Reflections on four years of research blogging
composition.al 2017-02-03
This blog just passed its four-year anniversary! Back in January 2015, I reflected on what my first two years of research blogging had been like. This post is an update on what’s changed since then and what hasn’t.
How popular is this blog, and who’s reading it?
I don’t use any website analytics tools, so I don’t have much information about how much traffic this blog gets or where that traffic comes from, other than what I observe on social media and in the comments here. One of the few signals that I have is Feedly follower count. In October 2014, when the blog was close to the two-year mark, someone pointed out to me that it had hit 100 followers on Feedly, and so I started paying occasional attention to that number. Today, at the four-year mark, I have 213 Feedly followers, so it seems like I’ve gained followers at a steady trickle of about fifty per year. For comparison with some other blogs I read, this makes my blog a bit less popular than, say, The Programming Languages Enthusiast (257 followers), about 25% as popular as Phil Guo’s website (871 followers), and an order of magnitude less popular than Julia Evans’s blog (5633 followers). I don’t know how representative Feedly follower counts are of general readership, but the relative popularities seem to be in the right ballpark to me, based on what I know about those blogs.
Incidentally, while I’m looking at the Feedly follower counts of blogs I read, I should urge all two hundred and thirteen of you to follow the apparently criminally underfollowed Counting from Zero, Janet Davis’s fascinating blog about starting a computer science program from scratch at Whitman College.
Why not use analytics tools? I’m not fundamentally opposed to the idea, but the ones I’ve tried so far have turned out to be depressingly focused on what I should be doing to make money off of my readers, and not very useful for actually helping me understand who’s reading my blog and why. I guess I’ll just have to resort to actually asking you those things. So: who are you? Why do you read this blog? I’d love to know!
What kinds of posts are popular?
At the beginning of 2015, the most popular post on this blog (again, based on the limited information I have from Feedly and social media) was “Your next conference should have real-time captioning” from May 2014. That post is now the second most popular: “Refactoring as a way to understand code” blew it out of the water in December 2015. I was kind of surprised by the popularity of the “Refactoring” post, because I thought I was saying things that were pretty obvious. In retrospect, though, I think that the obviousness was a big reason for its popularity, because a lot of people shared it with a comment along the lines of “I thought everyone did this” or “I’ve been doing this for years” or “Yes, definitely do this, of course”. To maximize popularity, perhaps my next post should be something like “Use version control”!
Out of the remaining forty-seven posts I wrote in the last two years, these are the others that seem to have been the most popular:
- “Say ‘experts’ instead of ‘smart people’”, April 2015
- “What’s the difference between LVars and CRDTs?” and “What’s the difference between inflationary and monotonic functions?”, summer 2015
- “An economics analogy for why adversarial examples work” and “Thoughts on ‘Safety Verification of Deep Neural Networks’”, November 2016
My theory is that popular posts tend to become popular for one of two reasons: either they’re about a topic that appeals broadly to a lot of readers (which is what I think happened with “Refactoring as a way to understand code” and “Say ‘experts’ instead of ‘smart people’”), or they’re about a relatively niche topic, but someone with a bigger platform than me finds them and shares them with people who are very interested in that niche (which is what I think happened with the others listed above).
Sometimes, posts that I think are good don’t get any traction. I’m not great at promoting my blog; I tweet about posts when they go up, but often that ends up being, like, in the middle of the night, or right when Beyoncé releases a new album, or something. And I don’t like the “In case you missed it, here’s another link to my post” genre of tweet, so I don’t usually do that. Thankfully, other people seem to be better at promoting my blog than I am!
Is two posts a month still sustainable?
Blogs so often seem to start out with an ambitious flurry of posts and then run out of steam. I didn’t want that to happen with this blog, so I decided that I would have a quota for number of posts I’d write: two a month, no more and no less. For the first couple of years, I was able to sustain that pace pretty comfortably, although I’d often procrastinate until the last couple of days of the month and then have to scramble to finish the posts, which is why so many of my posts are dated the 30th and 31st of the month.
That’s how it went for the first three years. Around the beginning of 2016, though, I began to get even more lazy. I’d allow a month to go by entirely, then hurriedly write posts and have to backdate them to the 30th and 31st of the previous month. Then the current month would end without me having written two posts dated for it, and the cycle would repeat itself. Some readers started to call me on this, which I actually really appreciated; it showed me that someone cared about my blog. I finally got caught up in October 2016, only to fall behind again in November. In fact, this post itself is an example of this pattern: as I type, it is February 3, but this post is dated January 30. How embarrassing!
In my defense, sometimes posts go up late not because I’m lazy, but because they’re legitimately hard to write and take me a long time. An example is “The saga of Accumulo bug 4379”, which describes events that occurred in July 2016, and which is dated August 2016, but didn’t actually go up until September 5. This post evolved a lot as I was writing it. It was originally going to be something like, “I had a good interaction with the Accumulo project!” Then, as I contemplated why the experience had been good, it changed into, “I had a mostly-good interaction with the Accumulo project, but here’s how it could have been even better!” Then I thought twice about that as well, and finally it changed into, “My experience with Accumulo bug 4379 led me to reflect on and reconsider my long-held assumptions about how open-source projects should handle bug reports and issue tracking!” I’m happy with what the post turned into, but because of all that, it took weeks to finish and was rather exhausting to write.
Many bloggers have a much more stream-of-consciousness style. For instance, Julia Evans will sometimes ask a question on Twitter and then later that day distill what she learned into a post. That would be really hard for me to do. For me, if I learn something new, sometimes it takes days or weeks before I can figure out what exactly it is that I learned, and how to do a decent job of explaining it.
Anyway, given that I’m still to able to cough up an average of two posts a month (albeit sometimes two posts that have to be backdated to the previous month), I’m going to try to carry on with this schedule for a while longer. Some of them will inevitably end up being “filler” posts, like a lot of the CFP posts are, for example. I’m okay with that.
What is a “research blog”, anyway?
Two years ago, I wrote that maybe only half my posts were about research, meaning my own original research. I listed several things that I’d been writing about instead: other people’s research, non-research things that I wanted the research community to know about, and so on. In response to that, Gabriel Scherer left one of his typically thoughtful comments:
I think the only reasonable definition of “research” is “what researchers do” (researchers being defined, of course, as “those that research”). All the things listed in your “Instead, ” sentence are research in my book, and even those posts “that weren’t research” could solidly be argued to be part of some people’s research (including at least you, apparently). For example, your interesting dicussions about Hacker School count as “diffusion”, helping other people to appreciate your discipline, and we have a checkbox about that in my lab’s yearly Activity Report.
After two more years of blogging, I’m starting to think that Gabriel is right about “what researchers do” being the only reasonable definition of “research”. I haven’t written a post tagged with “research” in a long time, but almost every post I write is about something that I do in my capacity as a “researcher”. “What even is OpenStack?” is not a research question, but it’s nevertheless a question that I wanted to answer in the course of a research project. Fiddling with git-filter-branch is probably not something that I’d ever mention in a research paper, but it is something that I needed to do in the process of creating an artifact to accompany that paper. My work organizing !!Con is not computer science, but I would have a lot less of a reason to be a computer scientist if events like !!Con did not exist. So I’m thinking of getting rid of the “research” tag entirely, in acknowledgement of Gabriel’s point that the whole blog is already about “what researchers do”. If what someone actually wants is my papers or talks, then there are tags for those things.