#119 Data-Driven Thinking for the Everyday Life

Hi, I'm Maggie, VP of Curriculum at datacamp, and I know firsthand how overwhelming it can be to keep up with the rapid changes in data and AI. But staying ahead doesn't have to be hard. That's why our curriculum team has curated the best learning content to help you and your team stay ahead, covering everything from data and AI literacy for busy executives to cutting edge generative AI for developers, and everything in between. Whether you're upskilling yourself or your team, we make learning practical, engaging and flexible. It just works. Power your data and AI transformation journey on datacamp.com. You're listening to Data Framed, a podcast by Data Camp. In this show, you'll hear all the latest trends and insights in data science. Whether you're just getting started in your data career or you're a data leader looking to scale data driven decisions in your organization. Join us for in depth discussions with data and analytics leaders at the forefront of the data revolution. Let's dive right in.

Welcome to Data Framed. This is Ritchie. As the new year begins, it's a great tradition to set some resolutions for what you want to accomplish in the coming year or what you want to change about your life. Just as in business, where the most successful projects tend to be the ones where you use data to track their progress, the same is true of personal goals. If you set quantifiable goals, then track your progress towards those goals, you're much more likely to hit them. I really love this idea of bringing data driven thinking into everyday life, and here to provide tips on how to do that is Gary Wolf, the director and co founder of Quantified Self. Gary's been involved in personal tracking since long before the existence of fitness trackers and smartwatches and other mainstream tracking devices. He's probably seen more personal tracking projects than anyone else on the planet, so I'm keen to hear his advice on the subject. To everyone listening, I hope you get some great ideas for how to improve your life this year. Hi there Gary. Thank you for joining us today. I guess to begin with, can you just tell me a bit about yourself and what you do at Quantified Self?

Well, I think the reason we're talking is that there's a growing interest in how people can learn about themselves and answer their own questions using empirical methods, using their own data. And at the Quantified Self, we're a users group, an international community of people who are interested in self tracking and self research. We've been around since 2008 and we do all kinds of things to support that practice and to support that community. We publish, we do research, including some academic research and some more journalistic popular research. But the main thing we do is we try to take note of and archive and document the discoveries that people are making about themselves using their own data. That's sort of our, our role as community organizers. And it's something that has kept us going and kept us interested for almost 15 years.

Brilliant. And maybe we can go back to the beginning. How did you come to start Quantified Self?

Well, by vocation, I'm a journalist and my background was researching and writing about the cultural effects of new technologies. And I was at Wired magazine for many years. I was a contributing editor of the magazine and I was the executive editor of the website and started Wired News and other titles associated with Wired Online. And in the course of all of that work over years, I ended up asking questions about not so much the explicit promises of new technologies, but their underlying long term personal and cultural effects. And in about 2007, one of the editors at Wired, a colleague of mine named Kevin Kelly, and I went through a kind of deliberate process where we put up some of the things on a whiteboard that interested us a lot at that time. The development of location services, which were pretty new. Small sensors, electromechanical sensors that were coming into all the devices. Also simply the spread of computing capacity into our phones. And we put a circle around it and asked what is the name, what is the title, what is the headline around the personal effects of all of these technologies? We were thinking of it as very personal computing. The name that I put on it at that time, Quantified Self, which is the mouthful, was really a synonym for personal computing. In a sense. It's what happens when these technologies come all the way in into our clothes, onto our bodies. But it's also almost the reverse because personal computing is about the technology. It's computing, but it's personal. Whereas Quantified Self is really about the self. And it's about what happens when these explicit measurement practices meet the self. And I, I just got deeply interested in that. I started to write about it and publish about it and meet people who were doing really interesting things. And out of that came a series of conferences and a forum and a website and, and the other ways that a community makes knowledge for itself. And I became kind of the steward of that community.

That's very cool. And I think now when you have things like fitness trackers being quite a sort of popular device that many people own, it seems that this Idea of tracking yourself is slightly mainstream, but back in 2007, I can imagine that was quite a radical idea. I'd like to learn a little bit more about what your community, what your users are actually tracking about themselves. So what are the most common projects?

Yes, that's easy to do, but I think to really understand what people are tracking and why, I'd like to just go back to what you said a minute ago, which is that in 2007, 2008, it was really avant garde to do something like this, and now with all the fitness trackers, it's really mainstream. I think that that's true. And at the beginning, Quantified Self was really of interest mainly to people who had scientific and technical training, because it was not very easy to make records of your observations such that you could go back and analyze them, reason about them using that whole range of empirical techniques that we have at our disposal when we do have some scientific and technical training. Simply making the observations was really quite demanding. Of course, you could put yourself on a scale and measure your weight, and there were pedometers that you could wear and you could keep track of your steps. But for the most part, these measurements weren't digital. So there was always an element of copying that you had to do. And so the record keeping was really demanding. And that was one of the things that was interesting about it at first, is that you saw people do really complicated and demanding things, but in the service of very commonplace questions. So unlike in professional science, where the kinds of questions you ask are driven by history of scientific exploration in your field, and a lot of them are very refined and specialized, right? Because a lot of people have been asking questions about physical chemistry or the nature of astronomical phenomenon or something like that. So when you come along and you get trained, you're really getting trained also, not just in the techniques, but also in what kinds of questions to ask, right? And from the outside, a lot of these questions are incomprehensible without training. Well, in the case of the Quantified self, it was really quite different. You had people who had really often quite refined techniques, but the questions they were asking were very much everyday questions. How can I be more fit? How can I handle the effects of some kind of chronic disease, right? Or pain or digestive issues? Or how can I handle my mental health in a different way, right? Or how can I feel better when I sit down to work every day? Or how can I sleep better? These questions were not super refined, super specialized, super academic questions. They were everyday questions. But they were being attacked with scientific and technical methods. What happens next? You have an entire industry that is created to make the process of collecting data and recording data easier. So Fitbit comes along and sells hundreds of millions of devices and Apple Watch comes along. And also the sensors in the phone become more accessible to apps. So a lot of the things that were difficult become easy or easier. But the questions, and this gets to your question that you just asked. What are people tracking? The kinds of questions people have stay the same more or less, because the questions are really fundamental, basic questions that people have in the course of their everyday life. So it's an easy answer to give you. What are people tracking? They're tracking how much they move, they're tracking where they go, they're tracking how they sleep, they're tracking what they eat. And really importantly, they're tracking their symptoms. If they're managing some kind of condition, it could be an acute condition, like maybe they're recovering from surgery and they're going through rehab. So they're tracking the progress of their rehab and some of the things that become very important depending on what kind of rehab they're doing. Or it could be something that goes on for many, many years because people cope with chronic conditions across decades. So those are the kinds of problems people are working on. It's just that they have really dramatically new techniques with which to work on them.

That's great. And I love the idea that the technology is really there now that you can apply these data techniques to your everyday life. It sounds like it's not exotic things that people are tracking. It really is health, fitness, mental health and all that sort of stuff. So stuff about your day to day life. So I'm curious about who your users are, like who is it that's doing this tracking?

The tracking that's being done is being done by hundreds of millions of people. So the quantified self movement has been catalyst. Sometimes I point out to people when they look at the quantified self and they see who's involved in it and they see that a lot of it is specialized practitioners or people who are involved in technology or they're involved in science, and they have a range of questions, not just about their own health, but also about the practice itself. So they're involved in the quantified self not just because they're interested, for instance, in dealing with chronic pain, but also because they're interested in the techniques of the quantified self itself. Right. So they have both of those interests. That's a small minority of the people who are doing the self tracking. So our mission is not really to be a landlord owning the space of self tracking. And if you're doing self tracking, we better find a way to be involved with you. We just have a different attitude. We feel like it's something that is going to be pretty much ubiquitous. There's going to be a huge range of practices and tools produced by many, many thousands of different kinds of actors in this space, individuals and companies, etc. Our role really is to hold a place for a discussion of what is the cutting edge of practice, what is happening that people who have this general interest in supporting self tracking need to pay attention to. So the people who are involved in our community still tend to be somewhat scientific, technical or allied health professionals who are turned on to the idea and want to know what's happening.

So it sounds like you have some sort of technical users, but a lot of people are just everyday people who are interested in this sort of thing.

Well, we have a forum. Forum.quantified self.com Let me just say that again. We have a forum, community forum that's@forum.quantified self.com and you see two things happening in that forum. You see people passing through quickly to ask a question like does anybody know if I can track sodium or potassium easily at home because I concerned about electrolytes? And that person is going to get a good answer at the forum, but they may never come back again because they had a question and they found us and they asked a question and they got their question answered. And it's a very open forum and we don't make you register and we're not going to be sending you emails for the rest of your life that says check out the quantified self forum for the latest news about potassium tracking. Right? We have no interest in that and it's not our thing. We're there to help somebody who may just be passing through at that moment. But then you also see people in the forum who are the people who are answering those questions. And those people have a special interest in self tracking for various reasons. Right? They're skilled practitioners and the forum serves both of those functions.

And so for people who are interested in this, because you're working with data, do you need any data skills to be involved in getting started tracking something?

It's a good question and I want to give a careful answer because the answer is yes. I mean, the simple answer is yes. You need skills in order to reason effectively about your own empirical observations. You definitely need skills. However, there's a common mistake that is made only by highly skilled people, and that is to assume that the main skills you need are data analysis skills. And so what can happen when people have really good data analysis skills is they simply start collecting data because they're confident that if they have a lot of data, they're going to be able to answer their question. So they pile up lots and lots and lots of data, and then eventually they get around to using their skills, they start to look for patterns in the data. And this is an important mistake to call out, especially given the nature of the podcast that we're talking on, because what tends to happen there is that they don't actually find meaningful patterns in the data. And all of the work that they've done collecting the data is for naught. And the reason that happens is that they haven't really posed a question, an answerable, empirical question that has guided their data collection. So somewhere back in the process, there was probably a problem with a protocol or some uncertainty, or maybe the data they collected wasn't really appropriate for the question that they eventually had. And no amount of specialized analysis in the end may help them at that point. So whenever I encounter somebody who really is pretty skilled with data analysis and wants to get into quantified self tracking, I always recommend that they put their data analysis skills on the shelf at first. They will come in handy, no question about it, but it's good to keep them on the shelf at first and to work on some of the upstream skills so that you end up in a place where you're more likely to find an answer.

That's really fascinating. And I guess it speaks to the difference between a sort of a data mining philosophy where I'll just collect all the data I can and then analyze it and there'll be some sort of pattern somewhere versus traditional statistics philosophy where, okay, I'll design an experiment and then I'll collect the data I need for that experiment and I'll analyze it afterwards.

Right? I'm going to jump in again and say this is another place where people can make a mistake and it's another thing that's quite specialized or special, maybe I should say about quantified self practice. Sometimes people think, okay, I don't want to just do data mining, but I want to have a really solid, really valid result. So instead what I'll focus on is my experiment design. And especially if they have good skills in that area so that they may be a scientist or have scientific training, they have high standards for Experiment design. So the next thing that happens is that they find themselves in a situation where they've got a fairly elaborate and hard to run experiment that we've even seen people who do things like self blinding, right? Like they want to experiment with different medication doses, so they break apart the pills and put them in special capsules and label them in bottles and have a friend or a spouse give them a different dose on randomly selected days. Here you get into these kind of really rigorous experimental designs and those also are often a dead end. Because what often happens in a situation like that is that the answer is quite ambiguous. Maybe the effect size is small and the time that they could commit to collecting the data was somewhat limited, only a few weeks, right? So now they don't have that many data points and they have a small effect size. And so there's high uncertainty and it was really taxing. So they can't really just keep doing it or do it again. And what's happened in that case is that they've sacrificed learning a little in the name of learning something really with high certainty. Whereas it's often much more advisable to have a more basic process that gives you a little bit of a clue and then incrementally work forward through your discovery until you reach a point where you're certain enough. And certain enough in personal science means something very different than certain enough means in professional scientific practice. Because you're trying to solve a practical problem, right? So let's say you're solving a problem of digestion and through some self tracking it seems to you that if you avoid dairy or you avoid nightshades, say your digestion gets better. Well, that's a great result and you have better digestion. There may be some doubt about whether that was really the answer or whether maybe the answer was more subtle than that. And you may not have to resolve those doubts because you're living with better digestion, or you may decide you need to resolve those doubts. So you may want to do another round six months later when your doubt is really refined. That's something that you wouldn't do typically. Well, you would do it as a scientist. This is what's most interesting, is that scientists who've been involved in this practice have given some talks where they say like, you know what, this is really like the back room of science. We do this kind of stuff all the time of like kind of half assed experiments where we could never publish something and it leaves us with a lot of doubts, but it gave us a little bit of a clue that we might be on the right track. Or it gave us a couple new ideas to try as we developed a protocol that was much more solid. Right. But we just. That stuff stays in the back room. Like you don't make any claims really publicly based on it, but it helped you take a step forward in refining your ideas. It's just that in personal science, the backroom science has more value because it's just for you as an individual and you don't have to. Nobody's saying that you have to then go up and stand up at a conference and say, here's my results. And I'm sure that it's true.

I love the idea of backroom science where it's just enough to give you an idea rather than have something that's completely rigorous. And the example you gave earlier about someone blinding themselves so they didn't know which pills they're taking, that actually sounds absolutely terrifying because I guess there's a lot of room for error in that case. So I'd love to talk more about the skills side of things, but can you give a practical example of what a first good tracking project to try is?

I can, and it's really interesting to think about this because it touches on some of the issues we were just discussing. A lot of people are really interested in their mental health, and measurement of mental health in clinical and academic research is actually quite hard. And yet self measurement of your mental states in personal science sometimes is one of the best ways to start self tracking because the phenomena that you're interested in are by definition subjectively noticeable. Right. So you're asking about things like stress or panic or a feeling of having a good level of energy to work, or really low energy where you don't feel like you can work or concentrate. These are things that getting an objective measure that would satisfy a clinical or academic scientist could be really quite hard. But as a subjective measurement that you're satisfied accurately reflects your own questions and experiences. It's really quite easy. And so one of the things that I think is worth considering as a first step is asking questions that can be at least partially answered through a single measurement once a day. So for instance, a question like feeling of restedness upon awakening. Right. Like on a really simple scale, maybe a one to five scale. Right. Or satisfaction with the day right before going to sleep, or if you want to track something more kind of labile, something that changes more, it could be very interesting, is a feeling of sadness or rumination, which is a symptom for a Lot of people of depression, a really explicit feeling of like, boy, I feel very upset by something that I can't stop thinking about. These sorts of self tracking projects really pay off for people for a couple of reasons. One, the act of self tracking itself creates a lot of learning like that, training yourself to notice what it is that you're experiencing. People value that a lot. Second, they're really easy to keep up for the long term because you're not making very many measurements and the measurements are quite reliable, like trustworthy. You can trust them and it allows you to notice when things change, which is an extremely important thing for people who are reflecting on their mental health. And finally, they can be represented analytically as a simple time series graph, which is one of the easiest ways to reason about empirical data. So you just have a magnitude on the Y axis and you have time on the X axis and you draw a line through all the measurements. And simply looking at the graph can give you insight into what's happening. And maybe I should say one more thing. Those are all immediate benefits that you get from doing that kind of tracking. It's like a taste, it turns you on to the practice. However, there's another benefit that kind of comes later, which is that you're well set up to do more complicated kinds of question asking, because now you have an ability to ask a question like, well, what is my typical range? And how often do I see a measurement that is far outside my typical range? Now, I'm very deliberately, as you probably noticed, since you think about data, I'm very deliberately not using statistical concepts here. I could easily have thrown in a statistical concept that is associated with typical range, right? But I've learned in working with people, including with skilled people, that as soon as you throw in a statistical concept, you invite them to do something more complicated. And so I actually think it's more interesting to think in these kind of more accessible general concepts of typical range. But I put it out there as a hint, because if you think about a future in which you've been doing very simple subjective self tracking for three or four or five months and you have a sense of that typical range, it might fire up your imagination for things which are much more complex, including experiments and interventions. But a naive approach is to just say, okay, here's a measurement that's outside my typical range. Let's see if I can figure out what was going on that day. You go back and you can look at your calendar, or you can look at your email, or you can try to find A way to say, oh, I think I know why that measurement was outside my typical range. It was the holiday season or I just traveled overnight and didn't sleep for 24 hours. Right. But as you get more and more skilled, you can ask different kinds of questions, like, well, it looks like I regularly have one or two measurements outside my typical range a month. So that could be just some noise around kind of my lifestyle that everything isn't constant all the time and I'm going to get knocked one way or another by, by the winds of normal life. However, look at this month. In this month I have six or seven measurements outside my typical range or I see a slow increase in the number of measurements outside my typical range over a very long period of time. Right. So here again, I'm just trying to fire up your imagination because I know you can think of statistical tools that would allow you to highlight those sorts of changes in measurements. But even without those tools, you can see that there could be some very interesting discoveries in that kind of data.

Absolutely. And I love that point you made to begin with, that just the act of tracking data forces self reflection and that in turn can lead you to ideas or lead you to make a change in your life. But from what you were saying, it sounds like maybe the best thing you can do is just to take repeated measurements over time and then just drawing a simple line plot, a time series plot, is going to give you a lot of insights. You don't need to do much more to get something valuable after this sort of project.

Well, you asked the question, which I think was a good one, is how do you start, how do you begin? And I gave that answer because it's a really good answer, both for people who do not have a lot of scientific or technical training and for people who do. So both types can get a lot out of doing this sort of long term, extremely low burden, subjectively noticeable tracking. However, there are some really nice things you can do as well as you get up to speed. Because tracking like that is really about refining your question. Right? It's about refining what do I want to track. I'm going to track, I'm going to make one measurement a day. I want to keep it going for a year. Let's really think about what that should be. Of course you can change it because it's, it's not that burdensome. Right. So you can change it. But I call that the foreground measurement. Again, avoiding kind of some of the more technical language about dependent and independent variables or whatever, which it's somewhat analogous to, but I call it the foreground measurement because it's what's in the foreground of your mind in terms of the question that you're asking. Well, there's also, we have access to what I think of as background measurements. So this is your Fitbit, right. This is your geolocation data, this is your photo stream and your calendar. You can also intentionally bring that kind of background data on board for yourself. Right now there's really minimally invasive blood glucose monitors. There are many kinds of things that you can begin to track without having a really hard question to ask them. And you run them in the background. And that's different than what I was talking about at the beginning where people just collect a bunch of data and put it in a pile and then go and look at for a pattern because now they have a specific purpose, a specific reason for being there, which is to help you explain variation in the foreground measurement. And so when you're looking at this time series graph of your foreground measurement and you're saying, wow, I think there's something going on here, some change that I should pay attention to. But what is going on now you have background data that you can consult and that's very interesting.

Absolutely. And one thing that's come up a few times is the idea that everything you do has to be based around a question. And I know on your website you have this sort of four stage process for these tracking projects that starts with a questioning phase. So can you just talk about those different stages?

Yes. So data, the way we think of it is really the formal record of your observations. Everybody makes observations in the course of daily life. But what's special about quantified self is that you actually make a record of those observations. And it's the formality of the record that allows you to use the quantitative techniques down the road. The actual physical act of making the record is it doesn't have to involve a numeral. Obviously some people can highlight days with a color or something like that. Right. And it still counts as an empirical observation because there's a formal mapping of how you make the record to the observation and the measurement that you're making. The important thing is that when you make that measurement, you make it for a reason. It's not a measurement if you don't have a reason. And yet the reason we make our measurements is often really not very well considered. So the reason somebody might make a measurement might be I got a Fitbit for Christmas and it seems like a neat device and therefore I ought to measure how many steps I'm making by wearing my Fitbit. And I never really learn anything from it. And people say this all the time about their devices. I look at it once in a while. It seems like it's basically stays the same a lot. And eventually I ran out of charge and I didn't feel like charging it and I left it in the drawer. Well, the problem there is not that the Fitbit is no good, the Fitbit is perfectly good. It's that the reason you were making the measurements had no real value to you or it only had a minimal value of appreciating the person who had given you the present and wanting to show that you valued the present. So we did that for a while and then the point was made and that's the end. So that's just a dead end. And it's always going to be a dead end. The only thing that makes self measurement and self research not a dead end is the value, the personal value, the emotional worth of the question that is driving the measurement you're making. And finding that question, it takes some time, it takes some thought because not only does it have to be important to you, that's the first thing, but it also has to be what we call tractable. So if you, you might say, well, it's really important to me to ask a question like what is the meaning of life? And I think life should have a purpose. And I'm not sure what my purpose is. And so I need to reflect on my purpose in order to have a better life. It's a kind of crudely simplified train of thought, but it's one that many people have had. But it's not a very good question for a personal science project because it's not very tractable using empirical methods. So we've learned from teaching for over a decade that if you say that to somebody, say students who are coming in to learn personal science as part of their allied health seminar or something like this, their question is, well wait, what does it mean to be tractable? Empirically tractable? That's just a bunch of words, right? It seems like it could almost be a tautology, like a good self tracking question is a good self tracking question. So here's where you find some skill and practice is needed. What makes a question tractable is that you can represent the question through a phenomenon, an experiential phenomenon, like I care about being healthier, but how do I know I'm healthier? Well, let's look at what represents Healthier to you? Oh, I feel healthier when I'm not eating ice cream out of the freezer late at night. Now, you've taken this very general concept of healthier and you've represented it through a very specific phenomenon. And then you can look at that phenomenon and ask, okay, is it a good candidate for self tracking? And then we have some practice kind of helping people define that. Like, it has to be noticeable if it's really hard to notice. For instance, if you have to use a really complicated instrument that's expensive to use. Like you'd say healthier. You could say healthier means that my cholesterol doesn't vary too much hour to hour and day to day. Now, we've done some of those experiments. I mean, I've measured my own cholesterol as often as once an hour. And it does vary a lot. And those cycles are really important and really interesting. But it's a very burdensome measurement to make. Like it involves an actual fingerprint blood test. So, like, that's a phenomenon that you can notice if you really put a lot of effort into it. But its noticeability is really low. Right. It's like you have to do a lot to notice it. Whereas something like eating ice cream out of the refrigerator at night is really easy to notice, you know, when you're doing it and you can make a record and there's lots in between there. Right. So when we teach people to do personal science and we teach them how to ask an empirically tractable question that's related to a subject of high personal concern, that's the kind of work we do. We talk about their concern. We talk about what phenomenon in their life can represent that concern. Then we talk about, well, wait, is that a really suitable candidate for self tracking? If it's not, sometimes you can find another way to represent. I'll give you an interesting example of that. One of my colleagues, co authors of the book that I've been working on, he was tracking headaches. And he began by noticing when he had a headache and how strong it was and how long it lasted. But it actually turned out to be really hard to notice those things because there's a lot of ambiguity there. Like, when does the headache end and how do you notice when the headache is over? And how do you know that your number four strength headache today is the same as your number four strength headache last week? Right. Like, you could have drift in how you're measuring it. And all these little problems, they add up to lack of trust. And I Mean, lack of trust you have in your own measurements. And if you start to distrust your own measurements, your motivation to make them goes way down, right? Because you're doing something that you're suspecting is BS and you're trying to believe in it, but you know that it's not working, and eventually you just stop doing it. So he switched the phenomenon that he was tracking from headaches to number of aspirin he was taking. Now, that's not a perfect representation, right? I mean, there's many ways that that could be slightly off here or there, but it was pretty good. A day in which he took no aspirin was pretty much a day in which he didn't have a headache. That was bad enough that made him want to take aspirin. And a day on which he did take aspirin was a day in which his headache was bad enough that it compelled him to take aspirin. And now we've got something. Now we've got an empirically tractable question that relates to a topic of high personal concern. This fiddling, this is core to the scientific process. Every scientific discovery that's ever been made probably has a really high amount of this empirical fiddling that goes on in the back room to figure out what exactly can we measure here that's going to cast light on our problem? But it's the part that's really hard to teach, right? Like, you can teach somebody how to use R, and you can teach somebody how to calculate a certain result or to transform their data in a certain way, but how do you teach them to ask the right question such that the answer they get is going to be really meaningful? And so that's where we focus.

That's a really interesting story. And I like the idea that it might take a bit of playing around to find just the right metric to get something that you can quantify from there. I'm curious to know if your friend managed to make any improvements with his headaches after doing all this tracking. Like, did it have an impact on his life?

He did. Very interesting discovery and very classic for personal science, I think, because it's a discovery about something that is known but not well known. And so if you don't know something, then it's not known to you, and therefore you can't use that information, right? So he found that by tracking how many aspirins he took, he was slightly more resistant to taking aspirins. So when he had a headache, he would ask himself, do I really need an aspirin for this headache. So he was taking less aspirin, and the number of days without aspirin was going up, but it was actually going up a lot, like he was doing a lot better. And he also learned from reading the fine print in a very long insert that it was not advisable to take more than a certain number of aspirin per week. And one of the side effects of taking too many aspirins is headaches.

Ah, I'm starting to see where this was going. So taking aspirin to solve the headaches was actually causing more headaches, then was.

Causing more frequent headaches. And so by. And you could see the results really clearly in the data. By reducing the number of aspirins he was taking, he was not only changing the measurement of how many headaches he was having, but he was actually changing the frequency of the headaches.

That's absolutely fascinating. And that's a pretty cool success story from actually just tracking what was going on, he managed to find a solution to his headache problem.

Well, it's also interesting in the sense that if you were to go to a scientific conference and try to present on this big discovery that taking too many aspirins can lead to more headaches, people would just shrug their shoulders and they'd be like, yes, well, we know that well enough that we print it on the insert in every bottle of aspirin. So how does that count as a discovery? But it actually counts as a discovery because he didn't know it. Many people don't know it, but he specifically didn't know it, and now he knows it.

That's brilliant. Do you have any other success stories from your community of people who've changed their lives?

Well, we've archived thousands of projects, so there's some that are fun to talk about because they're so clear, but they're also a little bit misleading because sometimes it sounds like, well, quantified self is just an intervention. You take step one, step two, step three, and your problems are solved. Which is actually not typical of what happens. What happens is you have a very gradual increase in your problem solving capacity relating to your own health, in your awareness of the things that influence it, and your capacity to manage the shifting tides of your own life and personal health. So I love to tell the stories, but I hate to tell them at the same time because they become too pat. Whereas really, what you want to encourage people to do is to turn on to that feeling of empowerment that you get from being able to solve your own problems. But with all of that said, we've seen some wonderful talks about incredible discoveries for people personally. For instance, a talk that I've never forgotten is by a person who had adult onset acne, which was really frustrating condition to suddenly develop. And through tracking symptoms and diet was able to find out that actually it was only Irish dairy that was causing her breakouts. She had eaten dairy her whole life and never had any problem. But it turns out that dairy in Ireland is typically produced with a different kind of cow than the dairy she was typically eating in the United States before she went to Ireland. And that there's an enzyme in the Irish dairy that she was specifically having this kind of autoimmune reaction to. And when she stopped, her acne went away.

That's. That's just an incredible story because it's such a niche thing and it's not something you would immediately go, okay, well, I've got some spots on my face. It must be the new dairy I'm eating. It must take a certain amount of experimentation to try and figure out what's going on.

Yes, well, this is very important because you can go to the bookstore and you will find dozens of books that tell you how to deal with chronic conditions by changing your diet. They'll tell you to eliminate carbs. They'll tell you to eliminate meat. They'll tell you to eliminate certain foods that have minerals in them. There's a million ideas out there about how we should change what we eat to change how we feel, and probably all of them are true for somebody. But the question is, what works for you? And it's not practical to say, I'm going to try every combination of every food. You have to learn how to do things a little bit more or a lot more powerfully and with some skill and use your inherent human kind of rational problem solving capacity. And the very first step in that is to have an outcome measure, to have a foreground measure. If you have a measure that tells you, how am I doing, Just that alone gives you a lot more reasoning capacity to find out, how are the things that I'm changing affecting me? And in diet, we see that all the time. Like, you really need to have a measure of something. You have to represent your questions about your health or about your diet with a phenomenon that you can track, such that you know whether things are getting better or worse or the pattern of them occurring is changing. And then you can use a lot of other techniques. They don't have to be formal experiments. I mean, one very common way to handle severe dietary issues is to do an elimination diet, right, where you eat only a couple of very simple foods for a while and then one by one you add things back in and you try to find the things that are affecting you. It's an extreme solution and it's one in which there actually ends up being a lot of noise in the reasoning process. And for the most part people have solved their problems with a much more rough hewn approach to reasoning, but one that is based on a really solid foreground measure.

So the idea is that if you experience symptoms, you can work backwards and see, well, what did I eat yesterday? And then than a candidate for working out what to eliminate rather than getting rid of everything at once.

Correct. And that process of tracking your symptoms, what you want is to build trust in your foreground measure so you really have a feeling that you know whether you're getting better or worse. And once you build a lot of trust in your foreground measure, then you can often use more rough, more intuitive processes for developing ideas about what the cause is. Now you can always test those ideas further, right. And we could start to talk more technically and say, well, it's a kind of Bayesian process where you have a, you have an articulated prior belief and now you can add certainty to the subsequent changes in that belief. Right. So it's all up to you. I'm not saying like you shouldn't try to be sure, but I am saying that this process of articulating what you think and how you know what you think is the secret recipe to making meaningful discoveries.

That's great. It does sound like there is potential there. I mean, you said it doesn't happen every time, but it does sound like there's potential of having some pretty life changing interventions through these projects. So I'd like to talk a little bit more about some of the practicalities since. Well, since you mentioned Bayesian statistics, are there any particular sort of tools or techniques that a lot of people make use of when they're doing these tracking projects?

Well, the kind of lingua franca of self tracking is the time series graph. So you're going to almost always see some sort of time series in self measurement. And then there's a set of really simple analytical steps that people may take with their time series data, which if you added those in, would give you 90% of the tools that are used to make discoveries. For instance, they might compare different windows of time where those windows are really chosen carefully. So yes, of course, it could be just month by month. Let's look at all the months. Or it could be seasons. Let's compare these three months. But often the windows could be something like, let's take a look. A certain number of days around the highest measurement magnitude. Right. So here was a peak. Or it doesn't have to be days. Right. It could be number of measurements. Right. So how long does it take to go up? And then how long does it take to go down again? Right. So the shape of the curve there could be very interesting because sometimes it doesn't matter if you have a high measurement, as long as it gets gradually high and gradually low again. Right. As opposed to peaking all of a sudden or dipping all of a sudden. Sometimes it's simply what is the lowest measurement of all time? Or what is the highest measurement of all time. This sometimes is a really important measure. Sometimes it's a condition, like in the measurements by location. So let's find out what locations really matter. If I care a lot about high measurements, do I have a lot of high measurements at this location? Right. You can see how analytically, anybody who can do basic analytical operations using tabular data can perform most of these operations. There's nothing very advanced about it except that there's a lot of judgment, like, for instance, making your bins or defining what a location is. These are really situated judgments where it's very consequential how you treat that right. So you may say, I'm going to bin them by work and home, but it turns out that actually your car was really important because your car was really dusty and like, you had a bunch of allergy attacks at work, but it's because you were driving your car to work. Right? So even though the analytical techniques are basic, the reasoning skill, you watch people get better and better at it. And there's room for improvement from all of us in terms of how we think empirically about the data we're collecting.

That point you mentioned about having to make judgment calls is something we. We talk about a lot on this podcast because it always seems, oh, yeah, I want to be a data analyst and just write some Python code or something and. And crunch some numbers. But actually, there's always the problem that you have to understand the context of the data. And really, you end up with a lot of edge cases where you're trying to make decisions in order to get a sensible answer. But I love this idea that you can get started by doing something very simple and just drawing a plot of how your numbers change over time, and you can gradually introduce more sophisticated techniques and get into things like calculating descriptive statistics and doing time series analysis and things like that. So there are different levels of analysis for different people who want to try this, for sure.

And we've seen very sophisticated things where, for instance, people who are tracking mental performance using some really widely available online testing tools want to take into account learning effects. Right? So these are tests that you probably get better at over time. So that's going to conceal some of the variation that may come from some of the things that you're really concerned about. And you can account for that if you're careful and skilled. Right. So we've seen people transform their data and do all kinds of really neat stuff in reasoning, but you don't have to. And sometimes like if you're highly skilled and you want to do this kind of tracking and you're prepared to do it, you can use one of these tests that has learning effects and you can simply account for those learning effects. Or if that is too complicated for you or you don't trust it, like you might even know how to do it, but you might be, you know what, I think that these learning effects, they themselves might vary in some way that I'm not taking into account. Or the different tests may have. The learning I do on one test may bleed over into learning on another test. And I didn't start them both at the same time. So these are complicated and challenging calculations that they're too hard for me to make. But I do know people who would feel confident in making them. My approach would be to refine my measurement and to say, okay, I better use a measurement that's a little less vulnerable to being doubted by myself so that I don't have to take the risk of doubting my analytical process. So again, that in itself is a judgment call. Right? And I just was talking with somebody and I think they posted about this on our Quantified Self Wiki, or well, the Personal Science Wiki Run out of the Open Humans Foundation. Some of your listeners might be familiar with the Oura ring. And the Oura ring is a self tracking ring that keeps track of a bunch of biometrics and produces a readiness score meant to reflect your general well being of the day to come. He recalculated his own readiness score because he felt that by taking his own approach to using heart rate and sleep time, he would have more trust in that data. And he did that by looking back over all of these measurements and his record of whether he was sick or healthy. That's an example of a highly skilled person who develops doubt in a ready made score that Aura is offering and is able to just refactor that for himself such that he counts on it more. And that sort of thing can be really rewarding because now he has a measurement that he trusts, whereas before he had a measurement that he didn't trust.

And it just seems trust is all important if you're going to do these things. Otherwise there's absolutely no point in tracking anything if you don't believe the results.

Distrust creeps in all the time and it is the number one thing that you have to pay attention to because sometimes it creeps in without your even noticing it. And then you're doing a bunch of things that you actually don't believe in and you only realize you don't believe in them after you've invested a lot of time and effort. So I think that's also a great lesson from all of our experiences. Be aware of whether you really have confidence in what you're doing.

Absolutely. It just seemed like for people wanting to try this, it would be helpful if they can get advice from other people who've done similar projects. So are there any communities for tracking that people should join?

I definitely recommend coming right to the Quantified Self forum forum.quantified self.com it's very low key forum, it's very on topic and it's not super high traffic. It's a place where you can ask a question and people who are lurking there and reading those questions will probably give you a hand.

That's brilliant. And in terms of sharing your results, so I know a lot of these projects are personal and okay, maybe you've solved a health problem or you've got fitter somehow and it's just for you. But I guess some people want to share what they've done with others. So where can people share the results of their project?

It's very much our job to help people learn from other people who have been making discoveries about themselves. So if you are interested in sharing your project, all you have to do is tag us Quantified Self on Twitter. Drop a note in the forum, send me an email@garyantified self.com any way you get in touch with us. We're interested in helping you share your project.

And I know we have a lot of listeners to the podcast who are interested in getting a job in data. And it does seem like a lot of these data tracking projects, the results would make a great addition to a data portfolio. So I'm wondering, have you heard about any instances where people have done this project and then they've shown it off and it's helped them get a job for sure.

And not only have people used their self tracking projects to identify themselves as somebody who knows what they're doing with data, but there actually is a specific domain within data science which is growing extremely rapidly, which is supporting discoveries by the customers of all of these self tracking tools. So all of those tools, the Fitbits and the Apple watches and the health apps, they all are providing people with insights based on analysis of the empirical observations that the devices are doing. And those insights are all being supported by data scientists, either behind the scenes or in the front of the stage, kind of making knowledge out of data. So I think if you're interested in quantified self and in personal science and in self research, that in itself is a domain that you can work in.

That's brilliant. Just being able to use your projects to make something better about your life and get a job, it's a sort of a double win there. That's fantastic. To finish up, do you have any final advice for anyone wanting to try a tracking project?

Well, just to enjoy yourself, I think, and to not spend a lot of time doing things in the hope of getting a reward months or years later when you've piled up a lot of data, but to do things that will feed back knowledge to you within days or weeks. If you do that, you can get into a cycle that leads to a lot of learning very quickly.

Thank you for your time, Gary. That was just a really fascinating session. So thank you for all your insights.

Thank you very much.

You've been listening to Data Framed, a podcast by Data Camp. Keep connected with us by subscribing to the show in your favorite podcast player. Please give us a rating, leave a comment and share episodes you love. That helps us keep delivering insights into all things data. Thanks for listening. Until next time.