When the boss comes and asks you if you're ready to do something with the Internet of Things thing she or he read about in an airline magazine, prepare to give them a very, very long list of things you'll need to do in order to get ready for the magical new world of measuring everything everywhere all the time.
Next, prepare to weave a dense web of suppliers to get anything done.
And then tell your boss it won't work the first time around anyway.
That was the basic message of a talk delivered yesterday by Gartner's Chet Geschickter during the Australian incarnation of the firm's 2015 Symposium.
Geschickter pointed out that the choices you'll have to make are many and varied, starting with:
- Should you build your own things?
- Or buy things off the shelf?
- Or rent things that thing service providers put in place?
- Will your things need power? Or will piezo-electric things that generate their own power by tuning into ambient vibrations do the job?
Once you've though that through, consider:
- How will your things connect to your business?
- Are you ready to handle diverse network connections?
- Are you ready to have lots of things connect?
- What happens if the network goes down and things stop talking?
- Is that data really coming from the thing you think it's coming from?
- If a thing had been compromised, how would you know?
If you can answer all of those questions satisfactorily, more challenges lie ahead, namely:
- What standards and ontologies will you use to describe thing-generated data?
- How are you going to analyse all the data things make? On-premises or in the cloud? With Big Data tools or with other analytics tools?
- Who can you buy this stuff from?
- Does anyone who sells this stuff have a suite or will you need to integrate?
- Will the vendors running in this space allow integration?
Next: how to act on insights generated by the combination of things and analytics. Are you willing to automate actions based on insights derived from big data? Or will a human need to be in the loop? How can you manage the risks thing-derived insights might bring?
Plenty of those questions won't be easy to answer, but Geschickter said one thing he's confident in predicting is that you'll need dedicated IT resources and infrastructure to manage any IoT implementation.
He recommended basing that effort on a network operations centre, only one tuned to handle high volumes of data coming over diverse connections.
And he also said prepare to fail and that pilot projects are essential.
“You will not get this right the first time,” he said. “There are too many variables.”
“This not a one and done thing,” he added. Repeated experimentation will be needed to learn what data matters and what data produces useful results.
Some of the questions above need to be asked before any project. Others will require novel thinking and acquisition of new knowledge. More than once, if Geschickter's correct. Which means you get to have more than one meeting in which you hear about the great ideas that emerge at the pointy end of planes. ®