Why low-code is transforming data analytics in IIoT

More than 65% of application development will be in low-code by 2024. What does this mean for data analytics in IIoT?

When low-code is applied to software applications, its aim is to simplify the development of business logic. For example, an IT personnel may want to build a web form to let customers enter their insurance claims. A typical low-code platform would provide a form template and guide the IT person to set up the fields, connect to a database, and set up notification for the insurance adjuster. The low-code platform would enable the IT person to do this quickly and without software expertise, but it trades off speed with flexibility.  A low-code software can only provide a limited set of business logic capabilities, so it is never as flexible as a general programming language.

Data analytics is a different problem.  It is not about simplifying business logic development - there is little business logic in a data analytical process. When a user programs a data-intensive application in code, up to 90% of her time is spent on programming, not on working with data. For this reason, working in a general programming language is extremely inefficient for data analytics tasks.

This is why nearly all engineering data analytics software are in low-code: think Matlab Simulink, NI Labview, Keysight ADS, Cadence ADE, etc., etc.  These software all use a particular form of low-code: functional block programming, and they all provide a set of customizable block libraries to help users apply advanced analytics techniques.

The following diagram illustrates functional block programming in an IIoT use case. Say an edge gateway acquires raw data from multiple sensor sources. It then down samples and averages the raw noisy data over a time window. After that it formats the data, adds in metadata, and sets alert threshold.

In this example application, the user will be trying to answer the following questions:

• What’s the property of the raw sensor data?
• What’s the optimal sampling window?
• What’s the optimal data filter - algebraic, RMS, frequency-based, or something else?
• What’s the optimal alert threshold?  Does it shift over time?

Note that none of these questions have anything to do with coding or coding skills.  Instead, they require the user to work with the data at different stages of the processing path, to visualize data at different nodes, compare raw data with filtered data, experiment with alert threshold, etc.  The user may also want to work with simulated data sources before applying it to real hardware.

Data analytics software is designed to help users to work on data. It organizes data processing in a visual, modular, and hierarchical flow. The user can easily visualize the entire data processing path in near-real time. She can tap in at any node to visualize the data, adjust the algorithms, and move the processing steps around. The key is simplicity and flexibility. 

Prescient Designer, Prescient Devices’ distributed low-code edge solutions software, further helps users to optimize data analytics at the edge and in the cloud.  It unifies the edge and the cloud by representing them as functional blocks in the same functional block diagram.

As illustrated in the above figure, the “IoTG-001” block represents data processing at the edge, and it can be associated with and deployed to thousands of edge devices. The “Cloud Analytics” block receives data from the edge devices and performs further data processing in the cloud.  So users often need to answer the following question: how to optimize the processing at the edge vs. in the cloud?

In Prescient Designer, any functional block can be moved between the edge and the cloud.  For example, if the user wants to see the effect of cleansing the data at the edge instead of in the cloud, she would simply cut the “cleansing” block from “Cloud Analytics” and paste it into the “IoTG-001” block.  When there are thousands of edge devices in the system, this could significantly reduce the workload in the cloud.  The key again is simplicity and flexibility -- when users can easily perform optimization experiments, they are more likely to produce better solutions.

IIoT systems are more about working with data than with business logic.  Low-code enables users to build data analytics faster and gain deeper insights, and it cuts 50% to 90% development effort. At 50% better efficiency, an organization would build an 8x competitive advantage in 3 years.  As an offensive or defensive strategy, every organization should consider the potential impact of low-code to its business.