1. "It was like a bar going BYOB but still charging you 5$ a bottle to drink it at their tables." Some restaurants call this a "corkage fee." You guys should have called yours a "cauldronage fee."
2. “Sometimes it is this way, sometimes it is that way,” - extra-stealth additional Kling reference
1. Wow, I didn't even know that was a thing! I should drink more.
2. I do have a lot of affection for Kling, even if I disagree with him more as we age. :)
3. Thank you :) Hopefully I will have another couple finished by the end of the month. Work really has been frying my brain lately, and in a suboptimal way, productivity wise.
I think Kling on most occasions is either clear-thinking or transparent about his thought process, or both, even if I disagree with him on stuff. Aside from maybe Bryan Caplan and Richard Hanania, he's the most resistant to social desirability bias of the "public intellectuals" I pay attention to.
Oh yea, I think he is great, and have been reading him for years. I just think he trusts government a bit too much, including government data, and think he is far too ok with the COVID response. As he said, his neighbor died from it, apparently, and I see how that affects you. Much like Tyler Cowan, however, I think living around the DC swamp has infected him over time. Or at least his view of what "normal people" are like is rather different from mine :D
I will admit, it also irks the hell out of me that he never addresses the problem of "How are the incentives of the Chief Operating Officer any different than, say, a president? Aren't you just adding a layer of bureaucracy without changing much?" with regards to his idea there, no matter how often it gets brought up. It probably shouldn't bother me that much, because it isn't like being a ground breaking theorist is what he hangs his hat on, but damn, that's pretty central to the idea. One should either address it carefully or shut up about the theory being worth while. It is like he doesn't even fear the ghost of James Buchanan coming back to clip him one round the ear hole.
I was going to bring up Arnold's COVID views as well--he got his fifth booster? On what evidence? Granted, I don't know his personal health situation but his age bracket doesn't put him in the danger zone, and besides, his personal risk calculation about COVID shouldn't influence policy. (My pet theory is that COVID rules formulators and commentators are mostly either older or have more comorbidities than the average citizen, but then, through some sort of illogical rationalization process, import the sensibilities they have about themselves--which are based on their above-average risk profile--onto to the average citizen, based on the incorrect assumption/conclusion that they themselves represent the average citizen. So you get something like, "I have more risk than the average citizen"-->but I am an average citizen-->so average citizens probably have my risk profile." There is of course the additional problem of being part of the DC-based laptop class and thinking everyone can self-isolate as easily as you can. Coyote Blog had some really good early insights on that.....)
Oh, yes, I also think he's out to lunch on the state capacity libertarian COO-ombudsman setup he talks about. Drain the swamp by filling it with more water. Okaaaaaay.
I think your pet theory has some legs. I also think that as people age they feel like more of the "establishment" in a sense, and so tend to defer more towards those parts of the established order that have not specifically pissed them off yet. I am often amazed at how many people say they don't trust e.g. the CIA, but oh, this other part of government, they are pure and honest. Partly I think it is compartmentalizing, and partly I think it is because doing so is scary, and results in tearing down your ideas of how most of the system works. (and possibly also being morally obligated to start doing something about it.) I think many of the old guard libertarians tended to think of the CDC as a bunch of scientists being reasonable and not causing all sorts of trouble like all those captured and corrupted bureaucracies. Stopping to think "Wait, why the hell was I trusting them? Should I have been doing that?" is really hard when you are scared and want someone to tell you what to do to avoid the plague.
Not to mention there seems to be a sort of gentleman's agreement among many like Kling and say Zvi (another one I read a lot but disagree with) to never mention the problems with vaccines, even though there is just a huge burning dumpster fire of studies and other evidence of the fact they are net negative value. I say gentleman's agreement, because I am pretty sure they are all seeing others write about the situation, but strangely act as though they know nothing. *shrug* maybe they really don't, but it seems strange to me that it is the case.
I miss Coyote... I hope he gets done sorting out these new businesses of his soon and gets back to writing.
In keeping with the theory of the gentleman's agreement, Arnold and Tyler have been silent about the Oster piece even though almost everyone outside their respectability windows has done so.
"And some models are so wrong they screw up your understanding of the world to the point you have no idea what reality looks like, and you have no desire to find out anymore because it would be too much work."
XD
Seriously though... It seems like it should be easier to do activity based accounting with computers & spreadsheets. Unless I'm missing something....
I think if you had a REALLY well designed and calibrated ERP system you could probably pull off activity based costing easier, although the data requirements to simulate it would be tough. (Though, hell, getting the ERP system to do reasonable planning runs by itself is beyond the ability of many companies.) I know one issue is that Sarb-Ox makes it illegal to tell supply chain and operations what the final price of goods is (or something to that effect) which means your system can't take that into account.
I suspect the biggest problem is that you would have to simulate basically your entire company to a very accurate detail level to get better answers than "Eh, the old man thinks this is probably the best thing to do." The sheer amount of explicit data you would have to gather, confirm, enter and update would be huge, and the amount of implicit knowledge you couldn't capture would probably kill the project if the pile of explicit didn't.
I think in some way this gets to the general omniscience problem in complex adaptive systems: once they get to a certain size humans can't wrap their heads around them, and the problem gets too big to solve even with a computer because humans can't wrap their heads around the system well enough to simulate it in the computer and make sure it works there. Once your company gets so large that the sales directors live 100's of miles from the plant, and have probably never seen it, and can't keep track of the day to day operation anyway, you are well past the point of being able to really fine tune things. I expect computers are making it easier all the time, but with ever decreasing returns.
Possible dumb ideas incoming warning: I passed Accounting 101 with a C, and haven't done any since.
"I think if you had a REALLY well designed and calibrated ERP system you could probably pull off activity based costing easier, although the data requirements to simulate it would be tough."
What if instead of trying to calculate everything, you just calculated the big inputs? Good enough is perfect IMO.
"I know one issue is that Sarb-Ox makes it illegal to tell supply chain and operations what the final price of goods is (or something to that effect) which means your system can't take that into account."
What about keeping two sets of books. One cost based, the other ERP. ERP is only used to help determine your focus.
"The sheer amount of explicit data you would have to gather, confirm, enter and update would be huge, and the amount of implicit knowledge you couldn't capture would probably kill the project if the pile of explicit didn't."
Again, just get the big inputs correct.
"I think in some way this gets to the general omniscience problem in complex adaptive systems: once they get to a certain size humans can't wrap their heads around them, and the problem gets too big to solve even with a computer because humans can't wrap their heads around the system well enough to simulate it in the computer and make sure it works there. "
Machine Learning sounds like a good candidate for this. It's a black box, but if it works, it works.
That's just it, you have to get many of the little inputs correct too to get things working at all, and many more to get it working nicely. Especially down at the actual manufacturing level, the amount of fiddly detail to get decent automated production planning is immense. Generally the best way to deal with it is in fact to just get the big stuff right, and then manually have humans doing the rest, but that cuts out most of the automation opportunities.
The big problem is that so much of the interaction is endogenous, and complex, so linear programming is brutal. As a result, many of the approaches I have seen go for a brute force, "simulate this 10^x times and pick the best outcome". So your machine learning guess isn't far off. The system runs, produces a proposal based on the (horrifically) large pile of data input, and humans look at some of the exceptions and tweak stuff.
The trouble is then lots of that data input is variable instead of fixed, particularly endogenously variable. Further, the data grows exponentially. Say I can make 100 products at my plant, and store them in two different distribution centers. I now have 300 product/location combinations that need planning. Add a warehouse, I now have 400. Add a product, 404. At one point at the company in the first anecdote, I was responsible for getting the system to plan over 52,000 product/locations. 52,000 combinations, with entries for every day, a year and a half in the future.
The background planning jobs ran over about 10 hours, start to finish.
Machine learning, or some other agent based simulation approach would probably work. (I have putzed around with some ABS myself here.) The trouble with the super black box nature (existing ERP systems get pretty black boxy pretty fast too, so it isn't unique) is that when your expected outcome from the plan varies from reality, you don't know why and can't fix it. This is a problem (again for both types) because reality and the system simulation of it always drift over time. Maybe your workers are getting a little lazier than before, or more motivated and less lazy. Maybe your machine resources are starting to run at lower capacity, and are just being hustled along by plant operators who know how to coax the most out of them. Maybe that was the case, and your operator retired and his replacement is clueless. Maybe solar flares.
Maybe you build a new plant in Korea that exactly matches your existing US plant down to the blue prints, but for some reason it refuses to produce the same purity of product even with American techs walking it through the process (happened there, too.)
Sorry, I am tired and starting to ramble. Long story short, it is an interesting problem in the field, and a really serious problem. Any consultant who says they can whip something together quick that will solve all your problems through algorithms and the hot acronyms of the day is lying to you :D
Worth noting too is that all the data is manually collected and entered, not machine gathered. Incredibly labor intensive and prone to error. Even getting decent definitions of what the data should be, how it is defined, what each bit influences in the system, is apparently a herculean task. I say apparently, because honestly I think it doesn't need to be impossible, but everyone responsible seems to think they can just whip it together as a one off instead of carefully producing rules, data dictionaries and maps of how things work, then making sure there is a process for validating and correcting the data. Trying to whip it off quickly ends up being phenomenally more work, quite quickly, but still, everyone insists on doing things half assed and ad hoc...
There is a really good chance of a forthcoming series on "How Not to Succeed in Your ERP Implementation!"
"Any consultant who says they can whip something together quick that will solve all your problems through algorithms and the hot acronyms of the day is lying to you :D"
Even with all the complexity, variability, and speculative nature of business, as the accountant noted you have that profitability target to at least work off of to coordinate everything and ensure you're moving in the right direction. Without that target, bureaucracies start to think that modeling and gauging productivity is not only possible, but simple and straightforward. The absurdity of such conceit never ceases to astound me.
Yea, and even getting that profitability nailed down at the sub-"entire business" level is really tricky. As ever, one of the biggest problems we face is measurement.
1. "It was like a bar going BYOB but still charging you 5$ a bottle to drink it at their tables." Some restaurants call this a "corkage fee." You guys should have called yours a "cauldronage fee."
2. “Sometimes it is this way, sometimes it is that way,” - extra-stealth additional Kling reference
3. Nice article. Thanks.
1. Wow, I didn't even know that was a thing! I should drink more.
2. I do have a lot of affection for Kling, even if I disagree with him more as we age. :)
3. Thank you :) Hopefully I will have another couple finished by the end of the month. Work really has been frying my brain lately, and in a suboptimal way, productivity wise.
I think Kling on most occasions is either clear-thinking or transparent about his thought process, or both, even if I disagree with him on stuff. Aside from maybe Bryan Caplan and Richard Hanania, he's the most resistant to social desirability bias of the "public intellectuals" I pay attention to.
Oh yea, I think he is great, and have been reading him for years. I just think he trusts government a bit too much, including government data, and think he is far too ok with the COVID response. As he said, his neighbor died from it, apparently, and I see how that affects you. Much like Tyler Cowan, however, I think living around the DC swamp has infected him over time. Or at least his view of what "normal people" are like is rather different from mine :D
I will admit, it also irks the hell out of me that he never addresses the problem of "How are the incentives of the Chief Operating Officer any different than, say, a president? Aren't you just adding a layer of bureaucracy without changing much?" with regards to his idea there, no matter how often it gets brought up. It probably shouldn't bother me that much, because it isn't like being a ground breaking theorist is what he hangs his hat on, but damn, that's pretty central to the idea. One should either address it carefully or shut up about the theory being worth while. It is like he doesn't even fear the ghost of James Buchanan coming back to clip him one round the ear hole.
I was going to bring up Arnold's COVID views as well--he got his fifth booster? On what evidence? Granted, I don't know his personal health situation but his age bracket doesn't put him in the danger zone, and besides, his personal risk calculation about COVID shouldn't influence policy. (My pet theory is that COVID rules formulators and commentators are mostly either older or have more comorbidities than the average citizen, but then, through some sort of illogical rationalization process, import the sensibilities they have about themselves--which are based on their above-average risk profile--onto to the average citizen, based on the incorrect assumption/conclusion that they themselves represent the average citizen. So you get something like, "I have more risk than the average citizen"-->but I am an average citizen-->so average citizens probably have my risk profile." There is of course the additional problem of being part of the DC-based laptop class and thinking everyone can self-isolate as easily as you can. Coyote Blog had some really good early insights on that.....)
Oh, yes, I also think he's out to lunch on the state capacity libertarian COO-ombudsman setup he talks about. Drain the swamp by filling it with more water. Okaaaaaay.
I think your pet theory has some legs. I also think that as people age they feel like more of the "establishment" in a sense, and so tend to defer more towards those parts of the established order that have not specifically pissed them off yet. I am often amazed at how many people say they don't trust e.g. the CIA, but oh, this other part of government, they are pure and honest. Partly I think it is compartmentalizing, and partly I think it is because doing so is scary, and results in tearing down your ideas of how most of the system works. (and possibly also being morally obligated to start doing something about it.) I think many of the old guard libertarians tended to think of the CDC as a bunch of scientists being reasonable and not causing all sorts of trouble like all those captured and corrupted bureaucracies. Stopping to think "Wait, why the hell was I trusting them? Should I have been doing that?" is really hard when you are scared and want someone to tell you what to do to avoid the plague.
Not to mention there seems to be a sort of gentleman's agreement among many like Kling and say Zvi (another one I read a lot but disagree with) to never mention the problems with vaccines, even though there is just a huge burning dumpster fire of studies and other evidence of the fact they are net negative value. I say gentleman's agreement, because I am pretty sure they are all seeing others write about the situation, but strangely act as though they know nothing. *shrug* maybe they really don't, but it seems strange to me that it is the case.
I miss Coyote... I hope he gets done sorting out these new businesses of his soon and gets back to writing.
Good stuff here.
I agree about the odd silence on potential problems with vaccines. Very weird. I wonder what the motivation is to be silent.
This reminds of the gell-man amnesia effect, but applied to government bureaucracies rather than media outlets
https://www.goodreads.com/quotes/65213-briefly-stated-the-gell-mann-amnesia-effect-is-as-follows-you
In keeping with the theory of the gentleman's agreement, Arnold and Tyler have been silent about the Oster piece even though almost everyone outside their respectability windows has done so.
"And some models are so wrong they screw up your understanding of the world to the point you have no idea what reality looks like, and you have no desire to find out anymore because it would be too much work."
XD
Seriously though... It seems like it should be easier to do activity based accounting with computers & spreadsheets. Unless I'm missing something....
I think if you had a REALLY well designed and calibrated ERP system you could probably pull off activity based costing easier, although the data requirements to simulate it would be tough. (Though, hell, getting the ERP system to do reasonable planning runs by itself is beyond the ability of many companies.) I know one issue is that Sarb-Ox makes it illegal to tell supply chain and operations what the final price of goods is (or something to that effect) which means your system can't take that into account.
I suspect the biggest problem is that you would have to simulate basically your entire company to a very accurate detail level to get better answers than "Eh, the old man thinks this is probably the best thing to do." The sheer amount of explicit data you would have to gather, confirm, enter and update would be huge, and the amount of implicit knowledge you couldn't capture would probably kill the project if the pile of explicit didn't.
I think in some way this gets to the general omniscience problem in complex adaptive systems: once they get to a certain size humans can't wrap their heads around them, and the problem gets too big to solve even with a computer because humans can't wrap their heads around the system well enough to simulate it in the computer and make sure it works there. Once your company gets so large that the sales directors live 100's of miles from the plant, and have probably never seen it, and can't keep track of the day to day operation anyway, you are well past the point of being able to really fine tune things. I expect computers are making it easier all the time, but with ever decreasing returns.
Possible dumb ideas incoming warning: I passed Accounting 101 with a C, and haven't done any since.
"I think if you had a REALLY well designed and calibrated ERP system you could probably pull off activity based costing easier, although the data requirements to simulate it would be tough."
What if instead of trying to calculate everything, you just calculated the big inputs? Good enough is perfect IMO.
"I know one issue is that Sarb-Ox makes it illegal to tell supply chain and operations what the final price of goods is (or something to that effect) which means your system can't take that into account."
What about keeping two sets of books. One cost based, the other ERP. ERP is only used to help determine your focus.
"The sheer amount of explicit data you would have to gather, confirm, enter and update would be huge, and the amount of implicit knowledge you couldn't capture would probably kill the project if the pile of explicit didn't."
Again, just get the big inputs correct.
"I think in some way this gets to the general omniscience problem in complex adaptive systems: once they get to a certain size humans can't wrap their heads around them, and the problem gets too big to solve even with a computer because humans can't wrap their heads around the system well enough to simulate it in the computer and make sure it works there. "
Machine Learning sounds like a good candidate for this. It's a black box, but if it works, it works.
That's just it, you have to get many of the little inputs correct too to get things working at all, and many more to get it working nicely. Especially down at the actual manufacturing level, the amount of fiddly detail to get decent automated production planning is immense. Generally the best way to deal with it is in fact to just get the big stuff right, and then manually have humans doing the rest, but that cuts out most of the automation opportunities.
The big problem is that so much of the interaction is endogenous, and complex, so linear programming is brutal. As a result, many of the approaches I have seen go for a brute force, "simulate this 10^x times and pick the best outcome". So your machine learning guess isn't far off. The system runs, produces a proposal based on the (horrifically) large pile of data input, and humans look at some of the exceptions and tweak stuff.
The trouble is then lots of that data input is variable instead of fixed, particularly endogenously variable. Further, the data grows exponentially. Say I can make 100 products at my plant, and store them in two different distribution centers. I now have 300 product/location combinations that need planning. Add a warehouse, I now have 400. Add a product, 404. At one point at the company in the first anecdote, I was responsible for getting the system to plan over 52,000 product/locations. 52,000 combinations, with entries for every day, a year and a half in the future.
The background planning jobs ran over about 10 hours, start to finish.
Machine learning, or some other agent based simulation approach would probably work. (I have putzed around with some ABS myself here.) The trouble with the super black box nature (existing ERP systems get pretty black boxy pretty fast too, so it isn't unique) is that when your expected outcome from the plan varies from reality, you don't know why and can't fix it. This is a problem (again for both types) because reality and the system simulation of it always drift over time. Maybe your workers are getting a little lazier than before, or more motivated and less lazy. Maybe your machine resources are starting to run at lower capacity, and are just being hustled along by plant operators who know how to coax the most out of them. Maybe that was the case, and your operator retired and his replacement is clueless. Maybe solar flares.
Maybe you build a new plant in Korea that exactly matches your existing US plant down to the blue prints, but for some reason it refuses to produce the same purity of product even with American techs walking it through the process (happened there, too.)
Sorry, I am tired and starting to ramble. Long story short, it is an interesting problem in the field, and a really serious problem. Any consultant who says they can whip something together quick that will solve all your problems through algorithms and the hot acronyms of the day is lying to you :D
Worth noting too is that all the data is manually collected and entered, not machine gathered. Incredibly labor intensive and prone to error. Even getting decent definitions of what the data should be, how it is defined, what each bit influences in the system, is apparently a herculean task. I say apparently, because honestly I think it doesn't need to be impossible, but everyone responsible seems to think they can just whip it together as a one off instead of carefully producing rules, data dictionaries and maps of how things work, then making sure there is a process for validating and correcting the data. Trying to whip it off quickly ends up being phenomenally more work, quite quickly, but still, everyone insists on doing things half assed and ad hoc...
There is a really good chance of a forthcoming series on "How Not to Succeed in Your ERP Implementation!"
"Any consultant who says they can whip something together quick that will solve all your problems through algorithms and the hot acronyms of the day is lying to you :D"
lol
Thank you sempai
Even with all the complexity, variability, and speculative nature of business, as the accountant noted you have that profitability target to at least work off of to coordinate everything and ensure you're moving in the right direction. Without that target, bureaucracies start to think that modeling and gauging productivity is not only possible, but simple and straightforward. The absurdity of such conceit never ceases to astound me.
Yea, and even getting that profitability nailed down at the sub-"entire business" level is really tricky. As ever, one of the biggest problems we face is measurement.