I worry that a/b testing will pick up some things but miss others. It might detect the percentage of people who click, for example, but it will probably miss how many of them are silently disappointed and say to themselves ,"this dynomight website sucks, I'm never going back". (At least, that kind of thing seems very hard to capture.)
nothing stopping you from a/b testing against longer term outcomes (return rates) in addition to initial click rates. also other instruments like dwell time, etc.
Sure there is, longer-term outcomes take longer! If I wanted to measure the impact of one style of title on my "reputation", feel like I'd need to use one style of title for a long period (a year?) and then alternate back and forth a bunch of times to average out any exogenous changes. Doesn't seem worth it.
(I guess in principle it would be fine if I had a panopticon and I knew, for example, if any reader came because some friend who'd seen title style A told them to. But that sounds terrible.)
well imagine you had two style of titles for every post. for every new user (cookie id or whatever they use these days) you assign them to one type of title forever more and serve that person that title style.
after a few months you go back and see how many still visit your blog. t-test and you're done!
That would be (much) better, but I think it would miss things like: Are people who see title type A more likely to tell their friends that I'm dumb and bad?
yes network effects/SUTVA are always an unknown confounder of any social experiment. but maybe you can make the assumption that they wouldnt simultaneously continue to visit the blog and bad mouth you? or maybe they visit your site to find more terrible titles to talk shit with 😆 all to say "How to title your blog post or whatever + Evidence!" is undoubtedly superior to "How to title your blog post or whatever"
Dang, this is absolutely going to shape how I title my posts in the future (whenever I can resist being cute and punny, so not toooo often).
It's funny, I think this advice was kind of implicit in an old post of mine about why Beeminder is targeted at nerds -- https://blog.beeminder.com/nerds -- citing Paul Graham and others on appealing strongly to a narrow niche. I just hadn't thought to extend it this way.
Another hobby horse of mine is how to name projects/startups/orgs/etc -- https://messymatters.com/nominology -- and now I'm pondering how much overlap there is between that and how to title pieces of writing.
PS: I love your "Taste Games" post and the title seems pretty great to me so now I'm curious what you might name it if you wrote it today. Also, can't you have the best of both worlds with a subtitle?
> My favorite thing category is “Lucid examination of all sides of an issue which finds some evidence pointing in various directions”.
Scott Alexander calls these "X: Much More Than You Wanted to Know". Which I think is quite good. But it's so I iconic that I feel like I can't use it without looking like I'm ripping off Scott Alexander.
This is a question I've struggled with—if I investigate a difficult question and come to a conclusion, should the title be the question, or the conclusion, or something else? "Are seed oils bad for you?" vs. "Seed oils are not bad for you" vs. "Thoughts on seed oil"
Good point, "much more than you wanted to know" is indeed a solution to that problem. (But I also agree it seems heretical to borrow it.)
I genuinely don't know if you should state the conclusion. Personally—I am a big fan BTW—I enjoy your stuff for the "tour" and the fact that it's sort of designed to allow me to make up my own mind. So I feel like a "question" type title most accurately reflects the value proposition. On the other hand, I'd read it anyway, and I wish this whole "genre" was more popular...
> I enjoy your stuff for the "tour" and the fact that it's sort of designed to allow me to make up my own mind.
I'm glad you interpret it that way (and thank you for the compliment, that means a lot coming from you!). I'm not *trying* to let readers make up their own minds, in fact it feels like the way I write is pretty opinionated and I could definitely be less opinionated. I'm guessing it reads that way because I try to be clear about why I believe what I believe, which in effect lets readers evaluate the evidence for themselves, even though that's not exactly what I'm going for. I just think it's good practice to be explicit about what evidence you're considering.
P.S. Yesterday I had to write the word "dynamite" and I couldn't remember whether it was spelled "dynamite" or "dynamight". I blame you for this.
Hmm! I don't think I'd suggest being less opinionated. Broadly speaking, I think you should write with whatever level of certainty best reflects whatever you actually believe. After all, that's what everyone wants to know! And as as long as you show your work and explain why you believe that, I don't think it really reduces value very much for anyone who might disagree.
Surprisingly interesting! I would be happy to read your thoughts if they came in brown paper packages sequentially numbered, but it's unlikely that I would have read the first one if you didn't have the quadruple (quintuple?) meaning in the title of the blog.
Oh, yeah, a "dyno" is a somewhat dorky climbing word used for any "non-static" move where you go from a stable start position to a stable end position via a non-stable intermediate state.
Nice post! My all-time favourite blog post title is 'Run by default.' I was quite disappointed to discover that I misremembered and the title is actually the (slightly worse, in my view) 'Running by default.' https://www.jefftk.com/p/running-by-default
Also an excellent example of the kind of situation where stupid puns ("keep it running"? "the world at my feet"? "running without borders"?) would be tempting (but worse).
This is really useful! I've been struggling with naming my weekly newsletter posts for Manifold Markets, and have defaulted to trying to think of a pun with the name of the newsletter (Above the Fold), which is a well that will run dry fairly quickly. "Above the Fold, Cards of Gold" was pretty decent, when writing about Trump's proposed gold card immigration system, and "Above the Folded Secret Ballot" was fine for the conclave, but last week's "Above the Fold, From Alcatraz to Zuppi" was pretty terrible, as was "Above the 49th Parallel" (Fold --> Forty...?).
I thiiiiiiiinnnkkk that's a situation where titles probably don't have as big a "classifier" function? I'd imagine that a strong majority of your audience is already familiar with prediction markets, so you just need to communicate, "This is the one about the Conclave" and people have most of the information they need.
So puns aren't necessarily bad. But I'd bet you could also go with maximally unclever titles like "The conclave" or "The gold card pathway to citizenship" or "The signal affair" and you'd still probably do well?
> So you don’t want people in the hate + click region.
I think this could use a qualifier. For instance: You don't want hate clicks if you want high quality good faith interactions. Plenty of people are fine with hate clicks, especially advertisers.
Oh, there absolutely are. There's no such thing as bad press. Clicks mean money. I call that subset of clickbait "outrage bait."
I see headlines all the time that are intentionally misleading or overstating in order to provoke people into refuting the author and each other in the comments. Outrage drives engagement.
If you haven't been aware of that in the past, keep an eye out for it now.
> Traditional advice says that you should put your main “message” in the title. […] It will definitely lead to lots of comments “refuting” you from people who didn’t read your thing.
Great piece, thank you. And it led me to the deep dive on seed oils….
Some of my favorite titles
- The man who beat roulette
- Why Isn't the Whole World Rich?
- Why parents bully their children
these theories of yours are begging to be a/b tested, which is entirely possible on the interwebs...
I worry that a/b testing will pick up some things but miss others. It might detect the percentage of people who click, for example, but it will probably miss how many of them are silently disappointed and say to themselves ,"this dynomight website sucks, I'm never going back". (At least, that kind of thing seems very hard to capture.)
nothing stopping you from a/b testing against longer term outcomes (return rates) in addition to initial click rates. also other instruments like dwell time, etc.
Sure there is, longer-term outcomes take longer! If I wanted to measure the impact of one style of title on my "reputation", feel like I'd need to use one style of title for a long period (a year?) and then alternate back and forth a bunch of times to average out any exogenous changes. Doesn't seem worth it.
(I guess in principle it would be fine if I had a panopticon and I knew, for example, if any reader came because some friend who'd seen title style A told them to. But that sounds terrible.)
well imagine you had two style of titles for every post. for every new user (cookie id or whatever they use these days) you assign them to one type of title forever more and serve that person that title style.
after a few months you go back and see how many still visit your blog. t-test and you're done!
That would be (much) better, but I think it would miss things like: Are people who see title type A more likely to tell their friends that I'm dumb and bad?
yes network effects/SUTVA are always an unknown confounder of any social experiment. but maybe you can make the assumption that they wouldnt simultaneously continue to visit the blog and bad mouth you? or maybe they visit your site to find more terrible titles to talk shit with 😆 all to say "How to title your blog post or whatever + Evidence!" is undoubtedly superior to "How to title your blog post or whatever"
> If you make a thing that you would love, then I guarantee you at least 0.0001% of other people would love it too. That’s still 8000 people!
Lol. I have interests so niche you would never have dreamed of them.
Dang, this is absolutely going to shape how I title my posts in the future (whenever I can resist being cute and punny, so not toooo often).
It's funny, I think this advice was kind of implicit in an old post of mine about why Beeminder is targeted at nerds -- https://blog.beeminder.com/nerds -- citing Paul Graham and others on appealing strongly to a narrow niche. I just hadn't thought to extend it this way.
Another hobby horse of mine is how to name projects/startups/orgs/etc -- https://messymatters.com/nominology -- and now I'm pondering how much overlap there is between that and how to title pieces of writing.
PS: I love your "Taste Games" post and the title seems pretty great to me so now I'm curious what you might name it if you wrote it today. Also, can't you have the best of both worlds with a subtitle?
> My favorite thing category is “Lucid examination of all sides of an issue which finds some evidence pointing in various directions”.
Scott Alexander calls these "X: Much More Than You Wanted to Know". Which I think is quite good. But it's so I iconic that I feel like I can't use it without looking like I'm ripping off Scott Alexander.
This is a question I've struggled with—if I investigate a difficult question and come to a conclusion, should the title be the question, or the conclusion, or something else? "Are seed oils bad for you?" vs. "Seed oils are not bad for you" vs. "Thoughts on seed oil"
Good point, "much more than you wanted to know" is indeed a solution to that problem. (But I also agree it seems heretical to borrow it.)
I genuinely don't know if you should state the conclusion. Personally—I am a big fan BTW—I enjoy your stuff for the "tour" and the fact that it's sort of designed to allow me to make up my own mind. So I feel like a "question" type title most accurately reflects the value proposition. On the other hand, I'd read it anyway, and I wish this whole "genre" was more popular...
> I enjoy your stuff for the "tour" and the fact that it's sort of designed to allow me to make up my own mind.
I'm glad you interpret it that way (and thank you for the compliment, that means a lot coming from you!). I'm not *trying* to let readers make up their own minds, in fact it feels like the way I write is pretty opinionated and I could definitely be less opinionated. I'm guessing it reads that way because I try to be clear about why I believe what I believe, which in effect lets readers evaluate the evidence for themselves, even though that's not exactly what I'm going for. I just think it's good practice to be explicit about what evidence you're considering.
P.S. Yesterday I had to write the word "dynamite" and I couldn't remember whether it was spelled "dynamite" or "dynamight". I blame you for this.
Hmm! I don't think I'd suggest being less opinionated. Broadly speaking, I think you should write with whatever level of certainty best reflects whatever you actually believe. After all, that's what everyone wants to know! And as as long as you show your work and explain why you believe that, I don't think it really reduces value very much for anyone who might disagree.
> dynamight
Missed opportunity? {{🧗,🧪🚫}{🤔,💪},🧨}
Much appreciated advice! I always have a devil of a time naming things, anything really. This helps, even if I am somewhat beyond helping :)
Surprisingly interesting! I would be happy to read your thoughts if they came in brown paper packages sequentially numbered, but it's unlikely that I would have read the first one if you didn't have the quadruple (quintuple?) meaning in the title of the blog.
> quintuple
{{🦕,🧗}{🤔,💪},🧨}
Ooh, I hadn't noticed "might" as in strength before. I guess that rock climber emoji is for dynamism?
Oh, yeah, a "dyno" is a somewhat dorky climbing word used for any "non-static" move where you go from a stable start position to a stable end position via a non-stable intermediate state.
Nice post! My all-time favourite blog post title is 'Run by default.' I was quite disappointed to discover that I misremembered and the title is actually the (slightly worse, in my view) 'Running by default.' https://www.jefftk.com/p/running-by-default
Also an excellent example of the kind of situation where stupid puns ("keep it running"? "the world at my feet"? "running without borders"?) would be tempting (but worse).
This is really useful! I've been struggling with naming my weekly newsletter posts for Manifold Markets, and have defaulted to trying to think of a pun with the name of the newsletter (Above the Fold), which is a well that will run dry fairly quickly. "Above the Fold, Cards of Gold" was pretty decent, when writing about Trump's proposed gold card immigration system, and "Above the Folded Secret Ballot" was fine for the conclave, but last week's "Above the Fold, From Alcatraz to Zuppi" was pretty terrible, as was "Above the 49th Parallel" (Fold --> Forty...?).
I thiiiiiiiinnnkkk that's a situation where titles probably don't have as big a "classifier" function? I'd imagine that a strong majority of your audience is already familiar with prediction markets, so you just need to communicate, "This is the one about the Conclave" and people have most of the information they need.
So puns aren't necessarily bad. But I'd bet you could also go with maximally unclever titles like "The conclave" or "The gold card pathway to citizenship" or "The signal affair" and you'd still probably do well?
> So you don’t want people in the hate + click region.
I think this could use a qualifier. For instance: You don't want hate clicks if you want high quality good faith interactions. Plenty of people are fine with hate clicks, especially advertisers.
I love the idea that there could be specific products that are marketed to the hate click demographic.
Oh, there absolutely are. There's no such thing as bad press. Clicks mean money. I call that subset of clickbait "outrage bait."
I see headlines all the time that are intentionally misleading or overstating in order to provoke people into refuting the author and each other in the comments. Outrage drives engagement.
If you haven't been aware of that in the past, keep an eye out for it now.
If I owned a company that made panties with a distressing habit of bunching up, I would think that was my target demographic right there.
> Traditional advice says that you should put your main “message” in the title. […] It will definitely lead to lots of comments “refuting” you from people who didn’t read your thing.
OHHHHHHHHhhhhhhhh
thank you very much