6 ms·

How to design perfect software products (2013)

93 points by bdon 8y ago

externalreality 8y ago
What is being described here is very similar if not identical to what can be found in Agile software and lean software development books that gained popularity in the mid 2000's.
- mr_vile 8y ago
  well, it's definitely not identical
  - TheOtherHobbes 8y ago
    No it isn't. Agile software development assumes management knows what it's doing - which is often not the case at all. Agile is primarily about managing down assuming a coherent goal exists - not about finding the best coherent goal when it doesn't, or even about being able to tell that it doesn't. Agile business management doesn't really exist as an established practice. Although you can certainly find books, consultants, and all the usual noise that use the term, the idea that you can recursively define your business strategy - not your implementation - by solving the problems that provide the best customer satisfaction at the lowest implementation cost isn't as common as it should be. The obvious complication is that the definition of "customer" varies. A bootstrapped startup needs to focus on real income-generating customers. A VC-funded startup is in a different market. The primary customers are actually the VCs and other investors - not product users, whether they're paying or not. Pieter Hintjens is always worth reading, even though his "solutions" are wildly oversimplified and contingent. Here he takes it for granted that solving customer pain points should always be the primary value-generating goal - which for better or worse happens not to be true, especially for startups.
- codingdave 8y ago
  Or, even simpler than the books, it is a long discussion that ends up rephrasing points from the Agile Manifesto.
revskill 8y ago
What people failed to define themselves is "Simplicity". That concept "Simplicity" needs context. Is it API surface simple ? Is it data model simple ? Is it Lines of Code short ? Is it extensible with small core and huge plugins system ? ... Is Wordpress simple ? Is MERN simple ? Is Rails simple ? There's no answer here. That's it. There's no "absolute simplicity", it depends on context.
- Razengan 8y ago
  For that matter, what about the definition of "perfect"?
- lucio 8y ago
  How many new concepts a new user needs to learn, and how many mental models a new user must develop in order to effectively use your software. Twitter is simple, Git is complex.
  - cheez 8y ago
    Git is pretty simple. https://github.com/git/git/blob/ecbdaf0899161c067986e9d9d564586d4b045d62/ll-merge.c https://github.com/git/git/blob/ecbdaf0899161c067986e9d9d564...
    - dreamcompiler 8y ago
      Git is one of those rare products whose internal implementation is simple but whose user interface is a complicated mess.
    - dsjoerg 8y ago
      http://stevelosh.com/blog/2013/04/git-koans/ http://stevelosh.com/blog/2013/04/git-koans/
  - rileymat2 8y ago
    I think this is close to a pretty good definition, but it should be restricted to people who already have learned the mental models of the domain. In that the mental models learned to use the software should match those of the domain.
stolee 8y ago
The comparison to hill-climbing algorithms is an interesting one, especially because the author ignores the fact that hill-climbing algorithms get stuck in local optima unless there is a way to “restart” the search from a new location. There is a time and place for big, risky changes to advance beyond what this “patch” approach can do.
- anbop 8y ago
  I think this is what happened to Amazon. A series of locally optimized decisions like commingled inventory and 3rd party sellers that has turned the site into a used car dealership in some categories.
simplyinfinity 8y ago
Can we get ( 2012 ) tag please
gfodor 8y ago
One problem with the meme that building products is about "solving problems" is that the biggest successes often come from recognizing very vague "problems" others don't, in the form of seeing fundamentally better ways of doing something -- without being able to grasp up-front the long term concrete benefits. You need to build and ship to learn them. So in those scenarios, people don't recognize what you are doing as a "real" problem, so it ends up being hindsight bias when it's recognized as such. An example in the article cited was Wikipedia: Wikipedia didn't "solve a problem" with existing encyclopedias, it recognized there was just a fundamentally better way to collect objective knowledge and disseminate it -- most people at the time were scratching their heads about what the point was, since from their perspective encyclopedias worked perfectly well within the scope of their design goals. I think the mental model that you should always be working on software where you have people directly telling you the problem they want to solve is a poor one -- it's great for a variety of domains, but often people want a faster horse or don't even see a problem with their existing horse. Even when you think you are recognizing a problem others don't see yet -- it can require a leap of faith to begin exploring to see if you can iterate and react to discovering something with big promise.
ncmncm 8y ago
The rarest skill, even in highly intelligent engineers, is to be able to imagine the world after a thing has been changed, and see what will matter to people then. Everyone else (including you, most likely) can only see the change itself. Is the change pretty? No awareness at all of whether the product after the change is nicer than the present product. It is a skill that can be developed, but it takes focused attention. The end result is to see how to alter the change to make the end result better to use, not just a thing with another PITA feature.

93 points by bdon 8y ago

What is being described here is very similar if not identical to what can be found in Agile software and lean software development books that gained popularity in the mid 2000's.

mr_vile 8y ago
well, it's definitely not identical
- TheOtherHobbes 8y ago
  No it isn't. Agile software development assumes management knows what it's doing - which is often not the case at all. Agile is primarily about managing down assuming a coherent goal exists - not about finding the best coherent goal when it doesn't, or even about being able to tell that it doesn't. Agile business management doesn't really exist as an established practice. Although you can certainly find books, consultants, and all the usual noise that use the term, the idea that you can recursively define your business strategy - not your implementation - by solving the problems that provide the best customer satisfaction at the lowest implementation cost isn't as common as it should be. The obvious complication is that the definition of "customer" varies. A bootstrapped startup needs to focus on real income-generating customers. A VC-funded startup is in a different market. The primary customers are actually the VCs and other investors - not product users, whether they're paying or not. Pieter Hintjens is always worth reading, even though his "solutions" are wildly oversimplified and contingent. Here he takes it for granted that solving customer pain points should always be the primary value-generating goal - which for better or worse happens not to be true, especially for startups.
codingdave 8y ago
Or, even simpler than the books, it is a long discussion that ends up rephrasing points from the Agile Manifesto.

revskill 8y ago

What people failed to define themselves is "Simplicity". That concept "Simplicity" needs context. Is it API surface simple ? Is it data model simple ? Is it Lines of Code short ? Is it extensible with small core and huge plugins system ? ... Is Wordpress simple ? Is MERN simple ? Is Rails simple ? There's no answer here. That's it. There's no "absolute simplicity", it depends on context.

Razengan 8y ago
For that matter, what about the definition of "perfect"?
lucio 8y ago
How many new concepts a new user needs to learn, and how many mental models a new user must develop in order to effectively use your software. Twitter is simple, Git is complex.
- cheez 8y ago
  Git is pretty simple. https://github.com/git/git/blob/ecbdaf0899161c067986e9d9d564586d4b045d62/ll-merge.c https://github.com/git/git/blob/ecbdaf0899161c067986e9d9d564...
  - dreamcompiler 8y ago
    Git is one of those rare products whose internal implementation is simple but whose user interface is a complicated mess.
  - dsjoerg 8y ago
    http://stevelosh.com/blog/2013/04/git-koans/ http://stevelosh.com/blog/2013/04/git-koans/
- rileymat2 8y ago
  I think this is close to a pretty good definition, but it should be restricted to people who already have learned the mental models of the domain. In that the mental models learned to use the software should match those of the domain.

stolee 8y ago

The comparison to hill-climbing algorithms is an interesting one, especially because the author ignores the fact that hill-climbing algorithms get stuck in local optima unless there is a way to “restart” the search from a new location. There is a time and place for big, risky changes to advance beyond what this “patch” approach can do.

anbop 8y ago
I think this is what happened to Amazon. A series of locally optimized decisions like commingled inventory and 3rd party sellers that has turned the site into a used car dealership in some categories.

simplyinfinity 8y ago

Can we get ( 2012 ) tag please

gfodor 8y ago

One problem with the meme that building products is about "solving problems" is that the biggest successes often come from recognizing very vague "problems" others don't, in the form of seeing fundamentally better ways of doing something -- without being able to grasp up-front the long term concrete benefits. You need to build and ship to learn them. So in those scenarios, people don't recognize what you are doing as a "real" problem, so it ends up being hindsight bias when it's recognized as such. An example in the article cited was Wikipedia: Wikipedia didn't "solve a problem" with existing encyclopedias, it recognized there was just a fundamentally better way to collect objective knowledge and disseminate it -- most people at the time were scratching their heads about what the point was, since from their perspective encyclopedias worked perfectly well within the scope of their design goals. I think the mental model that you should always be working on software where you have people directly telling you the problem they want to solve is a poor one -- it's great for a variety of domains, but often people want a faster horse or don't even see a problem with their existing horse. Even when you think you are recognizing a problem others don't see yet -- it can require a leap of faith to begin exploring to see if you can iterate and react to discovering something with big promise.

ncmncm 8y ago

The rarest skill, even in highly intelligent engineers, is to be able to imagine the world after a thing has been changed, and see what will matter to people then. Everyone else (including you, most likely) can only see the change itself. Is the change pretty? No awareness at all of whether the product after the change is nicer than the present product. It is a skill that can be developed, but it takes focused attention. The end result is to see how to alter the change to make the end result better to use, not just a thing with another PITA feature.