Chapter 12. The Economics of IoT Analytics

You sit in your office staring out the window at the building across the street. You feel the tap, tap, tap of your pen against your forehead as you fiddle with it lost in thought. Your boss left your office twenty minutes ago.

"We need a predictive maintenance program," he had said, "But it can't lose money! We need more revenue coming in. But I don't want you wasting your time on analytics that won't make us money."

You agreed and nodded your head with a thoughtful smile firmly held on your face. You know how to think about analytics but not necessarily how to look at it from a revenue and cost perspective.

"And I want you to keep trying new things! Let's keep up the experimentation. You are doing great! Just don't lose money," he continued.

Okay, sure

, you think.

How do I do that?

This chapter will explain how to think about a business case for IoT analytics projects. It discusses ways to optimize the return on investment by minimizing costs and increasing...

The economics of cloud computing and open source

At first glance, it may seem like cloud computing should be valued in the same way as on-premises systems. It may also seem that open source solutions always make the most sense as there are no license fees. It is free, right?

We will walk through some ways to think about valuing cloud computing and open source software for IoT analytics. Like most things, it is more complicated than it seems at first.

Variable versus fixed costs

Variable costs expand and contract with the amount of usage. Examples are automotive production parts for assembling vehicles, fuel in transportation, and cloud services. Fixed costs are constant no matter the amount of usage needed. Examples are data centers, rack servers, and equipment (or cloud environment) management.

In a traditional business case, fixed costs are assumed to continue until they are able to be sold at some anticipated amount. This rarely occurs in a short period of time. Variable costs are assumed...

Cost considerations for IoT analytics

There are some cost considerations associated with IoT analytics specifically. They are not unique to IoT but become more pronounced with the scale and processing needs associated with it.

Cloud services costs

IoT analytics requires multiple layers of cloud services. There is often an IoT hub, message queue, load balancing, compute, storage, ML service, data warehouse, and security services included in an IoT analytics solution.

As mentioned previously, these costs can add up quickly if not designed carefully and monitored closely. Make sure to include all services in a business case and use cloud flexibility to minimize costs.

Expected usage considerations

Model your expected IoT analytics usage requirements carefully. A business case should not include only the services used for analytics during data processing and storage. It should also include costs for both ad hoc analytics and data science modeling on the stored historical data.

Thinking about revenue opportunities

Once you have an IoT analytics platform set up, have cleaned and refined the data, used exploratory analysis and visualizations to understand it, enhanced IoT data with external datasets, applied geospatial analytics where appropriate, used data science to understand and predict, and organized your data lake for analytics efficiency; you can take a step back and think about additional revenue opportunities.

These are ones that you may not have considered before you had a deep understanding of the data. Opportunities may also simply not have been previously possible until an IoT analytics infrastructure was built.

Here are some revenue possibilities to consider:

• Extensions to your current business processes: These ideas closely align with what you are already doing and can add additional value to your customers. It is always beneficial to think of services from the perspective of the end customer. Put yourself in their shoes and think what would save you...

The economics of predictive maintenance example

Predictive maintenance is a common value proposition cited for IoT analytics. We will walk through an example as a way to highlight how to think financially about when it makes sense and when it does not.

Situation

The economics of predictive maintenance may not be entirely obvious. Believe it or not, it does not always make sense, even if you can predict early failures accurately. It many cases, you will actually lose money by doing it. Even when it can save you money, there is an optimal point for when it should be used. The optimal point depends on the costs and the accuracy of the predictive model.

The value formula

A formula to guide decision making compares the cost of allowing a failure to occur versus the cost to proactively repair the component while considering the probability of predicting the failure:

Net Savings = (Cost of Failure * (Expected Number of Failures - Expected True Positive Predictions)) - (Proactive Repair Cost * (Expected...

Summary

In this chapter, we discussed the economics of cloud computing and open source software. We reviewed cost considerations for IoT analytics and how to think about revenue opportunities. We also covered ways to think about an IoT analytics business case.

The economics of predictive maintenance was discussed as an example of how to view the costs and benefits of an applied IoT analytics ML model. An example framework for calculating the cost benefit was presented along with some R code, demonstrating how to generate a cost curve based on the ML model.