Evaluate Weigh the pros and cons of technologies, products and projects you are considering.

Democratizing analytics is key to bringing IoT success to SMBs

Technology has often been dubbed the great equalizer for SMBs, empowering them to compete with industry goliaths more effectively than they could in the past. The internet of things is the latest technological enabler that companies are eyeballing, and it is poised to change the workplace — especially as the number of devices deployed by businesses continues to grow.

The swelling IoT enthusiasm is understandable. Enterprises are using devices for every facet of business operations, and as sensors have become more affordable, companies are better equipped to utilize them to record everything from human interactions to machine activity. This gives businesses access to an unprecedented amount of raw data, which can be fed into their analytics to drive improvements throughout the company.

Whether IoT is enabling manufacturers to optimize factory operations or helping retailers improve the customer experience, the IoT hype has been fueled by the wild success of some of the early adopters. Additionally, the high-level IoT concepts that have graced news headlines recently are further driving excitement — stories about Tesla, Uber and traditional car companies like Ford building self-driving cars that rely on IoT or moves like Amazon extending Alexa to new products and devices.

But most of this hype is being driven by the investments of industry titans — the Googles, Amazons, Fords and eBays of the world. What about smaller businesses? Enthusiasm may be high, but you don’t often read about many small or midsize businesses taking full advantage of the wide range of data available to them to fuel IoT projects. If you could put numbers to the IoT success seen at small businesses, it’s likely that the facts wouldn’t match the fervor.

The IoT conundrum: SMBs face a data disadvantage

The big challenge many businesses face with fully utilizing IoT is that the definition of their data and analytics strategy is ambiguous in the first place. Some enterprises will say they have a reporting and analytics strategy, when all they actually have in place is a data visualization solution like Tableau — or worse, their strategy constitutes some sort of basic spreadsheet reporting. In these cases, their IoT initiatives are doomed from the start.

Even when businesses do have proper strategies, most lack the right personnel — data scientists — to successfully manage the process. Data scientists play a vital role in data and analytics strategies as individuals who can direct, interpret and validate the output. New tools and technology can help businesses process data, but the personnel are the brains behind the scenes and they aren’t easily replaced. That’s why some data scientists are commanding salaries and benefits of more than $250,000 annually.

The issue is only compounded further by the fact that developing and executing on a comprehensive data and analytics strategy is a real challenge. You need to be able to manage data effectively, from collection and access to cleansing and preparation. Then you must be able to determine which analytical model will yield the best predictions, and that requires data scientist expertise to train, test, evaluate and score the models. Once the data and model is set, you need to figure out how to operationalize your analytics and use them in production. Finally, the process itself must always be under review in the face of changing business conditions.

At the end of the day, all of these different elements of successful IoT deployment cost money and resources, and that’s why the real promise of IoT has been limited to industry titans thus far.

A new approach to IoT

When it comes to designing a data and analytics strategy that will maximize the value of IoT, every business faces its own unique challenges. However, the resource problem is persistent across most SMBs, and the answer is not as simple as hiring a large team of data scientists — that’s just not feasible. While analytics as a service can provide support for self-contained use cases like text-to-voice, “outsourcing” the entire process around strategic decisions is a no-go for most organizations as well. Even the influx of new tools that make data scientists more effective do not have much impact on the democratization of analytics for smaller enterprises.

One possible answer is automation. Now, you are probably thinking “isn’t that what the existing tools do already?” To earnestly make advancements in this area, the analytics process must be made capable of self-learning — and not just applying learning to the output of the analytics. This next-generation approach will apply meta-learning principles to machine learning, where learnings from one machine or entity are applied automatically to other machines or entities.

Meta learning will minimize the cost of running machine learning experiments by capturing learnings from prior machine learning experiments in the form of metadata, and then applying these learnings for future experiments. This is critical for several reasons:

  1. You can greatly increase the accuracy of the analytical models, which will directly impact your business outcomes.
  2. You can achieve faster outcomes, which makes your business more agile. Analytics are only useful if businesses get them while they are relevant, and improved agility helps enterprises realize and act on information even sooner.
  3. You can gain better control of your infrastructure and rein in the cost required to run the machine learning algorithms. This is especially important, as the volume of data available to businesses is growing exponentially alongside the proliferation of IoT devices.

Given the shortage of analytics skills — alongside the shortage of data skills — many businesses aren’t taking full advantage of the promise of IoT. However, if the industry can democratize analytics, more businesses may be better poised to capitalize. This next-generation approach essentially automates the data science lifecycle, greatly reducing the need for a high-cost and high-demand resource not available to most organizations.

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.