Connected Footwear Testing: Speed, Distance, Stride Length and Cadence
The Under Armour® mission is simple: Under Armour Makes You Better. Digitally, this means delivering reliable, accurate, and trustworthy data to our customers. In February 2018, Under Armour released its third version of connected footwear: the UA HOVR™ shoe. Not only do these shoes provide customers with a fantastic pairing experience and integration with MapMyRun®, they also provide incredibly accurate run data in the form of speed, distance, stride length and cadence.
Delivering accurate run data is not a trivial job. It requires years of research, massive amounts of data collection, and rapid iteration. This is how we do connected footwear testing at Under Armour.
Test, Analyze, Update, Repeat
Before starting connected footwear testing, we define the requirements for the product. For UA HOVR connected footwear, those requirements include:
- Functional for both running and walking speeds
- Inclusive of a wide variety of runners, running styles, and physical characteristics
- Sensitive enough to react to small changes in running
- Highly accurate data (speed, distance, stride length, and cadence)
Based on these requirements, we begin data collection. We design scalable and reproducible testing protocols to stress and improve the algorithms responsible for providing speed, distance, stride length, and cadence.
To ensure connected footwear works for all potential users, we need to test the product with a large and diverse group of people. Tall people, not-so-tall people, fast people, not-so-fast people, people that love to run, people that hate to run, people with immaculate strides, people with not-so-immaculate strides, etc. By testing against a population that represents a wide consumer set, we build confidence that our product will work for almost everyone. This concept is simple but, the execution is challenging.
After each round of data collection, we analyze it and use it to evaluate and improve the algorithms that calculate speed, distance, stride length and cadence. Then, after each update, we go back and iterate on our process, collecting more data and testing the new and improved algorithm. This process of constant iteration and improvement is always ongoing, and the accuracy of Under Armour connected footwear is continuously improving, even after the product launches.
After a strong candidate algorithm is identified or “locked in,” we characterize how this algorithm performs in a myriad of conditions and how it compares to other sources that produce similar data. This usually means we need to understand how connected footwear performs against GPS on its relevant metrics.
To start, we test how our algorithm performs under a variety of conditions and use cases. These conditions include interval runs, downhill runs, uphill runs, and trail runs. This is important pressure testing because it helps the team identify hard-to-find bugs in the algorithm. And, if needed, this affords the team the opportunity to “tweak” the algorithm that calculates speed, distance, stride length and cadence. The end result is a net positive for the connected footwear owner who will invariably run with their shoes on hills, trails, and in a variety of training conditions.
Next, we work to understand how our algorithm compares to other distance- and speed-tracking data sources; namely GPS. Currently, GPS is the de facto benchmark for tracking speed and distance. However, as many people have experienced, GPS often loses accuracy near tall buildings, densely wooded areas, and in bad weather. Therefore, while GPS is an accepted benchmark, it isn’t a pure gold-standard. The following figure displays the relationship between Under Armour connected footwear, GPS systems, and the truth.
Essentially, both connected footwear and GPS have some inherent level of error when compared with the absolute truth. However, that level of error isn’t always directly comparable. For example, I may run 1 mile (measured with a certified wheel) in my connected footwear and with a GPS tracking device. The former may read 1.02 miles while the latter reads 0.98 miles. Both devices registered a 2% error in this scenario, but one may overestimate while the other underestimates causing a 4% difference (not error). Got it? Good!
After extensive testing, we are happy to report that connected footwear compares favorably to GPS in terms of pure accuracy (absolute error compared to the truth). Moreover, when a runner begins a run in densely wooded areas or near tall buildings, connected footwear significantly outperforms GPS. So what’s the takeaway message? GPS and connected footwear are both viable options in terms of determining run-related metrics, but if you want a device that consistently performs regardless of the environment, with a very low overall amount of error, connected footwear is the superior option.
From Lab to Market
After years of testing and iteration, the product is ready to go to market. By emphasizing testing in real-world settings, testing against a plethora of runners at a wide range of speeds, and tirelessly iterating, we are confident we are delivering a product that generates speed, distance, cadence, and stride length data that our running customers can trust. This, in turn, helps ensure our runners have an amazing running experience.