SAG_Twitter_MEME_Working_Backward_May19-2(Or: What IoT can learn from Steve Jobs)

The complexity of the Internet of Things is staggering. Bringing greater insight into your business, by learning how operations deep at the heart of organization are really performing, requires a level of analytical proficiency that (until recently) was only found in either very large or very specialized organizations.

In order to notch the analytics up to the next level in your organization, you will need to double up on your efforts. All IoT use cases - and especially those in Industrial IoT - are doomed if the organization fails to get to grips with the increased complexity.

Take, for example, predictive maintenance. This is a use case that is on everyone’s minds. How many different assets do you have in your organization? Do you have historical data available from those assets? Do you know the normal behavior of those assets? How will you solve the challenges of identifying abnormal behavior?

But before you get bogged down in the details it is good to realize that there is a simple approach to addressing those challenges. It is called working backward.

When you were a kid you probably figured out that it was easier to solve a maze by starting at the destination and then working your way backward, right? Today you will see that it is easier to solve complex issues by tracking back from a desired outcome. If this sounds like good detective work, it is!

With predictive maintenance, the first question you must ask is “What outcome is it that any organization would normally like to achieve?”

You probably have the answer in mind, but it never hurts to make sure all your colleagues are on the same page. The answer might be as simple as preventing a machine or asset from failing by detecting early signs of problems - and taking an appropriate action.

Now the tougher questions need to be addressed. What kind of analytical model could give you such an insight? The data scientist in your team might tell you to set up a model to calculate remaining lifetime. The data scientist probably asked you to join a meeting with maintenance guys and together you would have a conversation on how - for example - increased vibration, temperature and energy consumption are good signs for the particular asset you have in mind.

“Fantastic you think, are we there yet?”  Well, not completely, you now want the engineers in your team to figure out how that behavior can be measured. The engineers might conclude that vibration can be measured if they had sensors able to measure acceleration and velocity.

You are now ready to trace your journey back from the sensor to the original goal of being able to do predictive maintenance. Where acceleration and velocity sensors help you to collect vibration data, the data scientist can then analyze and fit this data into a remaining-life model. This will help you to validate the assumptions of the maintenance guys - allowing you to do the early warnings that are expected in case of predictive maintenance.

The good part of this practical approach that it works in many situations - not just IoT - and many leaders have advocated it. So when you preach working backward to your team, you will be comfortable in saying: “Guys this is common knowledge, even Steve Jobs did it.”

This article was originally published in Big Data Quarterly.

Cumulocity IoT Free Trial

Leave a comment

Search B2B

Most Popular Blog Posts

Subscribe

Connect with us!