The 5 Types of Sensor Data Used by Businesses & Organizations

Making Sense of Sensor Data | Live Sensor Data | Historical Sensor Data | Analytical Sensor Data | Predictive Sensor Data | Sensor Data for Change | Which Type of Sensor Data is Right for You?

Text message alert from Kosmos

His eyes light up when he sees the freezer’s temperature appear on his phone. No more worrying about what is happening at his plant while he is away. Now this COO can walk into work and already know what is happening—no more unwelcome surprises.

The light bulb goes off when she exports all the hourly temperature and humidity readings from her storage facilities. Looking at the spreadsheet in front of her, this production manager realizes she’s just created her compliance report with the press of a button.

There’s so much talk about the potential and possibilities that IoT sensor data will unlock. It’s still striking every time I see first-hand that moment when a customer connects the dots and understands how sensor data will help them.

How to Make Sense of All That’s Sensed

Sensor data visualized in a city
Sensor data generates all kinds of opportunities and ideas.

Cisco says that 5 quintillion bytes of IoT data are generated every day and that 30 billion connected devices are expected by 2020.

These numbers seem impressive, but they’re not really helpful for any organization or team that’s exploring how connected sensors can help their work. In fact, these figures often just scare people off or make them hesitate.

If IoT involves such massive amounts of data, then it’s bound to be too complex and too much work for my organization to implement. Let’s see how these technologies shakes out for everyone else, and then we’ll see what works.

This wait-and-see approach can seem sensible, but I think it comes from a place of misunderstanding.

There are many different ways data generated from Internet of Things can be used. I always find it helpful to talk through the various uses of sensor data with anyone planning to collect it.

Let’s go through the 5 types of sensor data being used by organizations with examples from my work at Temboo. This will help you understand how sensor data can help your own work and how best to start scoping out your next IoT project.

Live Sensor Data: Just in Time

Kosmos dashboard on an iPad
Viewing the latest sensor readings of multiple devices on a Kosmos dashboard.

Live data is the most straightforward way to use sensor data.

You can see the current status of what you’re monitoring anytime from anywhere. And often there’s no need to even check on your devices because email or SMS alerts will notify you if sensor readings breach any thresholds you’ve set.

Recently, I helped a food production facility in the Bronx set up devices for “just in time” data.

Their 40,000 square foot building contains two large storage areas, one that refrigerates and another that freezes thousands of products every day. Ensuring these storage areas are the right temperature is critical to the business.

Thanks to the temperature and humidity sensors we installed and set up with Temboo’s Kosmos System, the COO and production managers no longer have to worry about their compressors failing without anyone knowing. They can check the status of their storage areas anytime. They’ll even get text and email alerts whenever temperatures start rising too high.

Variance Design has also been using Temboo for years to collect live data. They create and manage living plant walls for business and institutions, including Amazon, the Smithsonian, and the American Museum of Natural History in New York.

Their sensors measure light, temperature, and water pressure, triggering alerts whenever anything is going wrong.

These alerts are critical for ensuring the living walls that are located across the country stay alive. For example, if the pressure in their irrigation systems drops, the plants will die from lack of water soon, requiring the entire wall to be replaced.

In short: Live data is simple to understand. Every IoT system will provide this. It’s best for making sure systems are working and getting notified when they’re not behaving as they should.

Historical Sensor Data: Records & Compliance

Battery power monitoring on Kosmos
Viewing historical sensor readings graphed in Kosmos.

Tracking historical data with sensors is also straightforward but involves a bit more thinking.

What will historical data be used for? Where will it be stored? How far back should the data go and what time intervals (minutes, hours, days) should be logged?

The food production facility I mentioned in the previous section first implemented connected sensors for live data. They really cared about knowing when their refrigerator or freezer rooms were failing.

But then they realized that sensors automatically logging temperature and humidity every hour would help them with safety compliance. They had to keep records of these readings anyway and had been doing it manually for years.

Every hour an employee would have to enter the storage areas (opening the doors and causing the temperature to rise) to read the thermometer inside and then write down the reading and time on a clipboard.

These handwritten logs then had to be transcribed and formatted for record-keeping and compliance purposes. This important but rote task is prone to human-error.

But using wireless sensors means that this job is now being taken care of in the background 24/7. Whenever reports or records need to be generated, a spreadsheet with all the temperature values can be downloaded in seconds.

Skysmart has used Temboo for compliance data as well. An MRO (Maintenance, Repair, and Operations) business in the aviation industry, they’ve used sensors to log environmental conditions of their airplane parts storage facility, automatically creating records for their FAA audits.

In short: Historical data is great for record-keeping and compliance purposes. Most IoT systems will generate historical data, but you’ll need to decide for how long and at what intervals.

Analytical Sensor Data: Learning & Assessing

Co-workers looking at data

Generating sensor data for analytical purposes requires scoping out the problem you’re investigating and deciding what data can help address it.

Recently we worked with an obesity and diabetes research lab at the National Institute of Health to collect proximity sensor readings.

The lab is tracking activity levels in mice, and they are measuring this by looking at thousands of proximity sensor readings and their timestamps. By analyzing this data the scientists can then determine the activity levels of the mice.

They achieved this by setting up the sensors to record readings whenever a change in proximity was detected. Taking readings at regular time intervals wouldn’t make sense for what they wanted to analyze.

Stiga, Europe’s largest lawnmower manufacturer formerly known as Global Garden Products, and Electrolux, an appliance maker, have both used connected sensors and Temboo to generate analytical data for new product development.

It works perfectly!

-Varna Vallone, Engineer, Stiga

For Stiga, this involved installing sensors inside lawnmower engines to track RPMs, sound decibel levels, internal temperature, and more. Their engineers then analyzed all this data to help design quieter engines for new products.

The team at Electrolux added humidity and air quality sensors to their humidifiers and air conditioners. The sensor data collected was then used to plan new features for these products and develop connected ‘smart’ versions of them.

While analyzing IoT data generally requires human effort, Machine Learning and other techniques can help alleviate the burden. For example, Temboo’s Kosmos can use ML analysis of sensor data to identify anomalous readings from existing data sets.

In short: Analytical data requires some planning and thinking about what you’re investigating. More data points and more sensors are often needed to make analysis possible. And someone will need to work with the data to get results.

Predictive Sensor Data: Forecasting & Planning

Old fashioned Zoltar game
Sensor data can help generate better predictions.

Using sensor data for predictions is the next step up.

It requires analytical data to have been generated and then understood so that reasonable forecasting is even possible.

We’re currently working on a predictive maintenance project for a major US city.

The project involves forecasting behavior and maintenance issues on multiple water pumps and elevators at a 4 million square foot, 100+ year old building complex.

After collecting, analyzing, and understanding several months’ worth of data, the challenge becomes generating useful predictions and insights from this data. This involves understanding how the operations team currently works with the equipment being monitored.

Being able to predict when a water pump will fail or when a particular elevator will be highly used is undoubtably helpful. But how specific and confident do the predictions need to be?

The answer depends.

Since the water pumps are such critical infrastructure, it’s better to err on the side of caution. So if there’s any significant chance of failure in the coming weeks or months, that needs to be identified so that a maintenance checkup is scheduled beforehand.

On the other hand the elevators have more redundancy.

Each elevator bank has at least two elevators, which means any single elevator being at capacity is more of an inconvenience than a major issue. Predictions would be more useful when they have a high degree of confidence (to minimize distracting false alarms) or if they foresee that all the elevators at a particular bank are predicted to be at capacity.

Knowing what types of predictions will be useful and actionable helps determine the extent of the sensor data that needs to be collected.

In short: Planning for predictive data requires everything you need to do for analytical data plus figuring out what kinds of predictions will be useful. This will require understanding not only the assets being monitored but also how people interact with and maintain those assets.

Data for Change: Building Consensus & Making Improvements

Illustration of Kosmos
Using technology and data to improve businesses, organizations, and the world.

Do you want your company’s processes to run better? Do you want your organization to achieve its goals?

Obviously, the answer is yes. And using sensor data and IoT technologies can be of help.

The key to using sensor data to effect these sort of changes is understanding the human element. That’s always much more complicated than figuring out the technological element.

We’ve been working with municipalities and NGOs in the water space to help them push big changes forward. They may be using sensors to track methane extraction at wastewater treatment plants, monitor stormwater absorption in bioswales, or measure the volume of water going into the sewer system from pumps.

Similar to how predictive data requires understanding how people interact with what’s being monitored, using data for change necessitates understanding how people plan, budget, and manage what’s being monitored.

So in the water space for example, using data for change means you need to know what information is needed to motivate certain decision-makers.

You also need to understand how data needs to be delivered to your audience in the most legible way. And you need to know how new initiatives get launched and budgeted.

The human and organizational element can’t be ignored when pursuing data for change.

In short: Using data for change requires deep understanding of the people and organizations that will participate in the change. Learning about the internal politics and constituencies involved requires partnering with others. This is the most difficult but also the most promising use of sensor data.

What’s Your Goal & What Do You Need

Youth playing soccer
What goals do you want to achieve?

So what are you hoping to achieve with sensor data? And what do you need to do to make it happen?

Here’s quick guide for thinking about which of these five uses of sensor data you want to implement:

  • Live Data: I want to know when something is not working.
  • Historical Data: I want to keep logs of when something has and has not been working.
  • Analytical Data: I want to understand why something isn’t working.
  • Predictive Data: I want to know when something will stop working.
  • Data for Change: I want to change how something works.

Which of the above statements most applies to your situation? If you’re not sure, start with live data and then see where that takes you.

It’s always best to start simple, especially if your company is new to sensor data and IoT technologies.

The ultimate goal for every organization is to use the data they have to become better and more effective. And using data for change at scale is where IoT is headed.

But don’t underestimate the change and improvements that can come from even the most straightforward applications of sensor data collection. There’s always a first step. And getting a sense of where you can go and might go in the future always helps.

If you’re ready to start collecting any type of sensor data at your organization, contact us or sign up for a free trial of our Kosmos IoT System.