Website update in progress! You might be logged out of your account. If this occurs, please log back in.

Website update in progress! You might be logged out of your account. If this occurs, please log back in.

Sign Up

Please use your business email address if applicable

How to Implement IIoT Predictive Analytics Solutions Without Hiring Big Data Scientists

Background: Lots of Interest, Little Bandwidth

IIoT predictive maintenance is one of the hottest topics today. A comparison of Internet searches for “Industry 4.0” and “predictive maintenance” finds a spike in interest over the last couple of years (see Figure 1).

This increasing demand for IIoT predictive maintenance solutions has been addressed by numerous publications and analysts, among them Harvard Business Review1 and PwC2.

IIoT predictive maintenance is rapidly moving from strategy to execution. Senior executives are embracing the economic potential of increased uptime and higher production yield rates. At the same time, concerns are being raised about the availability of big data professionals.

In the Emory University The Future of IIoT Predictive Maintenance Research Study3, maintenance reliability professionals were asked to rate their attitudes regarding predictive maintenance deployment. The statement that generated the highest rate of agreement (5.8 on a 9-point scale) was “senior executives recognize the potential of predictive analytics.” At the other end of the spectrum, the statement that generated the lowest level of agreement (3.2 on a 9-point scale) was “we have sufficient staff of data scientists to deploy predictive analytics. Summary data from the study is shown in Figure 2.

Research from the Massachusetts Institute of Technology (MIT) conducted in 2012 is still relevant today and provides more context about the underlying concerns. The researchers note that data has become cheaper and there are new technologies to analyze data. However, tools “require a skill set that is new to most IT departments, which will need to work hard to integrate all the relevant internal and external sources of data.”

The weak link is data science talent. There is an abundance of unstructured data that requires deep expertise in machine learning and artificial intelligence (AI) if it is to be turned into valuable and actionable information. The dearth of trained data scientists has been well-documented. According to the report, The Quant Crunch: How the Demand for Data Science Skills Is Disrupting the Job Market, the demand for data scientists and data engineers will grow by 39 percent by 2020 in the U.S.

With the surge in demand for data scientists, it is difficult for industrial plants to compete with compensation packages provided by Wall Street and Silicon Valley.

Big Data Analytics Without Big Data Scientists

The lack of qualified data scientists is not a new issue. Part of the problem stems from a lack of education, with fewer than one third4 of global universities offering a degree in data science.

Let’s start by reviewing some of the current solutions. Global research firm Gartner has presented its approach in an article titled, How to do Machine Learning Without Hiring Data Scientists 5. It provides four strategies:

1. Turn existing staff into data technicians

A couple of years ago, Gartner came up with the term, citizen data scientist (CDS). It refers to a technician with mathematical capabilities who can be trained to perform data science roles.

Does this work? The answer depends on the expectations for the CDS. For example, a company that recruits and hires many data scientists and engineers may see few areas where someone lacking formal training can fill those roles.

2. Form alliances with academic institutions

Gartner suggests working with academic institutions that provide advanced degrees in data science. Some ideas include class projects, internships, and hackathons.

Although companies can utilize exceptional students to work on correlation benchmarking and data labeling to run supervised algorithms, keep in mind that this requires significant mentorship and is more the exception than the rule.

An internship program can provide great value to a company and a learning experience for students. However, it is a band-aid solution to a severe skill set shortage.

3. Use third-party consultants

It is no secret that with enough budget, companies can hire highly paid consultants to do the work of their employees. However, this is not a scalable solution and it merely camouflages a company’s inability to address its own needs internally. Shifting the burden from full-time employees to external vendor simply kicks the can down the road.

4. Purchase software applications

The idea that an off-the-shelf or even high-end software application can alleviate the need for big data scientists is wishful thinking. A plethora of software solutions exist, ranging from open source to custom applications. What do they all have in common? The need for skilled professionals to operate.

The Other Alternative: IIoT Predictive Maintenance as a Service

There is good news on the horizon. According to a Gartner report, in excess of 40 percent of data scientist tasks will be automated by 2020. Instead of band-aid solutions, industrial plants need to recognize that they are unable to change the long-term labor market or address weaknesses in the education system. The alternative is to build and acquire solutions that require almost no human intervention in the development and maintenance of the solution. Maintenance reliability technicians and engineers will not gain competencies in big data or machine learning and this reality needs to be accepted.

What is the alternative? Automated systems that use AI algorithms to analyze industrial plant sensor data and provide alerts of evolving asset degradation and failure. IIoT predictive maintenance surpasses traditional supervisory control and data acquisition (SCADA) monitoring because algorithms detect anomalous data patterns or patterns of anomalous behavior. With the traditional SCADA approach, only breaches of manually set controls are monitored and, in many cases, alerts happen too late.

Specifically, IIoT must be based on automated machine learning, whereby the specific algorithm applied to the data set is automatically selected. In reality, citizen data scientists and college interns lack the skill set for meta-learning, artificial intelligence, etc.

More and more companies are turning to cloud-based solutions, and with good reason. If you cannot bring the data scientist to your industrial plant, then you can bring your industrial plant data to the machine learning experts.


IIoT predictive maintenance impacts production yield rates, revenue, and bottom line profitability. It requires the adoption of new solutions and new ways of doing business. The shift from Industry 3.0 to Industry 4.0 is based on applying machine learning and big data to operations. Existing people, processes and technologies will not suffice.

As C-level executives prioritize IIoT and focus on uptime, it will require industrial plants to embrace solutions that incorporate the real-time analysis of operational data without adding to one’s workforce and disrupting ongoing production.







Deddy Lavid

Deddy Lavid,(Ben lulu) is a CoFounder of Presenso and its Chief Technology Officer and Vice President of Data Science. Deddy is considered a leading expert in the application of automated machine learning to industrial analytics. He has received several awards and prizes in the field of homeland security. Deddy authored twelve machine learning patents and published extensively on big data and academic research topics. Deddy holds a M.S. degree in Computer and Information Science and a B.S. degree in Computer Science from University of Haifa.

ChatGPT with
Find Your Answers Fast