Spikes from StumbleUpon and Digg Fall by 50% on Second Day

As mentioned before, I received two spikes in page views for my blog (credits to the US Census image). For this post, we take note of a striking similarity between the two spikes.

Blog Stats Two Spikes
The first spike (3/21/10) is caused by StumbleUpon, the second by Digg (3/26/10).

First, each spike lasts for roughly two days. But the more interesting fact is that, even though Digg brought in 3.5 times as many viewers than StumbleUpon on the first day, both had a near 50% drop on the second day. The percentage of views on second day to views on first day was precisely 50% on StumbleUpon and 48% on Digg.

This seems to imply one thing: StumbleUpon and Digg are fundamentally the same. They would seem to be different, for they have different modes of operation, i.e. random, recommended pages from a button versus lists sorted by time or popularity, and they have different user bases. The stats, however, show a disturbing similarity.

I wonder whether other sites that get hit by StumbleUpon or Digg experience this phenomenon too.

Digg Stats: Hits vs. Page

Yesterday I wrote a post on The Impact of Digg, and Digg vs. StumbleUpon, which involved stats on site interaction rates. Today, I shall look at Digg only, and how the page number of the Digg website impacts the number of hits. Basically, if a site is on the first page of Digg, it should get a lot of views. When it gets pushed to the second page, the number of views should decline, as only a fraction of the people who viewed the first page would click to the second page. Same goes for the third, fourth, fifth pages.

WordPress stat tracking allows me to see the number of hits from each separate url. And I see hits from digg.com, digg.com/page2, digg.com/page3, etc. So, what is this fraction of users that go beyond the first page? Below is a table of the number of hits from each Digg page url. This gives the number of people who clicked to my blog from the first page, the second page, etc. As I am compiling the data right now, the site is on page 11 of Digg, so I shall only go up to page 10. Also, note that there are many ways to access a site on Digg than directly from the front page or numbered pages, so the total views below will not add up to anywhere near 100k.

Page Hits Ratio to previous page
1 14855 N/A
2 5036 33.9%
3 1447 28.7%
4 1247 86.2%
5 641 51.4%
6 299 46.6%
7 385 128.8%
8 179 46.5%
9 196 109.5%
10 133 67.9%

Here are the corresponding graph for the Hits column:

Digg Stats: Hits vs Page

And for the last column:

Digg Stats: Hit Ratio vs Page

Quite a few surprises here! I shall try to comment in order.

First, from page one to two, we see that the clicks generated on page two is only a third of that generated on page one. That is, you’ll get about three times as many hits from the first page of Digg than from the second page. Roughly speaking, this also means one-third of users who go to the front page of Digg will continue to the second page.

From page two to three, there is an even larger drop—only 28.7% of those on page two continue to page three. But, from three to four, there is barely a drop. In fact, the number of hits generated by the fourth page is 86.2% of that generated on the third page. There are several ways to explain this, and each one most likely plays a role:

  • Dedication: Users having the energy to click to page three will probably have the energy to click to page four. Compare this to the first page. Someone might go to Digg, glance at the topics, then leave. But if someone is dedicated enough to go all the way to page three, there is a likelier chance that he will stay on Digg.
  • Post Timing: Depending on the number of sites that are dugg, a particular site might be sent back pages faster or slower.
  • User Timing: Perhaps more users are on at certain times of the day.

From four through six, the view count drops by about one-half each time. At seven, however, the view count INCREASES by 28.8%! Most likely it is a combination of the timing issues, plus the superstition of seven—a user might jump directly to page seven. Who knows?

From seven to eight we have about the same drop as from five to six. But at nine, the count increases again. And again, this is most likely explained by a timing issue. Page ten has a drop from page nine, but it is not a big drop.

Just looking at the first graph for a minute, we see the dropping rate almost flatten from three to four. After two sharp falls, it falls much more slowly.

The second graph has two points above 100%, which we have looked at before. It also has an generally positive trend: the later the page, the higher the retention rate. This makes logical sense, as explained in the bulleted dedication part.

There you have it. Overall, it’s mostly about the first page—it gets more views than the others combined. But weird things happen once you get into the later ones.

The Impact of Digg, and Digg vs. StumbleUpon

Ever since posting US Census 2010 Win, I began to receive a large number of hits on this site. I was first hit by a giant wave (see my previous post The True Impact of StumbleUpon) from StumbleUpon, a site which I use very often. But today, another wave hit, and this time, it was Digg.

Digg
First spike is from StumbleUpon, second from Digg.

I did notice quite a few interesting trends and correlations. And also, I know that, in dealing with real-time stats, the facts might change the next minute. I don’t use Digg (I registered an account but haven’t touched it recently), so you may expect me to be biased toward StumbleUpon. But it turns out that the facts show Digg users to interact with the site more.

StumbleUpon

Time Interval: 3/20/10 to 3/24/10 (full days)

Total Hits on “Census” image: 41,125

Max Hits on “Census” in One Day: 24,317 (3/21/10)

Total Hits on Homepage: 819

Total (Average) Homepage Hits Divided by Hits on “Census”: 1.99%

Explanation: A hit is a visit. “Hits on Homepage” refers to the number of times the homepage of my blog was accessed. It may include readers who did not arrive on it from StumbleUpon; however, that number should be fairly low, considering that the TOTAL visits per day prior to this was about 20-30. Also, if a user goes through more than one page (i.e. to access older posts), each page of posts will count as a hit to the homepage.

Digg

Time Interval: 3/26/10, 12:00 am to 5:22 pm, Central Time (UTC -6)

Hits on “Census” image: 33,922

Hits on “Census” image, minus StumbleUpon Hits: 33,335

Hits on Homepage: 959

Hits on Homepage Divided by Hits on “Census”: 2.88% 2.83%*

*Edit: The previous number used the wrong methodology: I was comparing total hits on homepage, including StumbleUpon users, to just Digg hits on “Census.” The new number, 2.83%, is the total, or weighted average for Digg and StumbleUpon that day.

Alright, I apologize for not having done extremely sophisticated statistics here, but I do notice that even on this part of one day, the chance that a Digg user clicks my homepage (2.88%) is significantly higher than that for a StumbleUpon user (1.99%). Thus Digg has a lower bounce rate than that of StumbleUpon, at least for this image.

This was surprising because I thought StumbleUpon had a significantly lower bounce rate and a significantly higher average time on website and number of pages accessed, from my earlier reading. Sure enough, I found several articles (such as this, this, and this), all pointing to the same understanding that StumbleUpon is better in these aspects for a website.

So why did the census post develop a different result? A few possibilities:

  • Samples not comparable—for StumbleUpon I measured over a period of five days, whereas for Digg, I measured over a period of less than 18 hours. I’m not exactly sure what difference this makes; maybe because Digg stats have not continued into the evening/night yet, the Digg users are spending more time. Perhaps looking at a site at night makes one more likely to go to the next website. I don’t know.
  • Assumptions outdated—I would normally not use this as an excuse, but it happens that the three articles I mentioned before were all written in 2007 or 2008; maybe by 2010 StumbleUpon’s user base has changed to be more like that of Digg’s.
  • Nature of content not predicted—It’s an image. Perhaps StumbleUpon users spend more time reading articles and less on images. Who knows? Plus, it is in the “Humor” category in StumbleUpon, so perhaps it was not as entertaining as other pages in StumbleUpon’s hilarious “Humor” section. I noticed specifically in my blog stats that the most accessed category tags were “Lol,” which is the section under which the picture was posted. Also, I notice that for Digg, I have approximately the bounce rate that is given by the articles, but I have a significantly higher bounce rate than expected for StumbleUpon. This also points evidence to a likely reason being the nature of content.

Whatever the case, go StumbleUpon and Digg!

Edit: An update at 10:12 pm

Update (3/27/10): For 3/26/10, complete day.

The stats for Digg as shown above went up to 5:22 pm on 3/26/10. This time, it’s the full day.

Time Interval: 3/26/10 (full day)

Hits on “Census” image: 81,474

Hits on Homepage: 1,906

Hits on Homepage Divided by Hits on “Census”: 2.33%

Interestingly, the interaction rate decreased over time. At 5:22 pm, as seen above, it was at 2.83%. From 5:22 pm to 12:00 am, the number of hits was the day total minus the number up to 5:22, or 81,474 − 33,922 = 47522. Likewise, the number of homepage hits in this time interval was 1906 − 959 = 947. This gives a division ratio of 1.99%, which is coincidentally the exact same as for StumbleUpon the past few days. When this 2.83% and 1.99% are weighted, the total ratio is 2.33%. This seems to imply that at night, viewers are less likely to interact with the site than they are during the daytime.