This is a bit of a double-entendre title.
First off, my blogging routine has somewhat ground to a halt and I’m not updating this site as much as I used to (or would like to). Work commitments keep me pretty busy and whatever online time I do have, I tend to contribute over at AcademicProductivity.com instead. So part of this post is to say sorry to those readers who might have been looking for updates on the UK microgeneration scene or other news.
But this post is also about trying to understand why my blogging efforts didn’t really take off. There’s any number of sites that will tell you what you need to do to be successful – post frequently, build a community, etc. – and yes, I’ve neglected much of that advice: mea culpa. That’s just big picture stuff though and I’d like to do a proper statistical post-mortem to understand why my blogging habits broke down. In other words, WWNSD*?
The record
Let’s start with some basic descriptive statistics of my posting record.
- First post = 2002-09-27
- Total posts = 143
- Mean lag between posts = 17.4 days
Next I plotted the distribution of posts and modelled it as a normal distribution with μ = 2006-11-05 and σ = 441.1 (the data passes a Shapiro test, p = 0.000434).
You can see that it’s not really normal and there’s a point where I tried to kick start my blogging again in early 2008. But using this distribution let’s us answer some simple questions.
For example, at what point should I have admitted defeat and stopped blogging? Using the language of statistical process control, there should be a sigma threshold at which point most of your good posts are behind you and you’re just in decline. My last post was published 61 days ago: at that point, assuming a normal distribution, 98.6% of the posts I was theoretically ever going to make were behind me. This corresponds to +2.2σ.
So that’s one useful finding already: by the time, you’re at the +2σ side of a normal distribution, you should start thinking about why you blog. Actually scratch that – start thinking about it at +1σ ; if you ever hit +3, find a new hobby.
Draft posts
In addition to the 143 published posts, I have 58 draft posts. To my mind, this is where the real autopsy begins.
One question I had was whether my blogging slowed down because I stopped having ideas, or because I just wasn’t follow through on them. The figure below compares the temporal distribution of both published posts and drafts and there is a significant difference between the two samples (using Kolmogorov-Smirnov test, p = 1.665e-05).
At first, this suggests that both modes of failure are contributing. The narrower distribution of the draft posts shows that the flow of new ideas did dry up at some point even though I kept posting after that. But anecdotally, I can remember that these most recent published posts were written on the spur of the moment: they never even entered the drafts queue. So while I did stop thinking of new draft ideas, I think new post ideas in general kept coming.
The bigger issue was the lack of motivation and the resulting failure to convert drafts to posts. Posts were taking simply too long to complete and the sight of an ever-increasing backlog made it difficult to know where to start.
So there’s another warning sign: if your drafts pile builds up to more than about 5 posts, consider pruning substantially. You can either delete the drafts entirely or use a link dump post to clear out the backlog. Just remember that if you’ve got more than a few drafts sitting around, it’s very unlikely that you will go back to these posts and if you do, you probably won’t remember why they were interesting in the first place.
Tits versus owls
I think the drafts dilemma is a really good illustration of why microblogging services like Twitter have taken off. For me, a quality post will take a couple of hours work. I like to check my facts before publishing, find a few interesting references, and it’s hard to really think through an issue and present something new and useful in less than 100 words. (My posts may have only averaged 252 words but there’s a big spread.)
In contrast, writing 140 characters takes very little time at all and even seasoned bloggers like Kottke seem to have adopted quite a brief style. Compare this with people like Andy Baio, Ben Fry or Errol Morris, all of whom blog infrequently (once a week to once a month) but with posts ranging from a couple hundred words to thousands. If tweeting is for the birds, then these guys are the owls: patient and knowledgeable.
Sources of inspiration
Next, I wanted to figure out how many of my posts are inspired by links. My gut feeling is that I often blog in response to things I’ve seen on the web and when I was focusing on microgeneration issues, these germs of an idea often came from traditional news media. So I looked through the draft posts and identified those that had (a) at least one link in them and (b) a news site link (specifically from the Guardian or the BBC). The result? 72.4% of my draft posts contained at least one link and 44.8% had either a BBC or Guardian link in them.
Is there any difference between the share of links in published and non-published posts? In other words, were the drafts “lazier” than the regular posts? Unfortunately I can’t easily tell since 89.5% of published posts have a link in them; they may not have started from a link but they did grow to include one. The presence of a news link is a slightly better measure in this regard, but the results still don’t show a significant difference (Chi-squared test, p = 0.389).
| Published | Drafts | |
| News link | 53 | 26 |
| No news link | 90 | 32 |
Timing of posts
How about the timing of the posts and drafts? Were there certain days that I was more likely to post than others? Well yes and no. First, a chi-squared test shows that there is no significant difference between the weekdays of draft and published posts (p = 0.972). However if you look at the overall distribution of the activity, there is a significant difference between the observed daily frequency and the null hypothesis that the posts would be equally distributed across all 7 days (p = 7.19e-11). As the figure below shows, Tuesday is more popular day for posting and Saturday decidedly less so. Apparently this early week preference is a common phenomena with Twitter as well.
As a final thing, I did a time series decomposition. This sort of analysis can help you identify seasonal, trend and irregular components in your data. After first grouping the data into weekly bins, I generated the following plot. For my data, there isn’t really anything going on beside the main trend. However it would be interesting to apply this analysis to a larger data set like Twitter’s feed or the posts of a more prolific blogger.
Conclusion
I think my experience has probably been fairly typically of the lower middle orders of bloggers. I started with a lot of enthusiasm and a clear idea about what I wanted to cover. But as time passed, other pressures crept in and I wasn’t able to keep up the pace, even as the site started to attract more interest.
The key issue for me is that I like to write longer posts and these take time. If it suits your personality, then you may be able to write a brilliant blog with short sharp posts and many people do. But if you’re trying to cover issues in more depth, then a much more concerted effort is required.
This analysis has been very helpful for me to understand my blogging habits and foibles. But I think there are three take-away points for any blogger:
- Monitor post frequency
You need to monitor the frequency of your published posts so that you can observe any decline in output and then actively decide what to do about it, rather than letting inertia take hold. A simple sigma-style analysis can be helpful: if you’re on the +1σ side of your post distribution, then start re-evaluting what you want to achieve with blogging. - Keep it fresh
You need to be ruthless in keeping your draft queue short so aim for fewer than five, and certainly no more than ten, draft post ideas. The author’s enthusiasm shines through on good blogs. If you’re trudging through six-month-old ideas just for the personal satisfaction of clearing out a backlog, stop: your readers don’t care. - Tits and owls
There appears to be two emerging styles of blogging. Tits are those people that thrive with a short style, using Twitter or just shorter frequent posts. By contrast, owls post less frequently but with detailed longer posts. New bloggers should experiment and get to know which style works best for you. Be aware though that if you’re an owl, you may have a harder job building up a following for your website. You’ll need to find an audience that’s patient enough to wait for your new posts and read their longer contents. These categories are also mainly applicable to amateur bloggers. Pros may have the time necessary to do one detailed post per day, whereas the rest of us might only manage one good post per week or per month.
Having now realised that I’m an owl, I plan to revamp the site in the coming months as a more static home on the web with less frequent, but more detailed, posts. Hope to see you back here then.
* What would Nate Silver do? Given the amount of wild speculation and general crap on the web, it seems like there’s an emerging desire for quality writing, supported by good data and analysis. Nate Silver at 538.com is a perfect example of this.
