Tellient Blog

Thoughts On Things

IoT Analytics

IoT Analytics vs Other Analytics: 4 Key Differences

Do you know which vendor can actually provide analytics for your particular IoT deployment? The data analytics industry is huge, and there are many subcategories. The most familiar to most people are web analytics and mobile app analytics, but there are companies serving practically every vertical where there is measurable and actionable data. Financial services, social media, server infrastructure and shop floor automation all have nuances that require a highly specialized approach to analyzing data.

These differences aren’t slight, like, say, the difference between a chihuahua and a labradoodle. A dog analytics provider could still tell you things like how much they eat, how fast they run and how much they sleep because they are both dogs even though one is small and annoying and the other is majestic and awesome. No, the differences between these environments is more like the difference between a chihuahua and a 747, where there is zero commonality. So is the relationship between IoT and the WWW. The tool required to measure each requires a different approach entirely. Here are some key differences:

  1. The devices are distributed vs. centralized. A website is a single destination that potentially millions of people can visit. This creates a funnel to a chokepoint through which all users flow. It is at that single chokepoint where behavior and usage can be measured. Web analytics, while not necessarily easy to do well, at least has the luxury of measuring one thing, a website. IoT analytics, on the other hand, has to measure millions of individual users of millions of things. This means instrumenting the code of millions of individual devices to report to the mothership. It is an order of magnitude more difficult to do, which is why Google Analytics doesn’t work for IoT.
  2. The data is heterogenous vs. homogenous. When you measure your website, you are measuring the same thing every other website owner is measuring. Pageviews, time on site, visitor flows and conversion rates are all homogenous indicators of website performance. A commerce website might be very different from a blog, but the measurement is the same because the there are only two things you can do on the web: Click on an ad or buy something. IoT analytics differs in that the variety of devices is greater, and how they are used is different from each other. They may or may not even be interacted with by humans. One size does not fit all. A good IoT analytics solution must be built from the ground up with this in mind.
  3. The purpose of the analysis is fundamentally different. There is no A/B testing of interfaces for devices without interfaces. While server (or web or app) analytics in general is a powerful tool to identify, say, code improvements or to analyze marketing campaign effectiveness, these concepts have little to do with IoT analytics. Companies use IoT analytics to act as a nervous system for products that previously could not feel. A smart, connected water heater (for instance) has no screen that users are clicking on, so the approach to measuring its usage has to be fundamentally different, but more importantly, the resultant data is put to much different use. MTBF, performance anomalies, service intervals and consumable maintenance are all concepts foreign to, say, social media monitoring or mobile app optimization. The goal of the analysis informs the underpinnings of the software itself, and IoT analytics is optimized for the very specific goals of devices and sensors.
  4. The algorithms are optimized for machine data.This may be the most obvious point of all, but is also the most important one. A really good financial data analytics tool would have a learning algorithm that helped you optimize your investing strategy based on mountains of financial market data. That algorithm would be completely useless in helping you A/B test the color of a button on your website. Similarly, when it comes to optimizing software to something really well, focus is helpful. This is the reason Google Analytics isn’t used for mobile applications and why Mixpanel isn’t used for web applications and why server analysis or application performance management software is useless for IoT applications.


IoT analytics is, somewhat unfortunately, a buzzword. Google the term and you’ll see how many companies without an actual IoT analytics solution are claiming they have one. These companies mistakenly think that their server monitoring software or their manufacturing automation software can be tortured into an application it wasn’t built for.

Similarly, the term “IoT analytics” could mean almost anything at this point. It may refer to shop floor automation measurement, sensor data aggregation, smart city monitoring, utility reporting or device analytics, among other definitions. It is important to understand that a tool built for monitoring traffic patterns in urban environments is not an appropriate tool for measuring usage of your company’s smart consumer lightbulbs. Sifting through the noise may add some overhead, but the likelihood that you can find what you need is very high if you are careful.

Download the Free Whitepaper: The Network Operator's Guide to Monetizing IoT Data

About The Author

Shawn Conahan

Shawn Conahan is the founder of Tellient. His mission is to make smart things smarter. (Just ask his modded Roomba named Robbie with adaptive mapping and navigation.) Shawn also loves infographics, and his all-time favorite is the Carte figurative des pertes successives en hommes de l'Armée Française dans la campagne de Russie 1812-1813 on his office wall.