3 ms·

You Release Late and Infrequently

65 points by liebke 17y ago

wingo 17y ago
I was about to raise the point that "release early, release often" is a phrase from free software, and wonder at its appropriation by commerce, but it turns out it's from ESR. Perhaps it is appropriate, then. (The majority of my hack is not for customers.)
gurraman 17y ago
On frequent releases: I absolutely love that there's a new interface tweak almost every time I login to my GitHub account.
- bradfordcross 17y ago
  yea, although a couple of the tweaks have ended up tweaking me out when i can't find something. :-)
10ren 17y ago
Wikipedia has links that go nowhere. But it does still sound a bit dodgey, I think because to get accurate feedback, you necessarily need to fool people. For example, if you had a link that stated it was a proposed feature, it would affect the clickrate. Or maybe it wouldn't? Maybe interest would be enough. Perhaps this could be experimentally tested.
- dstorrs 17y ago
  Personally, I think I would feel ok about putting the link up and having it go simply to a landing page that said "This feature is still in production. Please comment below on how we could best suit it to your needs."
decode 17y ago
Just this weekend I watched Machinima.com's documentary series on the history of Valve Software: http://www.machinima.com/film/view&id=42271 http://www.machinima.com/film/view&id=42271 One of the common themes in the series is Valve's frequent choice to delay shipping in order to get things right. Gabe Newell says something like, "Your software is late for a while, but it's bad forever." As far as I can tell, this has worked extremely well for them. Which makes me wonder what the difference is between Valve's games and Gmail. I think it has something to do with their games being about experiential storytelling, and Gmail is a tool to accomplish tasks. I imagine most software falls somewhere on a spectrum between the "tool" and "experience" endpoints and I wonder if other software besides games might be hurt more than helped by "release early and often."
- smokinn 17y ago
  The difference is that valve's software has to have a build installed on the client's computer or worse, a disk you shove into a console. On the web, your releases update 100% of your client base instantly. Though I do agree that depending on your market, "release early and often" can hurt you. Getting a bad reputation early can be hard to shrug off later.
  - bradfordcross 17y ago
    Yes, one part is the ease of productionizing for the web. That said, I think there are always techniques for using the early, often, listen approach with private betas, multivariate testing, etc. Release doesn't necessarily mean releasing to the public, and does not even necessarily mean releasing something user facing.
  - DougBTX 17y ago
    It's not just the medium. A game with plot and direction won't be the same when replayed. Imagine someone trying to repeat a joke to make it sound funnier, when you've already heard the punch line. So, danger of over-generalisation, but certainly games like TF2 can be improved with updates, extensions rather than revisions. Perhaps that's his point about bugs vs new features. Releasing often works when you are adding more good bits, not so much when fixing broken bits that should never have been released in the first place.
- winthrowe 17y ago
  Another story on valve that I read indicated that they do lots of testing and iterative design in house in the early stages, so they seem to have an internal 'release early, release often' culture, at least as compared to other large game company cultures.
  - far33d 17y ago
    This is also the philosophy @ Pixar - you can't release a movie twice, so it had better be right the first time. You achieve this by having 1000's of internal reviews, checkpoints, and pivots along the way where you trust your employees to challenge problems they see.
omouse 17y ago
The Wufoo example is awful. Shouldn't have taken more than 5 min for them to realize, "HEY the Flash experience on GNU/Linux sucks! Maybe we shouldn't use it so that we can have more customers use the thing!" Those 3 versions of GMail could have been one version that was most useful. There was no time constraint for it since it was a side-project. The links to nowhere sounds like a good idea except I don't think it indicates interest in the feature being built. A link's text is usually very short and you don't really get any demographics from a click. If you knew 15 to 20 yr olds were clicking a link but the rest of the site was mainly used by 30 to 40 yr olds, you might want to spin out the link as a separate site instead of just a new feature or something. But how do you know how interested people really are? :/
- bradfordcross 17y ago
  Interesting criticisms of PB's approach to building gmail. Can you share a link to a product you've shipped in the past 12 months and some of your experiences about building it?
  - omouse 17y ago
    Is that an appeal to authority I hear? I haven't worked on a product that has shipped in the last 12 months but I have worked with several programming languages and have read some Dijkstra and have done some literate programming, and have gone to university/college for a few years now. My criticism was based on the stuff PB said. All of that functionality could have been done in a single version but PB didn't want to waste company resources unless he could offer up proof that they were being used wisely. There's nothing wrong with that, but to say that it's a good development model is awful. How long did GMail stay in beta for? It never really shipped.
    - lrm242 17y ago
      GMail has 100+ million users[1]. Imagine how many they could have if they actually did ship something. Crappy development model or not, does it really matter? The thing is widely successful and any insight into how it came about is valuable as a learning experience. [1]: http://www.ft.com/cms/s/0/fc3d8a86-02a0-11de-b58b-000077b07658.html?nclick_check=1 http://www.ft.com/cms/s/0/fc3d8a86-02a0-11de-b58b-000077b076...
ntoshev 17y ago
I also feel it's a breach of trust when you mislead the users you have implemented something in order to test demand. Never mind the ethics, it seems it doesn't always work: http://news.ycombinator.com/item?id=960721 http://news.ycombinator.com/item?id=960721

65 points by liebke 17y ago

wingo 17y ago

I was about to raise the point that "release early, release often" is a phrase from free software, and wonder at its appropriation by commerce, but it turns out it's from ESR. Perhaps it is appropriate, then. (The majority of my hack is not for customers.)

gurraman 17y ago

On frequent releases: I absolutely love that there's a new interface tweak almost every time I login to my GitHub account.

bradfordcross 17y ago
yea, although a couple of the tweaks have ended up tweaking me out when i can't find something. :-)

10ren 17y ago

Wikipedia has links that go nowhere. But it does still sound a bit dodgey, I think because to get accurate feedback, you necessarily need to fool people. For example, if you had a link that stated it was a proposed feature, it would affect the clickrate. Or maybe it wouldn't? Maybe interest would be enough. Perhaps this could be experimentally tested.

dstorrs 17y ago
Personally, I think I would feel ok about putting the link up and having it go simply to a landing page that said "This feature is still in production. Please comment below on how we could best suit it to your needs."

decode 17y ago

Just this weekend I watched Machinima.com's documentary series on the history of Valve Software: http://www.machinima.com/film/view&id=42271 http://www.machinima.com/film/view&id=42271 One of the common themes in the series is Valve's frequent choice to delay shipping in order to get things right. Gabe Newell says something like, "Your software is late for a while, but it's bad forever." As far as I can tell, this has worked extremely well for them. Which makes me wonder what the difference is between Valve's games and Gmail. I think it has something to do with their games being about experiential storytelling, and Gmail is a tool to accomplish tasks. I imagine most software falls somewhere on a spectrum between the "tool" and "experience" endpoints and I wonder if other software besides games might be hurt more than helped by "release early and often."

smokinn 17y ago
The difference is that valve's software has to have a build installed on the client's computer or worse, a disk you shove into a console. On the web, your releases update 100% of your client base instantly. Though I do agree that depending on your market, "release early and often" can hurt you. Getting a bad reputation early can be hard to shrug off later.
- bradfordcross 17y ago
  Yes, one part is the ease of productionizing for the web. That said, I think there are always techniques for using the early, often, listen approach with private betas, multivariate testing, etc. Release doesn't necessarily mean releasing to the public, and does not even necessarily mean releasing something user facing.
- DougBTX 17y ago
  It's not just the medium. A game with plot and direction won't be the same when replayed. Imagine someone trying to repeat a joke to make it sound funnier, when you've already heard the punch line. So, danger of over-generalisation, but certainly games like TF2 can be improved with updates, extensions rather than revisions. Perhaps that's his point about bugs vs new features. Releasing often works when you are adding more good bits, not so much when fixing broken bits that should never have been released in the first place.
winthrowe 17y ago
Another story on valve that I read indicated that they do lots of testing and iterative design in house in the early stages, so they seem to have an internal 'release early, release often' culture, at least as compared to other large game company cultures.
- far33d 17y ago
  This is also the philosophy @ Pixar - you can't release a movie twice, so it had better be right the first time. You achieve this by having 1000's of internal reviews, checkpoints, and pivots along the way where you trust your employees to challenge problems they see.

omouse 17y ago

The Wufoo example is awful. Shouldn't have taken more than 5 min for them to realize, "HEY the Flash experience on GNU/Linux sucks! Maybe we shouldn't use it so that we can have more customers use the thing!" Those 3 versions of GMail could have been one version that was most useful. There was no time constraint for it since it was a side-project. The links to nowhere sounds like a good idea except I don't think it indicates interest in the feature being built. A link's text is usually very short and you don't really get any demographics from a click. If you knew 15 to 20 yr olds were clicking a link but the rest of the site was mainly used by 30 to 40 yr olds, you might want to spin out the link as a separate site instead of just a new feature or something. But how do you know how interested people really are? :/

bradfordcross 17y ago
Interesting criticisms of PB's approach to building gmail. Can you share a link to a product you've shipped in the past 12 months and some of your experiences about building it?
- omouse 17y ago
  Is that an appeal to authority I hear? I haven't worked on a product that has shipped in the last 12 months but I have worked with several programming languages and have read some Dijkstra and have done some literate programming, and have gone to university/college for a few years now. My criticism was based on the stuff PB said. All of that functionality could have been done in a single version but PB didn't want to waste company resources unless he could offer up proof that they were being used wisely. There's nothing wrong with that, but to say that it's a good development model is awful. How long did GMail stay in beta for? It never really shipped.
  - lrm242 17y ago
    GMail has 100+ million users[1]. Imagine how many they could have if they actually did ship something. Crappy development model or not, does it really matter? The thing is widely successful and any insight into how it came about is valuable as a learning experience. [1]: http://www.ft.com/cms/s/0/fc3d8a86-02a0-11de-b58b-000077b07658.html?nclick_check=1 http://www.ft.com/cms/s/0/fc3d8a86-02a0-11de-b58b-000077b076...

ntoshev 17y ago

I also feel it's a breach of trust when you mislead the users you have implemented something in order to test demand. Never mind the ethics, it seems it doesn't always work: http://news.ycombinator.com/item?id=960721 http://news.ycombinator.com/item?id=960721