The HN Effect in Numbers
blog.charleshooper.net
blog.charleshooper.net
One reason I want to know that is that for my own sites, I've generally found a large proportion of link-blog traffic to be very shallow, to the point where maybe one in 100 hits is a "real" hit. This is particularly the case at link-blogs without any summary blurbs. Slashdotters who don't want to read the article, for example, often just read the blurb and don't bother to click through, so a bigger proportion of those who do click through are at least somewhat interested in reading. A much larger percentage of people on sites like Reddit click through, but I don't think a much larger percentage actually read the articles; many are just clicking through to figure out what the article's about, since there's no blurb to explain it. And some don't even do that much, since a common Reddit use pattern is "middle-click on every article to open in new tab, then work through the tabs", often culminating in just closing the remaining still-unread tabs when you're done reddit-surfing, which results in a ton of spurious hits that involved no human eyeballs.
Anecdotally, the proportion of shallow-hits/not-even-real-hits goes up somewhat proportionally to traffic. Front-page of reddit delivers maybe 100x as much traffic as front-page of a more niche sub-reddit, but much less than 100x real readership.
Feedburner (now owned by Google) tracks stats on # of subscribers and is kind enough to let you see the history. In my case, I started out with about 10 subscribers and reached 44 on launch day. Today, I'm looking at 24 subscribers.
It has 76 votes, so that's pretty close to the 10 HN votes -> 1k uniques rule of thumb
Of those 7157, 602 clicked on at least one of two links at the bottom of the blog post to do more on the site, but I'd be surprised if more than a handful return in the future (other than thru another blog post).
The article I wrote seems to have come pretty close as well with ~9k uniques @ ~100 votes
Around 3500 uniques, from net +35 score (50 upvotes, 15 downvotes), of which around 200 clicked through to something else on the site (even if only loading the domain's root page), and around 100 clicked through to at least one other article. Seems like ratios in the same ballpark.
Unfortunately, it's still hard to tell how much of that is substantial engagement. I can imagine a decent number of people skimmed the first article, clicked to see what else was on the site, skimmed another article, lost interest and left, especially since reading reddit puts you in a sort of "idly click on things" mood, at least in my experience. I did get one email about the article with some thought-provoking comments, so that puts a lower bound of one engaged reader. ;-)
A pinging stats provider, like Clicky, gives you great insight into single-pageview engagement.
It's true that there's room for inaccuracy in the pinging measure. Some viewers with third-party (or all) JavaScript disabled won't be counted, viewers that leave the page open in a tab before they look at it may be over-counted, and readers who legitimately spend more than ten minutes reading a longer page are capped at the ten minute mark. However, averaging that across thousands of views is a very worthwhile measure of engagement.
I had a relevant post about that here recently (which was castigated for semantics more than it was actually discussed, unfortunately): http://news.ycombinator.com/item?id=1294969
The Clicky pinging data was invaluable then too. HN visitors to that post show an abysmal 95% bounce rate and an average 14s time-on-site in Google Analytics. Meanwhile, Clicky shows those same HN visitors spending an average of 3m29s time-on-site and truly bouncing from the page only 12.9% of the time. Even if you assume Clicky's numbers are wildly optimistic, it's nice to have that alternate insight.
http://news.ycombinator.com/item?id=1620333
Interesting tidbit, that is still going, it seems that the HN posting caused a bunch of other places to pick it up and link to it.
I've gotten closer with f(x) = 2658e^(-0.94)+30, but as x approaches infinity (or even 5) it starts to full apart. Any one good with the maths who wants to take a look?
I decided with this one to throw out the first datapoint (the day my site launched) since it was likely still growing during this period of (overall) decay. How significant is this?
The actual numbers: [2688, 1065, 452, 206, 138, 105] My function's numbers: [2688, 1068, 436, 188, 91, 54]
For easy playing: http://www.wolframalpha.com/input/?i=table[+2658e^%28-0.94x%...]
Hacker News may not generate as many clicks as Digg would, but visitors from here are way more interesting to me.
edit: It looks like the higher the original is voted the fewer people will visit through HN daily, which suggests that people use it more to see if they've missed something than to look at just HN daily.
edit2: > I really had no idea how many people would end up using it, but it seems that there's a few people at least.
That's a curious statement, can't you tell from the logs?
Probably referrer logs are one of your main sources of new visitors.
In raw numbers, yes. But I don't know how much traffic HN gets, so the raw numbers don't tell me much.
http://abiekatz.com/2010/08/20/introducing-a-new-theory-the-...
It got 5 karma points and 29 page views from Hacker News.
So making the front page is very important for getting traffic from HN. I am sure this is the case for any social news site.