Analytics for IoT – “Let me tell you the Facts of IoT Life.”

You knew that the Internet of Things was going to involve lots of sensors, lots of data and lots of new opportunities for understanding your business and customers. But did you think you’d need to learn about the Facts of IoT Life?

Research by Enterprise Management Associates (EMA) confirms that there is no more waiting for the Internet of Things. The sectors leading the way in IoT are industry, utilities, health care and manufacturing. The research also shows that most companies are connecting devices for geolocation, production, building infrastructure, smart homes, fleet vehicles and consumer durable goods. The fire hose of data is working just fine, thank you.

And that’s where the Facts of IoT Life come in.

Analytics for IoT data – Not the same old stuff

Analytics for IoT data is not the same as analytics for traditional data. Sometimes it’s a square peg in a round hole. As the photo shows, you can make that work, but it takes some doing.

For example, you don’t simply open a spreadsheet or run a couple of SQL queries, then sit down and analyze your IoT data. Some of the approaches and workflows for working with high-dimensional, high-volume data are easily adapted from existing applications, but not all. And some industries are discovering the differences between analyzing IoT data and traditional data the hard way.

To acquaint you with some of the Facts of IoT Life, I’ve put together a new eBook called Analytic Strategies for the Internet of Things – Getting the Most Out of IoT Data. In this post I’ll tell you a few of those Facts of IoT Life.

Where do I put all these data? In the data historian.

Most data sets generated by IoT systems and applications consist of time-stamped data, usually for large numbers of parameters and events. That’s one of the first big differences you’ll discover between ordinary data and IoT data.

Making data available for analytic software and routines is a job for the data historian, which is a standard database or data warehouse where data are indexed in several ways:

  • Time (What is the time stamp?)
  • Parameter tags (What is being measured?)
  • Data quality tags (Is the measured data point reliable and within valid bounds?)

The data historian also records events such as alarms (what occurred and time of occurrence) and failures (nature and time of failure). Not only is the nature of the data you’re collecting different, but you also apply different criteria in deciding how much of it to store.

Those data are not interesting…Oops, maybe they are.

The biggest Fact of IoT Life is that interesting patterns in IoT data look different from those you’re accustomed to spotting in, say, financial or manufacturing data. Until you know how to use IoT data in useful models, they won’t seem interesting and you won’t get much value out of them.

Your analytic strategies and goals with IoT data are similar to those with traditional data, but certain considerations are different because of the time-series nature of streaming, time-stamped IoT data. Our paper explores those considerations in five areas:

  1. Aggregating and aligning data
  2. Determining aggregation intervals
  3. Forecasting and predicting in the time domain
  4. Modeling trajectories and multivariate anomaly detection
  5. Finding inflection points

Don’t take the data to the math. Take the math to the data.

The next Fact of IoT Life is that those sensors, devices, machines and appliances have the potential to generate huge data sets. Do you want to have to constantly reel them into your data center or desktop before you can run analytics on them? That’s a lot of collecting and a lot of network traffic. Then you have to put them all somewhere. Storage may be cheap, but how long do you think you can keep track of where you’ve put data sets that have come in from all over the world?

In a Native Distributed Analytics architecture (NDAA), you can run analytics on aggregates computed in a database, gateway or device/sensor, wherever it may reside. With NDAA, there is no need to fetch data from the source; the math goes to the data set, wherever it is. Payload and latency are much smaller, and the role of the desktop is limited to building models and displaying results, rather than crunching vast data streams.

Citizen Data Scientists, right in the line of business

“Not every company can afford a data scientist,” says Shawn Rogers, chief research officer of Statistica, “which is a big reason why Citizen Data Scientists will become a big part of the data ecosystem as it evolves.”

That Fact of IoT Life has brought businesses to the realization that they can’t hire enough data scientists to do everything that is possible. Instead, they can train Citizen Data Scientists to do everything that is necessary. Line-of-business users who can meet their own analytics needs are less dependent on statisticians and able to accomplish more.

New eBook: Analytic Strategies for the Internet of Things

You’ll find more insights and elaboration on IoT data analytics in our new white paper, Analytic Strategies for the Internet of Things – Getting the Most Out of IoT Data. It will help you wrap your head around a few of the Facts of IoT Life that may not have occurred to you yet.

They may not make you giggle as much as the Facts of Life did when you were a kid, but after a certain age, you’ll probably find them more useful.

Download the eBook