    I'm not sure if I get it correctly. Does it only takes 6 votes per 30 mins to stay in the front?

    1. 3

      I think he means it takes 5.22 votes whilst in the "newest" section to actually get featured on the front page by the algorithm

      1. 2

        Yep, that's exactly it.

        The first time posts were seen in the featured section, they had an average of 5.22 votes and took an average of 27 min to hit the front page.

        From here, they received significantly more to stay at the stop, but I didn't include this in the stats. It seemed in most cases, once a post got to the front, it snowballed and received a lot more votes, securing it's place.


        1. 1

          Yeah, thanks. I don't know why I had the impression of it requires a load of upvotes to get into the front page. 5.22 seemed so small that made me wanna try.

          And now I know the best time to try it ;)

      2. 1

        I see. reach the front page, thanks.

  5. 1

    Love the data here! Very interesting. Thanks for taking the time to do this.

  6. 1

    Nice analysis. Just yesterday one of my side projects unexpectedly reached HN front page. And it was in the first position for a while. The best thing is that I didn't even submit, it was picked up by someone else.

    This is a screenshot of early FP reach

    Since the link from HN was pointing to my Github Repo, I didn't get direct traffic to the website. But still, I got a lot of visitors. This is a moment when the US was sleeping.

    Overall it landed ~8k UV to the website, coming trough the Github.

    And on Github, it landed ~14.5k unique visitors with 21.7k page views. It also brought ~1.1k extra stars on the Github repo, which resulted in total of ~1.6k stars.

  7. 1

    Cool analysis and very insightful, thank you for sharing! I've always heard HN is a great source of traffic, but I've never tried posting there... It would be interesting to get a breakdown of links by topic... to see which topic themes seem to resonate with their audience... is this data available?

  8. 1

    I've always suspected that weekends was the best time to post because that's when I've had most success. It's good to see that analysis supports that idea.

    Awesome work!

  9. 1

    This is agood example of data visualization.

    Congrats for the good work!

  10. 1

    Good stuff 👍

    Very thorough analysis. I was looking for something like that the other day but couldn't find anything useful.

  11. 1

    That was really insightful. Followed you on Twitter!

  12. 1

    Another good analysis - thanks Dan!

  13. 1

    Last time I was researching this, there were human editors who were deciding whether or not to put something on the front page. What about now?

    1. 1

      This maybe the case. No post was ever seen on the front page without a minimum of three votes (but these votes definitely could have come from mods).

      Quite a few posts were flagged, once being featured. I probably should have presented some data on this.

Recommended Posts