Solving the Internet of Things’ Last-Mile Problem

By Vaidy Krishnan 2015/11/09

With the rise of low-cost sensors, connectivity everywhere, and our fast-growing volume of data, the Internet of Things is likely to reshape the world as we know it. The possibilities are immense, but so are the challenges. Making the IoT work for the masses is more of a data challenge than a problem of things. We need to extract the data from devices then figure out what it all means.

So far, the market has been focused on getting smart devices online. We’ve seen little innovation to help us consume all the data that the gadgets and machines collect. As a result, many IoT solutions suffer from the last-mile problem. In other words, these solutions fail to help people see and understand the data they mine. But what good is data you can’t use? And if you can’t use it, why go through the trouble of collecting it?

So how do we democratize IoT data, be it from a smart home appliance, a wearable, or an industry-scale solution like GE’s Predix Platform? We need to address three hurdles that stand in the way.

1. Enable Deeper Exploration with Interactivity

We humans are innately curious. The answer to a question often leads to more questions. This happens when we analyze data. The more we see, the more we want to learn. That’s why we need a flexible tool that lets us sculpt and mold our data in different ways as our needs change.

Unfortunately, most IoT applications ship with one-size-fits-all views that lead to dead-end dashboards. They answer a predetermined set of questions; we can’t go beyond what we’ve been given.

For example, we might have an IoT application that looks at the historical activity data of a broken engine and predicts what conditions led to failures, and how often a failure is likely. But what if we want to drill down on the parts that fail the most often? Maybe we want to see which factories manufactured these parts and when, or which suppliers caused the most issues. What then?

In the rare case that we can ask follow-up questions, we’ll likely have to port our data or engage in days-long—or even weeks-long—development cycles. And since questions, by their very nature, involve exploring the unknown, it can be difficult to justify the costs for a big IT project. But if we don’t explore, we also won't know what we could have learned, or what opportunities we missed.

So how do we get meaning out of IoT data without having to fund a huge IT project? The answer lies in interactivity. When we can interact with our data, we can have a conversation with our data. We can explore all sorts of permutations and even discover surprising patterns.

Mac Bryla made some surprising discoveries when he explored the phone metadata of Australian journalist Will Ockenden. In the spirit of discovery, Will procured and released his own data to the public, and was rather surprised by what Mac learned in just 10 minutes! The viz below illustrates how Mac peeled back the onion, layer by layer. This is an amazing showcase of what’s possible when we are empowered to ask and answer our own questions. It’s a far cry from static, closed-ended views that limit discussions before they have a chance to take shape.

2. Allow Broader Analysis with Data Blending

Interconnected devices have changed daily life in ways we couldn’t have imagined even a few years ago. And deep within their data, they contain stories that have yet to be told. Uncovering these stories often involves combining IoT data with additional context.

Kaj Peltonen wanted to explore his Fitbit data for a possible link between his exercise regimen and sleep patterns. The native dashboards in Fitbit only allowed him to him analyze his fitness data in isolation. But he wanted to look at his FitBit data in a broader context. He wanted to know:

• How does physical activity during the day impact my sleeping patterns?
• Do I perform better when I have had ample sleep?

Thankfully, Fitbit allows users to export not only data on physical activities but also data on food intake, body measurements, and sleeping patterns. (An export is not ideal, but it’s sometimes the only way to broaden the scope of analysis.)

In no time, Kaj blended the data in Tableau and matched up his sleep patterns with his daily exercise regimen:

Kaj learned that a good night’s sleep (particularly on Monday nights) is often followed by an active day.

Imagine uncovering enterprise-level insights this way, by blending disparate data. Sensors embedded in a jet engine can help us predict when it might need service. The data can help us preempt failures and save billions of dollars. But what if we want to know how those savings compare to our projected budget by product and region? The ability to blend disparate data sources on the fly can help answer those questions.

3. Foster Sharing & Collaboration

Last but not the least, it’s imperative that we can easily share our IoT data and insights with others in a meaningful way. Now that we’ve gone through the trouble of collecting and analyzing the data, we want to maximize its impact.

Sharing allows others to explore new vectors and have their own aha moments. And as our IoT data starts to help drive decisions, things go from being mere devices that carry signals to being agents of empowerment that transform organizations. That’s how we close the last-mile gap.

We’ve only scratched the surface. As devices light up, things get interesting—if we can decode what the devices have to say. Once we see and understand IoT data, we’ll be able to make smarter decisions about our personal well-being, our professional performance, and the world we share. '

For more thoughts on working with IoT data, check out our whitepaper.


Submitted by patternproject (not verified) on

A wonderful read, esp the way it brings together IoT and Big Data, and how one can leverage the strength of the other.

Submitted by Barbara Campbell (not verified) on

The collaboration statement is so true. Job roles between work locations are lacking the time to pass the baton of information effectively. Results of hast seems to be more rework, duplicate work, defects and cancellations. Repeated errors are difficult to remedy when the we cannot view where, when, by who and on what job the event occured. The money spent on rework and removals could be better spent with knowlege of where the issue originally happened than fixing defects over and over and over again with little hope of improvement. Question: How can we get a Tableau Account?

Submitted by Vaidy K. on

Thanks for the feedback Barbara! You can create an account here: Among other things, this will give you quick and easy access to training, videos, events etc. and let you participate in the community as well...

새 댓글 추가 

non-humans click here