By John Kennedy Joseph, Director, Data Analytics
Can RFID and IoT sensors assure 100% read accuracy? If not, how can I close the gap?
Which type of analytics will generate the desired insights from tag and sensor data?
Can I get foresights in addition to insights from Visi-Trac Analytics?
A course correction for RFID / IoT Analytics
Gone are the days when business users need to pore over spreadsheet data, with or without charts, which may take days to understand and interpret. Analytics information is everywhere. Yet, project sponsors need to start tracking ROI and operational impacts as early as possible, and essential data is often not presented in a way to make rapid operational decisions. This gives rise to an opportunity for a course correction in the information being generated by RFID and IoT systems.
‘Last seen locations’ of specific equipment by make, model and serial number may be interesting, but the power of analytics comes from combining internally sourced data with data from public and third-party sources into insights to improve business operations.
Which Analytics should I focus on?
Following are four common types of analytics:
Of the four analytics disciplines, descriptive and diagnostic analytics are the most concrete and give hindsight into what has happened and why. These analytics are typically called business intelligence and are much used.
The other two disciplines – predictive and prescriptive – are a step up the analytics ladder. Both give insight, and even foresight, to support decision-making. Predictive and prescriptive analytics incorporate statistical modeling, machine learning, and data mining, to forecast what might happen and recommend actions based on that forecast.
Prevention is always better than cure. Treatment is better than post-mortem.
We’ll focus on the two key types – predictive and prescriptive analytics, here.
Is Predictive Analysis a ‘fortune-teller’?
The answer is No. We may be tempted to think of predictive analytics as a fortune-teller who tells you what the future holds. However, the truth is that no analytics can do that. Predictive analytics can only forecast what might happen in the future because all predictive analytics are probabilistic in nature. The three keystones of predictive analytics are decision analysis and optimization, transactional profiling, and predictive modeling. Predictive analytics exploits patterns in transactional and historical data to identify risks and opportunities. In other words, you take data that you have to predict data you don’t have.
Is predictive analytics the same as business intelligence?
The answer, unfortunately, is also No. Business intelligence provides information based on existing data, whereas, predictive analytics provide information that can be derived from the existing data. Later in this article, we’ll see some scenarios for using predictive analytics and how it differs from business intelligence.
In an ideal situation, we would prefer prevention rather than treatment. Rather than waiting for testing to be positive for a disease, it would be wise to look for prescriptions the moment we start seeing or are likely to start seeing symptoms. Prescriptive analytics goes beyond simply predicting options in the predictive model and suggests a range of prescribed actions and the potential outcomes of each action. Prescriptive analytics is a more advanced use of predictive analytics. Since a prescriptive model can predict the possible consequences based on different choices of action, it can also recommend the best course of action to achieve a pre-specified outcome.
Google’s self-driving car is an example of prescriptive analytics in action. The vehicle makes millions of calculations on every trip that help the car decide when and where to turn, whether to slow down or speed up, and when to change lanes — the same decisions a human driver makes behind the wheel.
In the energy sector, utility companies, gas producers, and pipeline companies use prescriptive analytics to identify factors affecting the price of oil and gas to get the best terms and hedge risks. Predictive and prescriptive analytics are co-dependent disciplines that take business intelligence to unprecedented levels. With both forms of analysis, business executives and leaders gain both insight and foresight.
How can I get foresights, in addition to insights, in RFID and IoT systems?
To understand better, let’s review a couple of scenarios to understand the possibilities.
Scenario 1: Tracking Equipment and Finding Variance in Read Rates
An RFID solution has been deployed to track equipment with appropriate analytics. Upon tracking the equipment, we find out that asset movements are not tracked as accurately as expected when equipment passes by a certain portal. It is a multi-location implementation and we see this behavior in most locations, but only in specific rooms. On analyzing trends, we find a commonality in that a particular reader/antenna combination exhibits this behavior under certain conditions.
Up until this point, diagnostic analytics gives us insight into the types of readers and antennas causing an exception condition. Along the way, we figure out remedies to handle the readability issue to make sure there is the required accuracy from tag reads.
Let’s see how the data collected so far can be used in prescriptive analytics. The next time we extend this deployment to other locations, we’re careful to make use of the insights and we plan in advance to mitigate any such occurrences. Thus, getting the foresight to take corrective actions based on the predicted re-occurrence of exception conditions.
Scenario 2: Tracking Vehicles and Monitoring Device Battery Life
Let’s expand to a larger IoT solution using GPS technology.
A global corporation has deployed high value assets across multiple countries. These assets, vehicles and equipment, are tracked using GPS devices. The objective of the IoT program is to reduce costs by increasing asset utilization.
By the nature of the technology, GPS devices need to have enough ‘sky visibility’ to see a minimum number of satellites to be able to report. Given weather conditions and latitude, sky visibility is not the same across the globe. Hence there is a difference in device reporting patterns. Another key aspect for devices to report accurately is their battery level. Some devices have solar powered rechargeable batteries, and some do not. Especially for those devices that do not have rechargeable batteries, it’s critical to verify the devices have sufficient battery life.
Using data collected from the asset tracking system, an analytics engine can predict battery life based on current battery levels. Based on this learning, the analytics engine can trigger proactive notifications of a list of devices which are likely to stop reporting in the next ‘x’ number of days/weeks.
These predictions help decrease assets’ non-reporting downtime. In other words, visibility of asset movements / utilization is improved when the IoT system recommends an action – to replace or re-charge batteries before devices stop sending data – based on predictions of battery behavior.
Prescriptive Analytics Drive Greater Productivity
We can extend this algorithm by combining other key factors as follows:
For rechargeable GPS devices, weather plays a critical role in making sure the batteries are (re)charged as required. If the weather is abnormally cold with very little sunlight, the recharge time is going to be long. With the amount of data captured in the past, the analytics engine can determine the expected re-charge time during different weather patterns and plan ‘prescriptions’ accordingly.
Yet another factor that can be added to predictive analytics is the forecast of equipment movements. The more that equipment moves, the more times the GPS device reports, and hence, the shorter the battery life.
Therefore, the analytics engine can understand the battery level in conjunction with the weather and the expected rate of movement of equipment to help predict the course of action needed, and recommend prescriptive actions.
Similarly, we can add any number of scenarios to this logic to capture outlier conditions and achieve greater productivity of the system.
This way we generate foresights in addition to insights through predictive and prescriptive analytics.