Manny Bernabe • 2021-11-11
The Internet of Things is coming. It will transform your business as surely as the internet did, and your business must plan for it. Treasure maps exist for a reason, though, and for you to deliver the treasure of IoT to your organization takes careful navigation.
There are three general paths this journey can take. Unfortunately, only one leads to success.
As stark as this may seem, when it comes to industry-disrupting technologies (internet, mobile-first, AI, IoT, etc.) there are more ways to lose than to win.
Let’s look at some of the blunders common to teams getting started with IoT analytics initiatives. I’m sorry to say that most of these come from first-hand experience, but glad to help you avoid some of the potholes on the path to success.
Have you noticed how Apple advertises its deep learning neural algorithm that powers Siri (Apple’s digital assistant)? Right. Me neither. The customer does not care about the technology, only what the technology will do for them. Good executives act as conduits for the customer, and want to see how your technology will ultimately benefit the customer and therefore the business.
As you advance past the PoC stage in your projects, start to pivot hard toward business value. If you don’t, you’ll find yourself lacking internal support and failing to get your project to production.
IoT analytics is a team sport. You’ll need executives, managers, engineers, and operators working together. When you bring the team together, everyone should have the same goal in mind. To align, your team will need a common vocabulary and understanding of IoT analytics. You will find that IoT analytics is a new enough space that in a room of 10 colleagues, you’re likely to get 10 different definitions of it.
As a first step, provide context for IoT analytics. What is it? Why is it important now? Do this for all your team members. Thereafter, you’ll want to host sessions tailored to specific roles (executives, managers, engineers, operators, etc.). This will help provide the right amount of detail for your audience.
The most important input to an AI solution is domain expertise. This expertise is scattered across your organization. The operator who’s been handling the machine for 20 years… the customer service rep who knows the most common customer complaints and questions… the marketing manager who knows what competitors are rolling out. Bringing these folks together is a top priority. Create experiences where these different views come together to learn and align on the value of your IoT analytics initiative.
There are two problems here. Firstly, you don’t know what type of data science team you need until you properly identify your challenge. Under the hood, you’ll find several subspecialties in data science teams, such as time-series and computer-vision experts. In rushing to build your data science team, you may find a mismatch between their skills and the organization’s needs.
Secondly, waiting for a complete data science team will slow you down. Data scientists are hard to find, hard to train, and hard to maintain. Particularly early on in your journey, don’t let this be an unnecessary bottleneck.
Let me know if you’ve heard this one before: Innovation leader spends nine months developing cutting-edge, deep-learning, neural AI proof-of-concept… that completely flops.
Writing data science code can be a prolonged process. Solutions that look easy on the surface will turn into a sinking pit of tangled vines and thorns. Before you spend serious time (and dollars) on production-ready code, be 100% sure that you are solving the right problem.
Map out the customer value earlier. Confirm your assumptions. Talk directly to the customer. The earlier the better.
Data is to IoT analytics projects as fuel is to rocket ships: you won’t get far without it. Have data analysts explore the breadth and quality of your data from an analytics perspective. Your current data is likely coming from existing business systems (accounting, customer services, quality control, etc.). These systems were not designed to optimize analytics, so your data will be lacking in meaningful ways for the purpose of IoT analytics.
Lack of quality data is the top reason IoT analytics projects fail. It’s not uncommon to have great business ideas and a valid statistical approach, only to realize that there is insufficient data to fully realize the opportunity.
It’s rarely too early to test your idea. It’s rarely too early to get feedback from key stakeholders and customers. Even early in the development cycle, you can find ways to prototype and test the most important aspects of your IoT analytics opportunity. Otherwise, you might find yourself having spent a lot of time and resources developing the wrong solution.
You won’t be able to avoid all of these issues. More success will come with experience and more “at bats.” Keep in mind that IoT analytics is a new muscle that is being built. With each IoT analytics initiative you undertake, you’ll get better at spotting these issues and avoiding them.
If you’d like to dig deeper into any of these issues or have questions about your specific IoT analytics project, feel free to email us at [email protected]!