6 ms·

Reversible and irreversible decisions (2018)

110 points by durmonski 4y ago

nerdponx 4y ago
Interesting framework. Perhaps one of the difficulties is that decision reversibility tends to lie on a spectrum between "reversible" and "irreversible", and sometimes it's hard to even know where on that spectrum it lies. Thus you end up having to figure out if a decision is "reversible", maybe even first needing to develop a decision framework in which to make such a determination. I think all of the truly difficult decisions I've made in work and life have been in this category. Still, I appreciate this concept and I think it's better to have the mental model than not.
- steveBK123 4y ago
  Right, Bezos & this blog are framing the world as black & white when its really just lots of gray. Many things that you could bucket into "type 2" decisions which are not one-way doors, effectively are. For example, orgs that tend to never go back and fix technical debt means everything is a one-way door. Further, design decisions that require substantial and growing effort once you go through the door .. are effectively one-way doors as well. Maybe ZIRP era over staffed FAANG could afford to build and rebuilt, have multiple competing redundant systems to see who wins, etc.. but most tech orgs just don't work that way.
  - robocat 4y ago
    You are framing it as black & white - I think miscategorising the article and Bezos. From the article: Decisions fall somewhere on the continuum from reversible to irreversible. You can tell where a decision lies on this spectrum by asking how much it would cost to undo. The higher the cost to undo, the more irreversible it is. The lower the cost, the more reversible it is. Bezos considers 70% certainty as the cut-off point where it is appropriate to make a decision. That means acting once we have 70% of the required information, instead of waiting longer.
- havkom 4y ago
  Yes. And when it comes to having multiple people and teams involved and starting work it may become costly to reverse an otherwise reversible decision.
simonw 4y ago
I've used a similar model successfully in work situations where a decision is taking too long to make but is something that's reversible: pick a path, and put a note in the calendar to review that decision in a couple of weeks time. If it turns out to be the wrong direction you can course correct then. This really helps if not everyone on a team is convinced that it's the right decision: they can commit to it knowing that there will be a chance to change direction pretty soon if it turns out not to work.
- ghaff 4y ago
  Way back when I was a product manager the team would constantly want decisions in this or that. The reality was that a lot of these decisions really didn’t matter or didn’t matter much and could be course corrected. So the best course of action was to make some decision any decision. It was usually more informed than flipping a coin but it often didn’t need a lot of deep study either.
FinanceAnon 4y ago
I use this model a lot in life. I like to remember about this when in arguments with someone - once you say something that you later regret, it's very difficult to take it back.
photochemsyn 4y ago
I've been learning a bit about distributed databases and of course AWS is a major user of such systems. The decision to do a mass migration from one distributed system to another, as an example, doesn't fit well with the kind of 'executive boardroom' mentality that this article is talking about. The fundamental issue is more just taking the possiblity of error and failure of strategy into account. Hence, having a rollback strategy in place beforehand is wise. This post discusses a few basic recovery strategies plus more complicated ones: https://aws.amazon.com/blogs/database/rolling-back-from-a-migration-with-aws-dms/ https://aws.amazon.com/blogs/database/rolling-back-from-a-mi... Bezos is likely well aware that the real world is always more complex than two simple choices, but that's too much information for a sales-confidence pitch to shareholders, I imagine.
scared333 4y ago
I guess russian leadership failed to consider this when they started their illegal imperial war to annex Ukraine. People of russia are almost irreversibly becoming the new 'nazis' of whose attrocities the kids will learn at school and of whom they will learn to despise for the rest of of their lives (while the earlier ones will be just another chapter in the history books that is read with the same enthusiasm that most study geopolitics behind WW1 currently). This is not to suggest that the attrocities (at this point anyway) would be comparable, but that history is moving ahead. Sorry about political angle, but due to my circumstances, I am very angry of what is happening, and I think these things cannot be repeated too many times.
- Animats 4y ago
  No, that's when you thought you were making a reversible decision but events played out in a way that made it irreversible.
- uri4 4y ago
  Everything with US is reversible, just wait until next election... I really hope not too many people will die until then. Iran, Afghanistan, Lybia, Syria, Jemen... another pointless war. World needs something better!
- creatorbytes 4y ago
  What Russia is doing is horrible. War crimes and all. However I wouldn’t go so far as call them nazis, think it dampens that metaphor when it’s over used. Russia feels NATO expansion is too close to their doorstep, he warned many times and this is the outcome. Obviously not condoning it, and it’s many millions of lives displaced and tens of thousands lost. Generations will be effected. Russia using the excuse of nazis in Ukraine, and the need to free the Ukrainian people is a tall tail. Though there is a small amount of truth, of actual nazis existing in some of the armed forces of Ukraine [0]. But by no means justifies what’s been done. [0] https://www.nbcnews.com/think/opinion/ukraine-has-nazi-problem-vladimir-putin-s-denazification-claim-war-ncna1290946 https://www.nbcnews.com/think/opinion/ukraine-has-nazi-probl...
  - 4y ago
city17 4y ago
I try to apply a similar mental model to new purchases. It can be tempting to weigh all the pros and cons of different products and find out a detailed list of all the possible alternatives. But often you just have to try a product to find out if it works for you. In that case it can help to think about what risk there is in just buying something that seems ok. If you can return it easily, or if it doesn't lose much value in using and you can sell it used for 90% of the new price then there's no real downside to just making a quick decision. Detailed analysis is only really worth it for more 'irreversible' purchases where it's either a big hassle to undo the purchase (buying a house) or there's a large cost (cars lose value quickly).
- lazide 4y ago
  What goods can you easily get 90% of the purchase price back? In my experience, unless you’re spending a lot of time and effort selling, 50-60% is the best you can hope for most of the time.
nonrandomstring 4y ago
Decision reversibility is much overlooked in resilience engineering. In a perfect world we'd have a kind of quantum uncertainty, where we forked reality into two streams, and maintained two options until one or the other proved a safe passage. Quoting from Digital Vegan; "Consider how the UK government let our drinking water reservoirs be sold off for property development, believing that advanced JIT (just in time) management technology, smart metering and so forth, would dispense with them. Then climate change came. Reservoirs are like power supply capacitors; they absorb as well as smooth out supply. Now in the UK we have housing estates built on flood plains. Rivers burst their banks with every downpour. Knocking down thousands of peoples' houses to regain reservoir capacity is much /harder/ than it was to sell the reservoirs to developers." The transition from a reservoir to a housing estate looks like a net gain in "order" (entropy reduction) because it seems to create value, but considering the system as a whole (cost of losing infrastructure) it increases disorder. Similarly our gushing project toward an "online cashless society" is playing with dangerous forces. It's a net destruction of wealth and order. It won't be cheap or quick to re-open shops, print and distribute cash, install ATMs and money handling facilities once the terrifying brittleness of a wholly digital economy becomes clear. Those imagining "Nothing can possibly go wrong with the all Bitcoin + Amazon society", haven't thought through the reality of what happens when tens of millions of people can't get food even though it's in a warehouse less than 100 miles from their house.
- verisimi 4y ago
  Is it impossible to conceive or state the idea that these governmental actions are by design?
  - DiggyJohnson 4y ago
    Why would an elite class of people or government want to collapse its own drinking water supply or financial system?
    - didericis 4y ago
      To control people through selective access to alternatives they've set up.
noduerme 4y ago
This may work for business, where only money is at stake. But the hardest decisions in life are the ones where both options are irreversible. e.g. choosing to have an abortion or have a baby with someone you barely know. There's no reversible way to check the other side of either door.
- JosephConradPl 4y ago
  I think using the word irreversible for both outcomes is entirely a matter of perspective. A person choosing to abort can still be a mother which would technically mean that option is highly reversible. Really just depends on what you consider the start and end states before and after either option is done.
  - noduerme 4y ago
    There are a number of reasons a person might not be able to get pregnant again after an abortion. Abortion always carries a risk of sterilization. Other life factors such as age or health can also mean that a pregnancy can't happen again.
fnordpiglet 4y ago
This was a key lesson I learned creating and deploying aws services at Amazon.
- iLoveOncall 4y ago
  Yeah, one-way door decisions VS two-way door decisions is how it's called in Amazon and while it's not a leadership principle in itself it is also ingrained in the culture.
  - fnordpiglet 4y ago
    We generally called them irreversible decisions and referred to the door analogy to explain it to folks unfamiliar to the culture.
lob_it 4y ago
Spatial data makes a lot of the 1 dimensional thinking more and more of a 20th century relic and even less of an opinion. "Those variables are unique to them" is an easy way to rationlize. The culture divide make so many variables moot/invalid, since lifestyle, language and customs may not even associate with any of the variables (one-way/two-way doors for example). As an individual, its easy to simplify. As someone trying to explain something, its easy to oversimplify, missing any/all of the details, experience and even the accumulated wisdom in basing the decision(s). Reversible/irreversible is not usually a carrot on a stick. Even with health, detrimental/beneficial usually has more merit than any cost weighted variables (you can never buy your health back). 2=1+1 has worked just fine in traditional science for centuries. The (2) is usually unique though. All that to say.... This article did nothing for me :)
- JosephConradPl 4y ago
  I'm not sure if you really understood anything about what the article is actually trying to say and instead you got caught up on the types of examples it uses to help illustrate the framework for making decisions. Could you clarify why exactly the article did nothing for you?
  - lob_it 4y ago
    "Make reversible decisions as soon as possible and make irreversible decisions as late as possible." The summary sounded like a rash judgement. Is A/B/C testing reversible, irreversable or a tool for clarity in goal making decisions? The article appeared to be 1 dimensional. I was fortunate enough to have a website that crested into the top 800,000 websites years ago (a solo endeavor)... And backups made every decision reversible, if that was even a thing. I did have to commit to goals though and not once was anything irreversable. The 1st world still bases decisions on the best data available.
midenginedcoupe 4y ago
The way I've framed it is the cost of making the decision should be in proportion to that decision turning out to be wrong. It's extremely rare for decisions to be as binary as one-way or two-way, but there's always a point on the sliding scale of cost to change later.

110 points by durmonski 4y ago

Interesting framework. Perhaps one of the difficulties is that decision reversibility tends to lie on a spectrum between "reversible" and "irreversible", and sometimes it's hard to even know where on that spectrum it lies. Thus you end up having to figure out if a decision is "reversible", maybe even first needing to develop a decision framework in which to make such a determination. I think all of the truly difficult decisions I've made in work and life have been in this category. Still, I appreciate this concept and I think it's better to have the mental model than not.

steveBK123 4y ago
Right, Bezos & this blog are framing the world as black & white when its really just lots of gray. Many things that you could bucket into "type 2" decisions which are not one-way doors, effectively are. For example, orgs that tend to never go back and fix technical debt means everything is a one-way door. Further, design decisions that require substantial and growing effort once you go through the door .. are effectively one-way doors as well. Maybe ZIRP era over staffed FAANG could afford to build and rebuilt, have multiple competing redundant systems to see who wins, etc.. but most tech orgs just don't work that way.
- robocat 4y ago
  You are framing it as black & white - I think miscategorising the article and Bezos. From the article: Decisions fall somewhere on the continuum from reversible to irreversible. You can tell where a decision lies on this spectrum by asking how much it would cost to undo. The higher the cost to undo, the more irreversible it is. The lower the cost, the more reversible it is. Bezos considers 70% certainty as the cut-off point where it is appropriate to make a decision. That means acting once we have 70% of the required information, instead of waiting longer.
havkom 4y ago
Yes. And when it comes to having multiple people and teams involved and starting work it may become costly to reverse an otherwise reversible decision.

simonw 4y ago

I've used a similar model successfully in work situations where a decision is taking too long to make but is something that's reversible: pick a path, and put a note in the calendar to review that decision in a couple of weeks time. If it turns out to be the wrong direction you can course correct then. This really helps if not everyone on a team is convinced that it's the right decision: they can commit to it knowing that there will be a chance to change direction pretty soon if it turns out not to work.

ghaff 4y ago
Way back when I was a product manager the team would constantly want decisions in this or that. The reality was that a lot of these decisions really didn’t matter or didn’t matter much and could be course corrected. So the best course of action was to make some decision any decision. It was usually more informed than flipping a coin but it often didn’t need a lot of deep study either.

FinanceAnon 4y ago

I use this model a lot in life. I like to remember about this when in arguments with someone - once you say something that you later regret, it's very difficult to take it back.

photochemsyn 4y ago

I've been learning a bit about distributed databases and of course AWS is a major user of such systems. The decision to do a mass migration from one distributed system to another, as an example, doesn't fit well with the kind of 'executive boardroom' mentality that this article is talking about. The fundamental issue is more just taking the possiblity of error and failure of strategy into account. Hence, having a rollback strategy in place beforehand is wise. This post discusses a few basic recovery strategies plus more complicated ones: https://aws.amazon.com/blogs/database/rolling-back-from-a-migration-with-aws-dms/ https://aws.amazon.com/blogs/database/rolling-back-from-a-mi... Bezos is likely well aware that the real world is always more complex than two simple choices, but that's too much information for a sales-confidence pitch to shareholders, I imagine.

scared333 4y ago

I guess russian leadership failed to consider this when they started their illegal imperial war to annex Ukraine. People of russia are almost irreversibly becoming the new 'nazis' of whose attrocities the kids will learn at school and of whom they will learn to despise for the rest of of their lives (while the earlier ones will be just another chapter in the history books that is read with the same enthusiasm that most study geopolitics behind WW1 currently). This is not to suggest that the attrocities (at this point anyway) would be comparable, but that history is moving ahead. Sorry about political angle, but due to my circumstances, I am very angry of what is happening, and I think these things cannot be repeated too many times.

Animats 4y ago
No, that's when you thought you were making a reversible decision but events played out in a way that made it irreversible.
uri4 4y ago
Everything with US is reversible, just wait until next election... I really hope not too many people will die until then. Iran, Afghanistan, Lybia, Syria, Jemen... another pointless war. World needs something better!
creatorbytes 4y ago
What Russia is doing is horrible. War crimes and all. However I wouldn’t go so far as call them nazis, think it dampens that metaphor when it’s over used. Russia feels NATO expansion is too close to their doorstep, he warned many times and this is the outcome. Obviously not condoning it, and it’s many millions of lives displaced and tens of thousands lost. Generations will be effected. Russia using the excuse of nazis in Ukraine, and the need to free the Ukrainian people is a tall tail. Though there is a small amount of truth, of actual nazis existing in some of the armed forces of Ukraine [0]. But by no means justifies what’s been done. [0] https://www.nbcnews.com/think/opinion/ukraine-has-nazi-problem-vladimir-putin-s-denazification-claim-war-ncna1290946 https://www.nbcnews.com/think/opinion/ukraine-has-nazi-probl...
- 4y ago

city17 4y ago

I try to apply a similar mental model to new purchases. It can be tempting to weigh all the pros and cons of different products and find out a detailed list of all the possible alternatives. But often you just have to try a product to find out if it works for you. In that case it can help to think about what risk there is in just buying something that seems ok. If you can return it easily, or if it doesn't lose much value in using and you can sell it used for 90% of the new price then there's no real downside to just making a quick decision. Detailed analysis is only really worth it for more 'irreversible' purchases where it's either a big hassle to undo the purchase (buying a house) or there's a large cost (cars lose value quickly).

lazide 4y ago
What goods can you easily get 90% of the purchase price back? In my experience, unless you’re spending a lot of time and effort selling, 50-60% is the best you can hope for most of the time.

nonrandomstring 4y ago

Decision reversibility is much overlooked in resilience engineering. In a perfect world we'd have a kind of quantum uncertainty, where we forked reality into two streams, and maintained two options until one or the other proved a safe passage. Quoting from Digital Vegan; "Consider how the UK government let our drinking water reservoirs be sold off for property development, believing that advanced JIT (just in time) management technology, smart metering and so forth, would dispense with them. Then climate change came. Reservoirs are like power supply capacitors; they absorb as well as smooth out supply. Now in the UK we have housing estates built on flood plains. Rivers burst their banks with every downpour. Knocking down thousands of peoples' houses to regain reservoir capacity is much /harder/ than it was to sell the reservoirs to developers." The transition from a reservoir to a housing estate looks like a net gain in "order" (entropy reduction) because it seems to create value, but considering the system as a whole (cost of losing infrastructure) it increases disorder. Similarly our gushing project toward an "online cashless society" is playing with dangerous forces. It's a net destruction of wealth and order. It won't be cheap or quick to re-open shops, print and distribute cash, install ATMs and money handling facilities once the terrifying brittleness of a wholly digital economy becomes clear. Those imagining "Nothing can possibly go wrong with the all Bitcoin + Amazon society", haven't thought through the reality of what happens when tens of millions of people can't get food even though it's in a warehouse less than 100 miles from their house.

verisimi 4y ago
Is it impossible to conceive or state the idea that these governmental actions are by design?
- DiggyJohnson 4y ago
  Why would an elite class of people or government want to collapse its own drinking water supply or financial system?
  - didericis 4y ago
    To control people through selective access to alternatives they've set up.

noduerme 4y ago

This may work for business, where only money is at stake. But the hardest decisions in life are the ones where both options are irreversible. e.g. choosing to have an abortion or have a baby with someone you barely know. There's no reversible way to check the other side of either door.

JosephConradPl 4y ago
I think using the word irreversible for both outcomes is entirely a matter of perspective. A person choosing to abort can still be a mother which would technically mean that option is highly reversible. Really just depends on what you consider the start and end states before and after either option is done.
- noduerme 4y ago
  There are a number of reasons a person might not be able to get pregnant again after an abortion. Abortion always carries a risk of sterilization. Other life factors such as age or health can also mean that a pregnancy can't happen again.

fnordpiglet 4y ago

This was a key lesson I learned creating and deploying aws services at Amazon.

iLoveOncall 4y ago
Yeah, one-way door decisions VS two-way door decisions is how it's called in Amazon and while it's not a leadership principle in itself it is also ingrained in the culture.
- fnordpiglet 4y ago
  We generally called them irreversible decisions and referred to the door analogy to explain it to folks unfamiliar to the culture.

lob_it 4y ago

Spatial data makes a lot of the 1 dimensional thinking more and more of a 20th century relic and even less of an opinion. "Those variables are unique to them" is an easy way to rationlize. The culture divide make so many variables moot/invalid, since lifestyle, language and customs may not even associate with any of the variables (one-way/two-way doors for example). As an individual, its easy to simplify. As someone trying to explain something, its easy to oversimplify, missing any/all of the details, experience and even the accumulated wisdom in basing the decision(s). Reversible/irreversible is not usually a carrot on a stick. Even with health, detrimental/beneficial usually has more merit than any cost weighted variables (you can never buy your health back). 2=1+1 has worked just fine in traditional science for centuries. The (2) is usually unique though. All that to say.... This article did nothing for me :)

JosephConradPl 4y ago
I'm not sure if you really understood anything about what the article is actually trying to say and instead you got caught up on the types of examples it uses to help illustrate the framework for making decisions. Could you clarify why exactly the article did nothing for you?
- lob_it 4y ago
  "Make reversible decisions as soon as possible and make irreversible decisions as late as possible." The summary sounded like a rash judgement. Is A/B/C testing reversible, irreversable or a tool for clarity in goal making decisions? The article appeared to be 1 dimensional. I was fortunate enough to have a website that crested into the top 800,000 websites years ago (a solo endeavor)... And backups made every decision reversible, if that was even a thing. I did have to commit to goals though and not once was anything irreversable. The 1st world still bases decisions on the best data available.

midenginedcoupe 4y ago

The way I've framed it is the cost of making the decision should be in proportion to that decision turning out to be wrong. It's extremely rare for decisions to be as binary as one-way or two-way, but there's always a point on the sliding scale of cost to change later.