Analytics: Hacker News v.s. a tweet from Elon Musk | Hacker News Reader

Analytics: Hacker News v.s. a tweet from Elon Musk

simonwillison.net

159 pointssimonw53 comments

Adding another data point here:

I had a very good post here on HN in the last year (about 1600 upvotes), and this was the only place where I shared it or saw it posted. I know a few people discussed it on Twitter, too, so a good fraction of my traffic was likely "sourced from Twitter" by these metrics.

Per Cloudflare, I had about 180k visitors during the 2-day period when that blog post was on the front page here, and another 60k during the following week. I don't have any trackers on the blog so I don't know the traffic sources or any other more granular breakdown. At the time, my baseline was about 400 visitors a day, which I attribute to RSS feeds and bots. I know that Cloudflare overestimates unique visitors, since it is just using layer 3-4 factors, but I'm not sure it overestimates by that large of a factor.

I would assume that the HN impact for his post was likely underestimated thanks to secondary sharing effects. A lot of links go from HN to Twitter, Reddit, and other communities, where they get second chances at "going viral."

What are layer 3-4 factors for unique visitors if you don't mind me asking?

not familiar with the term, but i assume layer 3/4 of the OSI model, so tcp/ip , that is if a single user connects through different network to your page, cloudflare count it as multiple users.

> Notable that sticking at the top of Hacker News for a day really does drive an enormous amount of traffic—18% of the traffic you get from the second most followed account on Twitter (looks like Barack Obama is still number one).

That's pretty incredible. I'd be really curious to know how this compares to being on Reddit's front page. I guess Twitter's click-through rate is just particularly poor (or maybe specifically Elon's is, maybe even the view count metrics are inflated).

It was a 2.37% click through rate. That is pretty solid, where do you know of that gets click through rates consistently higher than that?

lots of organic content on social media gets similar levels

It’s probably higher, given that Twitter has been miscounting the views metric (inflating it) for some time.

Tweets on private accounts with zero followers also track 40-50 views, for eg.

How is that 18% number derived? I can't seem to get 18% from the figures in the table above that bit.

They used 132k/721k. They either used the second number for "Direct / None" by accident or mislabelled the table.

Ah that makes sense, thanks. The 'Plausible' analytics link shows the same numbers so the table is not mislabeled, the 18% thing is very likely just a mistake. The number (if we accept all the other assumptions) is a bit under 7%.

Nice post, that is a very interesting breakdown of traffic:

Twitter: 721k (after elon tweeted him)

Direct / None: 132k (this includes traffic from Mastodon)

Hacker News: 49.5k (18th all time highest ranked post)

other: ~60k

Direct is probably also from chat apps and stuff. I remember that I posted the article in one of my groups.

Fascinating

The views are nice, but the time on page stats from social media tend to be horrible. One of Simon's screenshots shows over a million pageviews with 43 minutes of time on page. This isn't a long blog post, but it is fairly information dense. Out of 1 million unique views, this is only enough minutes for a couple dozen people to have actually read the entire post.

Most of my blog's traffic comes in from Google, and I most of my posts that see a reasonable amount of traffic will have at least average 1 or 2 minutes of time on page.

If I get a spike in traffic to a page from Twitter, the average time on page will drop to 1 or 2 seconds.

When someone comes in from search, they are seeking information about something specific. The folks dropping in from Twitter are infinitely more likely to immediately click on something else.

I don't know that I write much that would be of broad interest to the Hacker News audience, but I would be much happier to see traffic from here than from Twitter. I would bet Simon's 49.5k clicks from here got him way more engagement than 712k clicks from Twitter.

I'm very skeptical of that 43 minutes number. It doesn't make sense as a sum-of-all-time number, but it also doesn't make sense as a average-time number either.

I'd like to hear from Plausible about what that's meant to mean, because as it is I've been ignoring it as possibly a data collection bug of some sort.

I never have much trust in the time on page or time on site numbers. In your screenshot, the number sure looks like it is meant to be a total.

I use Matomo for my analytics. I think their little Javascript doodad attempts to report back time information after 15 seconds. Anyone who clicks away before that timer goes off is recorded as a zero.

I get a sort of warm fuzzy feeling when I see high time on page numbers, but zeroes never bother me. There are a huge number of reasons why a time can't be captured, but when a time is captured, it is almost definitely someone who had eyeballs on your page.

You're doing a good job, Simon! I would be thrilled to see a million clicks from Twitter on one of my posts. I am just more excited about your 50k clicks from Hacker News!

> Time on Page

> The average time people spend on a particular page on your site. This is calculated as the difference between the point when a person lands on a particular page and when they move on to the next page.

https://plausible.io/docs/metrics-definitions#time-on-page

"I use Matomo for my analytics. I think their little Javascript doodad attempts to report back time information after 15 seconds. Anyone who clicks away before that timer goes off is recorded as a zero."

Why in the world do they measure time on page like that? All they need to do is compare timestamps between hits...

There isn't always a second hit.

Plausible, like many other client-side analytics packages, by default only measures time on page for people who view more than one page. The time they report is the time between clicks.

So for big viral social media moments, the stat is basically worthless, since the vast majority of those visits bounce off the article page.

Simon, I wrote my intro philosophy class paper on your blog post. Submitted it yesterday. Super funny.

I hope you get an A :)

Will you post the paper somewhere?

Very interesting! But apart from visitor numbers, how high was the number of emails or requests that you received because of the wide circulation of that post? I could imagine that there was some interesting requests for comments?

I had a conversation with a journalist this morning (one I'd talked to before about other AI stuff), and I'm set to do a short live TV news interview at the weekend - but not much aside from that.

Fellow blogger here.

If that happened to me, I'd consider that to be a big deal. The fact that it's happened multiple times is awesome!

Here's the TV appearance! https://simonwillison.net/2023/Feb/19/live-tv/

Craziest to me was that 82% visited on mobile (with a 2:1 iOS:Android ratio). Perhaps this is a consequence of most visits being from Elon's tweet.

Also had a 2:1 desktop:laptop ratio -- would have guessed more laptops. (The author questions how Plausible determined this.)

The author put the Plausible logs on GitHub here: https://github.com/simonw/i-will-not-harm-you-unless-you-har...

> Also had a 2:1 desktop:laptop ratio -- would have guessed more laptops. (The author questions how Plausible determined this.)

For my personal analytics without third party dependencies for privacy reasons I use the fact that Javascript allows access to the battery API, so a laptop uses a desktop browser, with a desktop resolution and zoom, while still having a battery.

This is kind of in line with the statistics for one property I manage: 88% of visitors use mobile devices (26k visits/month).

I've heard several times that at this point most people do most computing on phones. Consumers increasing don't own desktops/laptops at all.

Interesting that both Plausible and CloudFlare are using ClickHouse as a backend:

https://github.com/plausible/analytics

https://news.ycombinator.com/item?id=24846300

https://github.com/ClickHouse/ClickHouse

I don't think OP is getting any money from plausible but wish they were - first thing I did was search them to see what it supports. Awesome advertising!

Sorta off topic, but how do some Substacks get so many subs? I seen posts go viral on Reddit, Hacker News, Twitter,etc yet the associated Substack posts have just 1-5 likes, yet other Substack blogs get hundred likes per post. Is it just friends and connections, or are external traffic sources not that useful? Where does the traffic come from if even front page Hacker News for 3+ hours is not good enough to get more than 1-5 likes?

Super interesting! Can you share the code you used to extract the metrics from the cloudflare graphql API?

Your site works great when permitting only first-party images/css/html. I use Firefox's tracking protection + ublock origin + umatrix. I don't allow first-party JS by default, or any third-party stuff. I spoof the referer. I see JS from plausible.io tries to load.

I wonder how that skewed the requests. Do you have data on that? I'm sure there are some % of HN visitors that do the same. Less likely the case for Elon Musk followers. I don't know how you would get 'time on page' data without JS.

The best data I have on that is the delta between Cloudflare numbers and Plausible numbers - but that's not quite the right comparison, because for Cloudflare I have all traffic to my site, not just traffic to that specific page (which I can get from Plausible).

What about device OS or browser data? User-agent should be enough for a meaningful breakdown.

People mainly use the twitter app. HN is just a website. I'd expect more desktop users from HN.

EDIT: oh I see. It's in the github repo and vercel app.

name visitors

Mobile 778,678

Desktop 101,216

Laptop 47,781

Tablet 16,967

Yeah the device data from Plausible is here: https://i-will-not-harm-you-unless-you-harm-me-first.vercel....

I think you can reasonably assume that the cloudflare traffic from that time is almost entirely headed to this page, minus your baseline number from other posts. The "total requests" will be completely skewed, but "unique visitors" should be okay.

A lot of the HN crowd seems to use adblockers and script blocking.

a 2.37% click through rate is far from ideal. Twitter analytics specifically does not mention clicks on links for this very reason. instead they simply track all clicks... links, profile, hashtags, etc.

That is what you would expect. An impression is merely having the tweet show up on your feed , even if you only see it briefly. Whether or not you read or click, etc is secondary.

I think that is a fantastic click through rate for something that passively shows on a feed. That's probably about as high as you can realistically get to be honest. Most posts get sub fractions of 1% click through rate.

you are wrong, sub 1% is closer to what you would get for an advertisement.

For any random feed link that was not requested, just passively shows up to everyone that happens to follow Elon, it is a fantastic click through rate. Please show some data on passive feed click through rates on social media sites.

Given that Twitter just changed the algorithm to artificially inflate Musk's tweets, I think a tweet from him could well be considered a paid advertisement at this point. A very, very expensive paid ad.

44 billion dollars expensive?

For what it's worth I would _kill_ for a 2.37% click through rate on my blogposts when they are linked on social media. Turns out most people on siloes like Twitter want to stay in the silo instead of actually reading things on people's own websites.

Actually very surprising to me. I am well aware of the click rate of advertising. But for something like this, I would expect closer to 10%. Turns out most people just don’t read the content. And headline dictate how our society are being informed.

How interesting to see the analytical breakdown. Thanks for sharing!

This only tells us that HN (and Mastodon) is a tiny beehive in a garden with the amount of traffic generated compared to the traffic generated by one tweet on Twitter by Elon Musk in a single day and puts the months of 'Twitter is dying' narrative to rest.

Mastodon comes no where near to that and has to be lumped up with 'other sources' and even separating that would amount to even less traffic than reported here.

Case closed.

Mastodon fits in the "direct / other" category because it deliberately opts out of referrer based analytics by including rel="noreferrer" on all outbound links.

Unless you manually add UTM tags to the URL, right?

I’m not so sure this is great news for Twitter. Twitter can drive 5x as many clicks as… a niche tech website that has 0 business plan. It essentially puts a cap on Twitter, the most active/popular account can drive less than 1m clicks, and you can’t reproduce that because that account is owned by a guy literally tweaking the algorithm to prefer his own posts. Now think about advertising where the click through is going to be an order of magnitude lower, and the reach is going to be orders of magnitude lower. It’s not really setting the world on fire…

plausible.io <3

Imagine had he 301'd that page to an Elon giveaway page after it was tweeted and then claimed he was hacked. he could have gotten away with it too. The economics and irrevocability of crypto , plus Elon's huge brand and credibility, create all sorts of new incentives. It's like the Bill Gates email scam of the 90s but multiplied by 1000x.

Reply on news.ycombinator.com