“You’ll never find your limits unless you’ve gone too far.” — Aron Ralston, Between a Rock and a Hard Place
The grand vision of industrial IoT is trillions of dollars in economic value and hundreds of billions in needed investments. That prospect alone is enough to make both the end user industry segments and the technology providers alike salivate at the prospect of what the IoT revolution will unleash. With tech vendors pushing hard to enhance IoT adoption and end users all too willing to indulge on the tech provider’s dime, nothing seems quite as right as an IoT proof of concept (POC) to begin with. The end game thus becomes navigating past the righteous POC phase before entering the territory of IoT at production scale. The IoT POC becomes the gateway which we need to mandatorily transcend on our way to IoT heaven.
There however remains a small glitch! The POC purgatory is holding us captive. It’s akin to crossing a chasm that can’t be breached. The POC approach is emblematic of the “look before you leap” psyche that typically pervades any big ticket technology initiative. It’s an eerie reminder from three decades back at the onset of the ERP revolution, where the initial approach was to begin just with financial accounting functionality (despite all ERP talk about integration capabilities!). It was a while before the course correction was applied by implementing all core modules together. The floodgates eventually opened and we then saw the mass adoption of ERP.
A survey by McKinsey revealed that 71% of IIoT initiatives are stuck in POC purgatory (41% in limbo at pilot stage and another 30% in startup phase). Another study by Cisco indicated that 60% of all IoT initiatives stall at the POC stage and that only 26% of companies have an IoT project that they classify as a complete success. While there are definitely signs of some green shoots, those scalable yet limited IoT initiatives are clearly not enough for the trillions in economic value to be realized over the next decade.
Here’s why isolated IoT POCs are a bad idea and key takeaways to avoid the same.
Missing the forest for the trees: Often, the way the POCs are structured is that they work hard to prove that the technology works. What this eventually leads to is that the plumbing works just fine, but the mains remain off. Focusing the POC on purely the technology aspects without a business metrics yardstick is a sure short recipe for disaster. It does get the juices flowing of the technology folks, but the initiative doesn’t get the management buy-in owing to a lack of an associated business case justifying the ROI.
Takeaway: Conduct a detailed diagnostic assessment to pick the right business use case which addresses key business metrics and delivers the right business impact.
Taking a scatter-gun approach: Very often, IoT initiatives get initiated by the diktat of an enthusiastic but largely hands-off management team. The folks entrusted with driving the initiative missing any leadership guidance end up spawning multiple POCs (often with different vendors!). Quality and focus are sacrificed at the altar of quantity and giving everyone a shot at the freebie POC.
Takeaway: Ensure a holistic approach with a concerted focus (you can’t possibly take four different tech stacks to the production phase). Pick complementary partners wisely after a thorough evaluation and co-create with them.
Inhibiting the law of accelerating returns: Ray Kurzweil’s law of accelerating returns states that while we humans are linear by nature, technology is exponential. Constraining an IoT initiative within the bounds of a narrowly defined POC translates to holding it captive, clipping its wings and not allowing it to take flight — which goes against the law of accelerating returns. The end result is that the end outcome has limited overall appeal since it focuses on a very tiny dimension, works off a limited sliver of data and is poor on meaningful insights.
Takeaway: Ensure that even the POC has a reasonable scale (functionality and complexity), uses adequate data and is supplemented with the right context to be representative of real-life scenarios.
Taking the road to nowhere: Quite often, there is an excessive focus placed on the POC without defining a clear roadmap on how the POC will transcend into a full-blown production rollout. Once the POC is over, the metrics measurement is hazy, the next steps are unclear, the major stakeholders are not aligned and the nuances needed to scale the POC have not been addressed. The POC stays stuck in a quagmire.
Takeaway: Define the roadmap from POC to production and ensure adequate socialization with key stakeholders who will ensure successful transition to production.
The (false) lure of machine learning nirvana: Crawl, walk and run — that’s how life works. Expecting machine learning to give instant gratification is a blunder. Machine learning projects take at least a few months of arduous efforts to collect and curate the data and tweak the algorithms to eke out the relevant insights. Even then there is no silver bullet, no single big-ticket item. It’s all the small analytic victories that add up over time, so expecting a single isolated POC to magically deliver insights is a myth.
Takeaway: Have realistic expectations of what machine learning can deliver and in what time frame. Ensure that the use case has enough fuel (i.e., data) to fire the machine learning algorithms.
In closing, although the industrial IoT revolution is for real, we are bound to witness these early false starts and growth pangs. Significant barriers related to leadership, skills and expertise, ROI quantification, security and implementation still need to be surmounted to realize the true promise of industrial IoT. As with other tech adoption, we will eventually cross the chasm to scale the industrial IoT initiative — and hopefully quite soon. Until then, we are stuck between a rock and a hard place!
All IoT Agenda network contributors are responsible for the content and accuracy of their posts. Opinions are of the writers and do not necessarily convey the thoughts of IoT Agenda.