• United States

3 types of IoT platform analytics

Nov 26, 20184 mins
AnalyticsInternet of Things

Enterprises rely on 3 types of IoT analytics from their selected IoT platforms. By testing multiple IoT platforms, an enterprise can identify best-in-class analytics capabilities.

brain team iot analytics
Credit: Getty Images

Enterprises rely on their IoT platforms for many services. One of the most important is analytics. In layman’s terms, IoT analytics is the science and art of trying to find patterns in the massive quantity of data generated by connected assets. Or a more careful definition from MachNation’s IoT platform testing lab might be, analytics is the ability of a platform administrator or operator to monitor trends, identify abnormalities, and produce business insights from ingested IoT data.

As a first step to identifying a best-in-class IoT platform for analytics, an enterprise should deploy the various platform analytics services. In particular, the enterprise should at least configure an on-platform analytics service for live streaming and stored/historical data; configure a platform for live streaming external analytics service integrations; and then export on-platform data to an external analytics service. These configuration tests will help an enterprise determine if an IoT platform vendor has designed exceptional or lackluster management tools and usability into its platform.

After an enterprise tests the analytics configuration processes, it should also evaluate 3 types of analytics capabilities. Let’s take a look at the 3 types of IoT analytics and how an enterprise can identify a platform that provides best-in-class analytics microservices.

1. Descriptive analytics

Descriptive analytics is the most basic form of analytic insight that allows users to describe and aggregate incoming IoT data. Descriptive analytics — even calculations as simple as mean and standard deviation — can be used to quickly make sense of collected data. In a connected factory use case, description analytics might be used to answer the question, “What are the average pump temperature, flow rate, and RPM over a 30-minute time period?”

When identifying best-in-class descriptive analytics capabilities on an IoT platform, enterprises should evaluate:

  • On-platform descriptive analytics capabilities: The ability of a platform to perform descriptive analytic inquiries, such as aggregating or calculating basic statistics of ingested data points across sensors, device, or groups of devices as well as visually presenting the results.
  • On-platform data lake / big-data storage capabilities: The ability of the platform to both store and query against very large quantities of ingested IoT data including table-based data stores with greater than 10 million rows or unstructured data stores with greater than 50 million records.

2. Predictive analytics

Predictive analytics seek to model future data and behaviors by analyzing historical data. Regression analysis such as linear regression is an example of predictive analytics. In the same use case, predictive analytics might be used to answer the question, “What is the estimated time-to-failure for a pump that is demonstrating a 20% increase in measured temperature?”

When identifying best-in-class predictive analytics capabilities on an IoT platform, enterprises should evaluate:

  • On-platform predictive analytic model building: The ability of the platform to automatically or through programmatic-interfaces generate a predictive model of the underlying platform-ingested IoT data. Models such as linear or polynomial regressions are typical, although more complex modeling choices are available in sophisticated platforms.
  • On-platform predictive analytic model operation: The ability of the platform to utilize either a platform-generated or platform-integrated data model (such as R or Python) to classify data or identify outliers through anomaly detection. Users should place emphasis on the ability to manage models such as model versioning and updating as well as the ability to integrate a predictive model within a complex event processing (CEP) framework.

3. Prescriptive analytics

Prescriptive analytics are analyses to help enterprises optimize a future direction to be taken. Image processing, machine learning, and natural language processing are some of the techniques used to complete prescriptive analytics. Prescriptive analytics might be used to answer the question, “To maximize pump uptime and minimize service intervals, what is the maximum allowed temperature increase for a pump before a preventative pump servicing must be scheduled?”

When identifying best-in-class prescriptive analytics capabilities on an IoT platform, enterprises should evaluate:

  • On-platform prescriptive analytic model capabilities: The ability of the platform to utilize either a platform-generated or platform-integrated data model, such as R or Python, to optimize a business outcome or relevant KPI. A prescriptive model should maximize or minimize a business-relevant KPI, such as time-to-delivery in route planning or equipment uptime for predictive maintenance.

Analytics help enterprises create business value by better understanding data. While there is no perfect IoT platform, some platforms are built with higher quality IoT analytics microservices than others. Savvy enterprises will test IoT platforms’ descriptive, predictive, and prescriptive analytics capabilities as well as the ability of a platform to integrate with third-party analytic solutions. They will also fully test their ability to use platform tools to configure on-platform analytics services and export data to external systems. The best way to understand the capabilities of an IoT platform is to use them.


Steven Hilton is a co-founder and President at MachNation, the leading insight services firm researching Internet of Things (IoT) middleware and platforms. His primary areas of expertise include competitive positioning, marketing media development, cloud services, small and medium businesses and sales channels.

Steve has served on Cisco’s IoT World Forum Steering Committee where he was co-chairperson of the Service Provide and Security working groups. Steve has 25 years’ experience in technology and communications marketing.

Prior to founding MachNation, he built and ran the IoT/M2M and Enterprise practice areas at Analysys Mason. He has also held senior positions at Yankee Group, Lucent Technologies, TDS (Telephone and Data Systems) and Cambridge Strategic Management Group.

Steve is a frequent speaker at industry and client events, and publishes articles and blogs in several respected trade journals. He holds a degree in economics from the University of Chicago and a Master’s degree in marketing from Northwestern University’s Kellogg School of Management.

The opinions expressed in this blog are those of Steven Hilton and do not necessarily represent those of IDG Communications, Inc., its parent, subsidiary or affiliated companies.