How do we get value from IoT analytics? Do we need to hire someone to handle this?

I hear these questions all the time.

Our customers are aware of the promise and potential of getting more information about their physical products, processes, and facilities. But the savvier ones also realize that IoT analytics themselves aren’t the answer.

It’s how you understand that data and use it to drive action that really counts. That’s the hard part.

We can use connected sensors to automatically gather readings, but that information won’t automatically lead to improvements.

How Do You Get Insights Out of IoT Analytics?

The answer is often to use other data that you already have or already understand. Then, by comparing these sets of data, an organization can find what they’re looking for—and also things they never expected.

IoT analytics can’t exist in a vacuum. They need context for the information to be valuable. And while you can get very fancy with statistical analyses, calculating R², and feeding information into Machine Learning models (all good things!), the first step can be much more straightforward.

Just ask: What else do we know or have access to that is relevant to the IoT analytics we’re collecting? Then start looking at your data with that in mind.

Need some examples? Of course you do!

Measuring Elevator Usage

One of our customers is using our Kosmos IoT System to learn more about how elevators are being used in their building complex. It’s a commercial building with over 30 elevators, including freight and passenger elevators. Most of the tenants are manufacturing firms.

The property’s head of operations previously had no data about elevator usage. He’s hoping to use this information to help with planning maintenance, placing new tenants properly to avoid congestion, and generally to get a better sense of what’s going on at the massive site his team manages.

So how does he makes sense of the IoT analytics that Kosmos provides?

Well, what other source of information does he have access to that may be relevant?

The answer is really simple: a calendar.

The elevator data shows when the elevators move from one floor to the other. By comparing this data against a calendar, the operations team can start to see how elevator usage differs across nights & days, weekdays & weekends, holidays, and seasons.

And more insights can be found when they look at the IoT analytics in light of other events on their own and their tenants’ calendars. What does elevator usage look like on days when major shipments are received or sent out? How do public events affect elevator congestion? What changes when tenants move in or move out?

The connected sensor data that’s tracking the elevators isn’t very useful until it’s analyzed in the specific context of this particular building site.

How’s My Indoor Air Quality?

Recently we’ve been running some experiments on indoor air quality, specifically CO2 concentrations. There’s a growing body of evidence that stuffy indoor settings impair cognitive performance and that, yes, long meetings in closed conference rooms may be making you dumber.

Across New York City we have volunteers who have set up CO2 sensors in their offices with the data being collected by our Kosmos IoT System.

Understanding CO2 measurements is simply based on levels. Higher readings (roughly above 1,200 parts per million) are worse, and lower readings are better. But we need other information to determine how to address high CO2 levels.

As with the elevator example above, calendar information is useful. CO2 levels are expected to be higher the more people are present in a space.

But information about the offices themselves is necessary. Does this space have windows? If yes, what kind are they and are they ever open? Are the doors open or closed? How high are the ceilings? What sort of HVAC system is being used?

Without this information you can’t know how to fix high CO2 levels. It could be something as something as leaving a door open or installing a fan. You might need to modify your ventilation system or office layout. Or maybe there’s simply too many people packed into too tight a space.

The main point is that even when IoT analytics clearly indicate that there is a problem, you can’t understand how to fix the problem until you look at the other, contextual data.

And while this may sound obvious, that’s the point. You start with the most obvious inferences and insights about the data at hand, and from there your analysis grows more sophisticated and detailed as more data is brought into the picture.

Improving Stormwater Management

One of my favorite projects we’re working on right now involves mitigating stormwater’s impact on the Gowanus Canal in Brooklyn. The waterway is a formerly industrial neighborhood that now has a fast-growing resident population.

In order to reduce flooding and combined sewer overflows into the canal, we’re working with a local organization to study how trees can best absorb stormwater during rain events and mitigate its impact.

To do this we’ve placed soil moisture sensors by trees around the neighborhood, and Kosmos is collecting that data automatically for several months across different seasons.

But the soil moisture data itself won’t answer any questions. We need to place it in context.

One source of contextual information is the NYC Street Tree Map. This tells us information about each tree that we’re monitoring, like what species it is and how large it is. This information is useful for determining which types of trees best absorb rainwater.

Another source of contextual information is local weather data. There’s a weather station in the vicinity of the neighborhood, so we’re using that to learn the cumulative rainfall for every rain event and its intensity or speed.

This rainfall data helps us distinguish between a neighbor watering a tree and a true precipitation event. It also helps us determine how much water each tree can absorb, how quickly it can do it, and for how long it can retain rainwater.

All this information helps our customer determine the best tree planting policies to advocate for in their neighborhood. It also helps them acquire new grant funds to pursue these policies since they can back them up with data.

Where to Find Contextual Data

So where do you look for the contextual data that will help make sense of your IoT analytics?

Well, the possibilities are endless and depend on the problems or issues you’re interested in. But I’ve deliberately kept my examples in this post fairly straightforward. Simple is usually the best place to start.

All sensor data is going to have timestamps, so thinking about calendars is a great place to start. How could the time of day, day of the week, or the season shed light on what the IoT analytics mean? What events could be affecting what you’re monitoring?

Weather data is also generally easy to get and is going to play a role in many situations, particularly with anything outside.

You should also just think about any data that you already have and is easy to get. Examples from my experience include sales data, worker schedules, accident rates, raw material costs, and more.

IoT Analytics in Context: What Next?

Once you’ve started looking at sensor data in context and understanding it more, it’s really important to share your insights and how you arrived at them.

This is important for two reasons. First, sharing your insights will help bring about the changes or improvements they’re pointing to. And secondly, sharing your thought process will spur others to bring their own ideas to the sensor data collected. It often causes colleagues to realize that there is other data they’re familiar with that should be analyzed against the IoT analytics.

If you have a problem you’re working on where connected sensors could help, please get in touch. We’re always interested in tackling interesting challenges and helping our customers think through how to get lessons from their data.