PARTNERS Blog

Apr 10, 2017 Five Essential Analytics for IoT Applications

Trends / Future / Data Conference / Technology


Practices, Processes, and Technologies for Better Decisions


The following is based on notes and slides from a presentation by Mike Gualtieri, Principal Analyst Forrester, “Five Essential Analytics for IoT Applications,” at the Teradata PARTNERS Conference 2016 in Atlanta, GA.

In the data and analytics world, the Internet of Things (IoT) is more popular than the Kardashians. That’s just a guess, but with “82% of enterprises stating that they are interested in IoT” according to Mike Gualtieri, Principal Analyst Forrester, it would be hard to argue with that statement.

The interest in IoT spans across industries and around the world, and for good reason: Internet sensors, devices, and applications are being incorporated into everything imaginable from jet engines to refrigerators to smart dust. As the number of Internet-enabled devices grows, the data associated with them compounds as it follows suit. The result is an ever-growing volume of data, with more and more coming quickly.

 

Data is Like a Raindrop

The issue for enterprises is that while all data is born fast, most analytics are done later. As Mr. Gualtieri noted, “data is like a raindrop.” It is initiated far away and travels a great distance before it has any impact, creating ripples in a puddle for instance. Enterprises must act on a range of perishable insights quickly in order to get value from the data. The time-to-action requirement determines how quickly data must be analyzed. If your goal is real-time, then your data is highly perishable. That’s where IoT applications can excel. Where devices have sensors and may have controllers, IoT applications are not smart without the “brain” that comes through the analytics. This, generally speaking, allows them to operate very quickly while being broadly applied and dispersed.

In Gualtieri’s view, an IoT application must be capable of the following actions:

  • Learn about the device
  • Detect context in real-time
  • Decide to act or not in real-time
  • Adapt logic over time to improve application value

 

If You Can Measure It, You Can Use It

While most applications are blind and not designed with the idea of learned logic, with sensors you can make them see. With that, Gualtieri identifies five types of essential analytics for IoT applications.

1. Streaming Analytics – Streaming analytics can detect and act on real-time perishable data. Its use is increasing dramatically—from 24% in 2014 to 42% in 2015. Think of it as continuous ingestion with continuous analytics. It can provide the ability to filter, aggregate, enrich and analyze a high throughput of data from disparate live data sources. From there it can identify patterns, detect urgent situations, and automate immediate results. A great example is the ability to warn drivers of a dangerous curve based one car’s experience. When a car’s ABS or traction-control is activated, the system could look at current road conditions, compare historical data about that location, and warn following drivers of the danger ahead.

2. Machine Learning – Machine learning builds learned logic predictive models from historical data. This is typically done in batch mode and expressed in different ways. Algorithms analyze data to find models that can predict outcomes or understand context with significant accuracy and then improve as more data becomes available. The presentation identified three types of machine learning models:

  1. Classifiers – predict characteristics
  2. Recommenders – suggest the next best action
  3. Cluster – identify patterns in the data

3. Spatial Analytics – Spatial analytics detect and identify activity at a specific location. According to Gualtieri, “location is critical for context in IoT apps.” It includes the tools, techniques and technologies to understand the fixed and changing spatial relationship among physical objects. Common sources of spatial information include the familiar GPS, IP address, RFID, and cellular data.

4. Time-Series Analytics – All data has a historical timestamp. By batch processing time-series data, trends can be identified. Tools, techniques and technologies can be applied to analyze time-ordered data and find statistical patterns and/or forecast future values in time.

5. Prescriptive Analytics – The power of analytics lies in decision-making. Prescriptive analytics aim to improve and accelerate decisions to act—or not! Ultimately, success depends on the collective efficacy of decisions about customers, operations, and strategy. According to Gualtieri, “If the ability to rapidly model, improve, and change decisions is critical, it will be infused in the application.”

 

More Data, Better Decisions

While many IoT apps existed before the term Internet of Things, since the dramatic growth of applications, devices and the subsequent data, many have been reclassified by their creators as “apps.” In the end, however, Gualtieri used this definition: “IoT analytics are the practices, processes, and technologies that use data, analytics, math, experiments and human insight to improve the efficacy of decisions made by human and/or by decision logic embedded in applications.” One thing for sure is that the growth of IoT isn’t going to slow anytime soon. Knowing and applying the proper analytics will enable organizations to acquire the most useful information to their objectives.


Visit the PARTNERS archives to view archived presentations.

PARTNERS ARCHIVES

Get Updates

You have been subscribed.
Thank you!

Please enter the fields below and click subscribe to receive email updates on PARTNERS.