Fitness, data, analytics and impossible dreams

Tracking ProgressI don’t want much. I just want to train, get fitter and stronger, and live a healthy life for as long as I can. To that end, I exercise regularly, I’m fairly careful without being anal about what I eat and I look for ways to be more active like walking to the shops instead of driving.

But I’m also a professional nerd and I like to use technology to help me with my goals. And, over time, I have tried to consolidate my data in order to overcome the explosion of different data sources that fitness and exercise enthusiasts have faced over the last few years.

But I still can’t use that data to figure out the difference between the days when I’m working at my best and those where my performance is sub-optimal on the road or at work.

I’m a Fitbit user but that’s not really important for this as I think the problem I’m describing is not a platform one but an intelligence one.

Looking at the data

Let’s take three easy-to-capture metrics: sleep, exercise performance and diet.

Most fitness trackers are able to capture data about your sleep. For example, you can easily track the time you fell asleep, the time you woke up, and sleep quality. Sure, different devices will measure this in different ways but the data passes the “good enough” test. Realistically, the only way most people could get better sleep quality data is to have cameras and other monitoring equipment on hand.

jawbone-up-appExercise frequency and performance are also easy to capture. A decade ago, taking data from a sports watch to a computer for analysis was difficult and only available in the most expensive devices. Today, you can get that data for under $200 on your smartphone.

Nutritional data is the most difficult data to capture. While there are some very comprehensive nutritional databases around (MyFitnessPal has the best in my view) you still need to weigh foods. Also, none of the apps I’ve seen let you enter what time you ate the meal. Most simply allow for breakfast, lunch and dinner, and snacks.

Of course, those three categories – sleep, exercise performance, and diet – each has a number of more granular dimensions. And it’s the ability to look at those and find correlations that deliver value.

Data – know thyself

Back in the late 1990s I worked at a manufacturing company on a project, building an enterprise data warehouse. This was simply a database that collected data from a number of data sources within the company. We then used some tools to conduct analysis on that data.

This was important to the business as it allowed us to find correlations and other relationships between bits of data that had been siloed from each other.

But perhaps the biggest lesson I learned from that work was that if I knew and understood the data, I could understand the business. After all, data doesn’t lie. Our interpretations might be erroneous but data that is collected using accurate tools never lies.

Today, a group of dedicated (obsessed?) people is doing similar things with their lives. The Quantified Self, or QS, movement is all about quantifying as much as possible in order to optimise our lives.

the-plateau-effectHugh Thompson and Bob Sullivan wrote a QS meeting they attended in their book The Plateau Effect.

“One programmer/presenter showed how he logged the pace every mile he had run for months and mapped them to air temperature, precise time of day, liquid intake and dozens of other factors. He had determined that he runs best at ten fifteen in the morning”.

I’m not suggesting we record every single possible data point you can but a lot of the data we could use is already available to us.

For example, as a runner I’m interested in the time, duration and distances I run, what I’ve eaten, and my sleep. All of these are captured without too much trouble. But when I look back at two runs on the same course a year apart and my performance is significantly different, how do I know why they’re different?

If I have access to the data, then I can start to better understand differences between good and bad days.

I was unwell the other day

A couple of weeks ago, after a heavy fortnight of work and travel I came down with a bad cold.

That can happen. The combination of a couple of red-eye flights resulting in two lost nights of sleep along with a busy schedule had me a little run down. Interestingly, when I looked back at my heart rate data for the week, my resting heart rate was slightly elevated.

Most nights my resting pulse is around 52-54 beats per minute. But when I was unwell it was up at 56 beats per minute. That’s not a big difference but with many months of heart rate data recorded, that subtle two beat difference seems significant to me.

I had a measure that seems to correlate with my health quite strongly.

What I was missing was an alert that warned me I was not operating at my best and that a rest day was in order. Instead, because I didn’t get a warning, I went for a nice run and probably contributed to further weakening my system.

Wearable devices and health data are immature

img_0540.jpgWe are still very much in the infancy of how wearable devices can help people improve their health and fitness. Excluding elite athletes with their teams of biomechanists, doctors and specialists, most of us can only use off-the-shelf gear like fitness trackers, smart scales, body composition measurement devices like the Skulpt Chisel and low-cost devices like blood pressure cuffs and pulse oximeters if you really want to get fancy.

But even with all that what we’re missing are tools that let us take all that data, bring it together and analyse it.

Then I could answer a couple of questions.

  1. Why didn’t I perform as well as expected on a specific training run or race?
  2. How can I predict when my bad days are coming so I can modify my training program or diet?

For this to work, we need the companies creating our fitness trackers to stop locking data in proprietary databases.

One of the reasons I stopped using a Nike Sportwatch was exactly that reason – the data couldn’t be easily accessed from other applications.

myfitnesspalSome platforms are quite open. Strava talks with almost every major running tool out there. Fitbit is reasonably open and MyFitnessPal is probably the most open in both receiving and sending data to other applications.

All this works through APIs (Application Program Interfaces – tools used by software developers to send data between different programs). However, there’s a dearth of tools for taking data from different sources for actual analysis. The tools that are out there are mainly about single ports of call for storing data.

What I really want is someone who can grab all the detailed information and overlay a reporting tool that lets me search for my own correlations and data relationships. I want to be able to find what foods work best for me before a run, how much sleep is optimal for me and I want to predict when I need to lighten my training load.

I want to be able to use the data I collect. Is that too much to ask?

Leave a Reply