after conducting a dozen interviews with various systems thinkers from around the world and across many disciplines one of the things that became very clear is that feedback loops of various kinds are really critical to all systems thinking uh again whether it's education engineering agriculture uh economics all of the above feedback loops and understanding uh what feedback loops are and how to use them is the way to go now you might think that I might be jumping into virtuous cycles and Vicious Cycles and we will into that but it actually has more to do with measurements insights so a feedback loop is comprised of a few things that you need to know of there are just a few core ideas that go into a feedback loop that may or may not be obvious unless you have actually studied this um and put it into practice so first is measurement time so there is an appropriate time in place to make a measurement and then there is uh you have to glean actionable insights from that measurement and then you have to make sure that the right people get that information information and know what to do with it there's also an optimal length of time between measurement and action um so for instance some feedback loops need to be very tight very short sometimes seconds or minutes um depending on what you're doing like if you're operating a machine um sometimes it needs to be seconds or subseconds um for instance feedback loops inside of machines uh or programs but in other cases feedback loops are much longer uh either due to the complexity of collecting the data or because you actually need to wait to see what the impact was so this is things like GDP measurements um quarterly reports those sorts of things but the the point of the matter is you do need to collect information and then make sense of it and make sure that it's also disseminated to the correct people and that you're optimizing for a kpi some kind of measurement that you're trying to change uh which we'll talk about uh later in the video so when I talk about measurements usually we're talking about first principles which is putting a number to something you can't always always put a number to something you often have to look for proxies um but in general the better data you have and the more accurate data you have the better decisions that you can make and so that it always comes down to measurements now in some cases there are proxies and I'll talk about proxies versus direct measurements in just a second um but uh the the tldr is you need to put a number onto what it whatever it is that you're trying to optimize or increase or decrease so if you're trying to increase subscriber growth or increase revenue or decrease costs or decrease losses those sorts of things generally it comes down to comes down to numbers now human feedback is something that we can actually integrate more and more today with the use of generative AI such as chat GPT and other llms so for instance you can use these and in conjunction with things such as rubric tests so that you can create uh objective measurements of qualitative data so you can transition from from qualitative to quantitative data but we need to I need to make sure that that I emphasize the point that qualitative data and quantitative data are both data um there are there are many purists out there whether uh in various stem fields or whatever that kind of believe that math is Superior in all ways but um one thing that actually emerged in many cases was that uh you need to collect data from people and it's going to be squishy it's going to be subjective and it's going to be qualitative and sometimes that is actually Superior information one of the lessons that was taught to me by one of the systems thinkings interviews that I did was a principle actually that Jeff Bezos uses at Amazon which is if the humans give you a story that conflicts with the data trust the humans and so that's counterintuitive um to many Tech heads and scientists and and so on and so forth and so again human feedback is absolutely critical and and often if there's a conflict if the numbers don't align with what the people are saying that's usually where you're going to get the best Insight so it's not just a matter of trusting one over the other and saying oh well clearly you're wrong it's saying hey what's this discrepancy here um that actually kind of emerges as one of the most interesting sources of valuable data where it's like hey this measurement that we're making looks good but the people are saying something different so what's the actual story here another key component of the feedback loop is actionable insights basically whatever feedback you give what whatever whether it's a report or a chart or something it has to result in a decision change or a behavior change if there's no point if there's no desired Behavior change or decision change then you shouldn't be doing that feedback um if you're just giving feedback for the sake of giving feedback because you think that you're supposed to or because the charts look pretty it is literally just a waste of time um you know you might say okay well this data is exactly what I expected so I'm going to keep doing the behaviors that I'm doing that's fine um if you're using it as validation but again if you're not paying attention to the data if it's not telling you anything meaningful then it's basically just causing confusion um so the one way that you can think about this is It's Superfluous information and Superfluous information has a cognitive load it has a cognitive cost um which means that okay if you're going to have the cost just by virtue of you you want data to you know I don't know sooe your own anxiety or whatever um you also need to measure the effectiveness of the feedback and so you can't just measure a thing that you want to change you also have to look at okay did providing this feedback at that time in this format have the desired effect and so in that case you're kind of taking an even bigger step back and measuring the effectiveness of the feedback loop itself so you're measuring the measurement or you're measuring the feedback uh and that also serves as a as a key under underlying principle because a lot of people get fixated on feedback for feedback back sake and that doesn't help anything and in fact it can make things worse as we'll talk about with with Vicious Cycles near the end of the videos but uh just a a simple example I remember in fourth grade I had a friend who was an immigrant from Mexico and he was like a CD student I was an AB student but here's a fourth grader who's like what eight or nine years old and he gets a d with no explanation and so I remember he was just crying in the back of class and so what does that do it's like oh well you're a failure and nobody's going to explain to you what to do better and so guess what he just checked out an 8-year-old or a nine-year-old checking out because of bad feedback and so this is you probably hear me on on this channel and others where I constantly criticize uh education and grades this is one of the reasons is because um providing a number with no context or no actionable Insight or guidance or whatever is actually doing more harm than good um and it's true for 8-year-old old and 9-year-old children it's true for adults it's true for friends and family employees employers and so on now one thing that's really important is is understanding that a loop has intrinsically a temporal component and so what I mean is that there is a temporal Horizon uh Beyond which the measurement doesn't really mean anything if you measure GDP today it's not really going to mean anything in 50 years if you measure the position of a robot's hand right now it's not really going to mean anything in 2 minutes because it will have moved on um so that's what I meant when I say there's an optimal time frame in which to make a measurement and provide a feedback and there's there is such a thing as too fast there is such a thing as too slow and so you uh with whatever system you're working in whether it's agriculture or uh education or whatever you need to keep in mind okay what is what is the decision Point what is the original event or action um then when can I expect to actually measure that result or what how long does it take to get that good measurement and then how long does it take to provide that feedback and then you know do a course correction and so these temporal Horizons are actually the one of the most important parts it's time and measurement are kind of the two key components of feedback loops another example you know harping on education grades are a proxy for learning but then learning is a proxy for Success later in life and so this is one of the reasons that education is so contentious is because it's like well what are you measuring are you actually measuring the child's learning are you actually predicting how they're going to do in life and is it having the the desired impact later in life because if you just give a a child a d or an F with no explanation coaching or feed or other context for that feedback then you're actually setting them up for failure later in life because they just have internalized I'm an idiot I'm a failure I don't know anything because of a monotropic grade which doesn't measure the rest of their quality as a person um and so again I have a I haven't asked to grind with education in case you didn't notice but that is an example of a very longtime Horizon likewise parenting behaviors which can have very long time Horizons that play out over decades Financial policies school policies those kinds of laws uh you know deregulation for instance uh repeal of the glass Eagle act here in America that had an impact uh you know more than a decade later and so the longer the time Horizon also the more complexity and noise is introduced into the feedback and so it's like well was that thing that happened a decade ago or two decades ago really the cause here or was it only one of many causes uh and it can be it can be more difficult and so that's why one of the key principles that I that I have introduced here is in general as a as a rule of thumb it is always better to shorten feedback loops to whatever the most concise feedback loop is possible based on those temporal constraints so you might have heard of good heart's law good heart's law is a really important idea which basically says that once a measurement becomes a Target it ceases to become a good measurement and this is because of perverse incentives where you're trying to optimize for that one thing rather than the underlying truth of the thing and so again using grades as an example grades are a proxy for learning which is a proxy for uh success later in life um and so it's a proxy of a proxy which means that grades are generally a bad measurement um now un on the contrary though you can have direct measurements such as you know how long did it take from the time that you know your customer ordered their product to the time that it arrived at their door that is a very objective end to end measurement that is that is super direct and that is like okay you want to measure delivery time you got you got an exact number you know the exact results and you can unpack what you know like what expenses and other uh variables went into that and so while good Hearts law is a thing because if you're measuring GDP that's not necessarily what you're trying to optimize for it's a proxy if you're measuring uh grades it's a proxy it's a proxy of a proxy in some cases if you're trying to measure uh Revenue growth for the company quarterly reports annual reports again these are proxies or Aggregates um of other numbers which don't necessarily speak to the underlying the actual performance metrics and so uh this is the key thing that I wanted to drive home which is that proxies are useful in in only in cases where you cannot make a direct measurement where like you know in the case of grades you can't actually open up a child's brain and look at the synaptic connections that are made and say oh like this is you you actually know what you're talking about we've actually made the the the uh neurobiological changes that we want to make um and then of course likewise if you give a child a report card and that looks and they feel like they're an object failure um it's a bad measurement because it doesn't take the child's emotions or sense of self into account so again education as it is today is a very flawed system let's say um now more specifically though in business or Logistics or whatever you can usually put an objective measurement like how much how much uh crop yield did this acre of land produce you know was it one ton of potatoes or 50 tons of potatoes if you can directly measure the output of something the the the like empirical uh uh result whether it's mass weight energy time that's usually going to be a better metric and so in those cases good heart's law is much much less likely to apply so one of the interesting ideas that came up in one of my interviews was the idea of a kpi tree so a kpi tree is not when you measure one thing and say okay that's the be all end all of it like you know what was our revenue for this year or what was the um you know average time for delivery or what was you know the average customer review instead a kpi tree is where you say okay what are the other kpi that uh that give rise to this value what are the uh positive correlates and the inverse correlates what are the caus causal and causitive relationships that create this result and so um an example is when you have aggregate kpi where it's like okay um just using using a system that I'm in right now which is um YouTube you're watching me on YouTube so I can look at all kinds of variables such as um subscriber growth click-through rate uh watch time uh how far video spread you know what's their reach and engagement and those sorts of things and so there's there's all these little you know kpis and numbers and all kinds of data all over the place that you can look at and you know I watch I watch other creators and and tutorials and coaches on here and they say oh well you know if you're if you lose viewers at a certain point in a video then look at what happened at that point in the video and remove that from your future thing so then you make a change and you say okay what was the down Downstream impact of this did your did your ad Revenue go up or down did your subscriber growth go up or down and so then it's like okay well what is it that you're trying to optimize for because like subscriber growth is one thing ad revenue is another watch time is another clickthrough rate like so but then it's like what what exactly is it that you're trying to optimize for and you can see how all of these various kpi kind of fit together to form a bigger picture and so that's what I mean by kpi trees and of course it's going to vary from industry and and even um from one Department to another whether you're in logistics or marketing or legal or whatever you know cuz there's always going to be various things that you're trying to optimize for values that you're trying to increase or decrease you're trying to decrease ship time you're trying to increase engagement or whatever but there's going to be multiple variables um pretty much all effects that you get are multivariate but at the same time there's usually also going to be key signals or essential signals in that data where it's like okay okay the rest of this information is just noise and there's really one thing that you're trying to optimize for and in in the case of YouTube the one thing that I'm trying to optimize for is subscriber growth so it's not by accident that my main channel has um at at time of recording over 130,000 subscribers and it's also not an accident that this channel that you're watching at the time of recording has almost 8,000 subscribers this channel started at something completely different this channel started as a channel exploring autism in ADHD and then I uh part of part of my autism is that I'm a systems thinker I'm a systemizer and so I made a couple videos about that and that got way way better views and way better subscriber growth so what I what I did was I pivoted my channel towards that essential signal so it said okay based on the first 30 videos that I made which ones got the most subscribers and it was Far and Away the systems thinking videos so here we are and so that is an example of identifying the the the key the quintessential kpi which is subscriber growth and then all all of the other kpi under that such as click-through rate conversion rates and those sorts of things that lead up to that um video length is another kpi uh watch time all of that so I hope that example really kind of uh gives you an idea of what I mean by a kpi tree where for me the root of the tree is subscriber growth and then there's all these other little things that kind of filter down into that kpi so speaking of there's usually also um on top of having key measurements or you know like a root KP I there's also Key activities there's quintessential activities that uh generate the most value and so again using myself as an example because as a systems thinker and as someone who's here on YouTube what I discovered was that the most valuable activity for me was time in front of the camera um I did some math and it was it was very Fuzzy Math um and I I kind of calculated out I was like okay based on annual growth and revenue and all this sort of stuff or at the end of the year um every hour that I spend in front of the camera roughly translates to $1,200 so I was like I make $1,200 an hour in front of the camera now obviously there's a lot of other activities that go up to that I need to make these slide decks I need to you know manage patreon and Discord and upload videos and make thumbnails there's all kinds of stuff but the quintessential the most valuable activity that I do um is time in front of the camera everything that that I do uh revolves around that because again the key kpi is subscriber growth and you only get subscriber growth if you make videos airgo my key activity is making videos seems kind of obvious when you put it that way um but it one it was not obvious because I was trying to do all kinds of other stuff I had offers coming at me you know unsolicited offers for startups and Partnerships and spe speaking engagements and I tried all those because at first I thought okay well that would be really lucrative if someone's going to give me you know a few thousand dollars to come do a presentation but then it turns out actually that presentation takes weeks and weeks of planning and negotiation and I could have spent that 20 40 50 100 hours making videos and I would have had much higher return um by focusing on my key activity and so that's where that's why I said like it you really do have to measure everything and also be very clear about what it is that you're trying to optimize for um and you need a theory of that and so for instance here on on YouTube um the best book that I read and I I don't have a slide for it but the best book that I read was automat a customer so automatic customer uh is all about subscri subscription based models whether it's social media or LinkedIn or newsletters or whatever and it said ah the number one thing that you're trying to optimize for is subscriber growth I said got it cool here's all the principles that go into that so then you let everything else go all of those other distractions the the invitations to speak and other stuff non-essential activity yes it could be lucrative yes it could be valuable but it doesn't speak to my main kpi which is subscriber growth so some stories from the front line and this actually came up multiple times across uh multiple interviews was some form of blame shifting or dodging responsibility and so if you're a systems thinker and you're whether you're a consultant or a business analyst or whatever you might uh encounter some resistance when you go talk to people and say hey I'm uh I'm trying to you know optimize you know the shipping times here or I'm trying to optimize the turnaround time here and most people will be defensive because they take it personally um and this is this is just human reactions um it's human nature and what I the way to get around that is you say look everyone can be doing exactly what they're supposed to be doing at 100% efficiency but if the system is wrong if they're not getting the information that they need at the right time if they're not getting the materials that they need at the right time then the system is is to blame and most people are not systems thinkers and so they say okay well I'm part of the system so therefore it's my fault um but really you like you just driving home the point that it's like look this is a bad design it's not your fault we just need to get the right information so that we can fix the system um that you're that you're a part of and again this is where uh uh prioritizing human feedback is more important than um than prioritizing numbers for the sake of just numbers now both are important I'm not going to say oh well you don't like the numbers so just ignore them or you don't like what the human is saying so just ignore them you need all of it because even a bad measurement is a measurement or even a negative result is still a result it's all data um now again you need to make sure that you're making the correct measurement and that the data that you're getting is accurate data and that it means what you think it means um but also you kind of have to read between the lines especially where humans are and so uh when when a human gives you feedback you have to keep in mind their context what do they know what is their what is their education what is their expertise and you know the the good old saying um to a hammer every problem as a nail right whether you're an engineer or a trucker or you know a carpenter or whatever like people are equipped with the tools that they're equipped with mentally or otherwise and there's nothing wrong with that but you know they will they will be able to give you insights and it's it's a matter of asking questions which is why um in my five pillars of systems thinking communication is the Far and Away the most important pillar is because all systems are human systems whether it's the producers the consumers the vendors or whatever and so being able to communicate with humans is the most quintessential skill of being a systems thinker and that that allows you to do these things such as have a meeting with someone and overcome their defensiveness and really get to the key issue of what's going on there okay so as promised let's talk about The Virtuous cycle so this is really what you're aiming for and so you know you can you can read the slide if you want to but I'll just tell you my story uh while you read the slide if you pause it if you want um so once I figured out what the actual core Behavior was that you know made me successful on YouTube and it's not just me it's all people that are successful on YouTube um is you you you identify the key kpi you identify the Key activities and you identify the key principles okay yes I realized subscriber growth is key that's how you make a living on YouTube great okay cool how do you do that you get better you produce content you have to produce content time in front of the camera so then you experiment you say okay well what content gets the most results and that's where you do those experiments like I did on this channel where I switch from autism and ADHD to systems thinking why because I I sent out something into the universe and then I got the signal back where one video was doing 10 times better than the rest of the videos cool okay not by accident but by systematically experimenting I figured out what it was the value that I was adding that people wanted more of and so that's that feedback that's that virtuous cycle and now the channel is growing it's monetized um and you know more subscribers means more views which means more reach which means that you get more click-through which means that you get shown to more people you get more Impressions and so on and so forth and that is a virtuous cycle The Virtuous cycle means that it continuously grows forever and ever and ever and this is also how you get exponential results so exponential results Bank on using tools Technologies and techniques that do not have constraints YouTube is a global platform with over 2 billion users and growing so therefore it's like okay my total addressable audience is up to two billion people I have 8,000 subscribers is right now on this channel which means my my room for growth is several orders of magnitude so it's like cool I have an infinite basically an infinite field and you know in a few years in 5 10 years maybe all 8 billion humans will be using YouTube who knows and then it's like once you get to 100 % of all humans like that's that so that's The Virtuous cycle but getting to that exponential growth or exponential growth rather is the way to get to that and you do that by integrating uh continuous feedback continuous improvements in order to uh basically maximize or optimize your value now the opposite of this is the Vicious Cycle so the Vicious Cycle is where something bad happens and it leads to another bad thing which leads to another bad thing and it all gets worse and worse and worse over time so I already gave you the example of where you give a child a bad grade and then they internalize that it changes their identity it changes their emotional uh engagement to school and they get worse grades and then that serves as proof that I'm an idiot I'm useless I'm never going to succeed in school so I just give up entirely that is an example of of a vicious cycle and why uh scho the school system is really bad for some students another example of a vicious cycle is debt so so particularly national debt is is the kind of thing that we're seeing in a runaway kind of Snowball Effect right now and the way that it works is that you start taking out some debt and then you end up having to cut some costs elsewhere because now you're instead of paying for essential goods and services you're paying off the debt and then that causes you to retract from other spending elsewhere which causes the economy to slow down which then causes you to have less Revenue to pay off the debt which means that you need to take out more debt and then you end up spending more money servicing the debt rather than on on essential goods and services which means the economy slows down even more so on and so forth until you end up with this collapse of the economy which if you watch my other channel prag pragmatic Progressive um this is one of the things that we are possibly heading towards here in America is uh is a debt Calamity so those are two examples of Vicious Cycles where uh feedback is either the system is operating kind of on its own and you either don't integrate the feedback or the feedback is bad or the feedback has a negative impact and then you misinterpret what to do with that feedback so for instance debt goes up it's like oh well let's decrease spending no actually According to some theories what you should do is debt goes up as you increase spending to keep the economy hot so that you can outgrow the debt or instead of taking on more debt you take on less debt um but again because of the feedback loops and because of all the perverse incentives and politics in particular this rarely happens until it gets bad to the point that you need austerity measures and deleveraging so where I'll end you with is uh develop a theory so when I've talked about you know grades are bad because this or kpi is good because this or subscriber growth is good because this or debt is good because of that um it all comes down to axiomatic statements and so a theory is a it's a math-based kind of explanation as to how things work which is both predictive meaning you you can anticipate the future with the theory but it's also prescriptive which means it says hey I'm going to I'm going to prescribe a certain set of actions or principles or rules of thumb to get the result that you want and so in in my case um the Axiom that I came up with was my time in front of the camera is the most valuable use of my time time in front of the camera time in front of the camera time let everything else go um as much as possible obviously like I said um you know there's other activities that lead up to this and there's a few activities after such as you know editing and you know uploading and that sort of thing but really it's time in front of the camera is the most valuable activity that I can be doing um likewise whether you you know you're trying to optimize for education or optimize the economy there are theories of these things and some of those things are much more complicated than just making videos on YouTube um again because of uh the the complexity of the system whether it's the number of people participating and the number of of of data streams or the time Horizon both the economy and education have much longer time Horizons and this is one of the reasons that YouTube is such a good platform is because I get instant almost instant feedback I usually get feedback within a within a couple days of how well a video is doing and it's like Okay cool so I have a theory and then I you know make a video I test the theory did it work okay cool update the theory and move on um but this this consciously and and deliberately developing a theory um is something that was implicitly done by a lot of the systems thinkers that I talk to where it's like in one example there there was a lot of cost being accured from the logistics department and so the theory that was developed was oh well we have all these trucks showing up at the same time and so then they have to wait so now the theory is make sure trucks Don't Wait and it's like oh that's a really simple axiomatic statement waiting trucks are bad um you know for me time in front of camera is good waiting trucks are bad that's what I mean by axiomatic statements and and once you distill it down if you watched my other videos on here you know that distillation is a key cognitive skill so once you look at all the data once you take all the stories into account once you look at all the kpi and you do some of those tests and experiments your brain will distill it all down um deliberately with help or or through incubation period and then you synthesize these axiomatic statements which can then you just write it on a board or you know write it on a card or it becomes you know a guiding principle for business practices and so then it's like ah waiting trucks are bad so then you tell your doc workers and your doc manager and your Logistics uh managers if a truck is waiting that's bad fix that and then you just give them that principle and they can operate by that rule of thumb everever and and on since then um and for me likewise if I find myself doing anything other than sitting in front of the camera uh recording a good video then I know that that's that activity is probably a waste of time um so yeah thanks for watching I hope you got a lot out of this let me know again feedback I read I read pretty much all the comments um and if you want more jump in on my patreon I do have some videos um in the backlog there as well as a monthly Town Hall that we run on Zoom webinars um it's basically a QA session I'll give a short presentation kind of my current thoughts on AI and whatever else is going on and then I'm also on Discord all day every day you get all of this um via my patreon so cheers have a good one