Manufacturing analytics is a $6 billion market that’s expected to grow to $36 billion (+19%) by 2030. With those kinds of numbers, you can expect a flood of vendors to the market, all hoping to capitalize on manufacturers’ long standing need to better understand and utilize the vast amounts of data at their disposal.
But this presents a problem: many of those rushing to solve complex problems will offer simplistic solutions that belie the effort needed to properly address data quality and analytical challenges. How can supply chain and operations leaders prioritize strategic initiatives based upon both value and effort when the latter is often obfuscated or misunderstood by people touting the former?
Data as (By)Product
The answer is to treat analytics and data as products like any other manufactured by organizations. A product is evaluated based on its market potential and its cost of production. Data can and should be similarly evaluated. What is its value? And is the cost of production commensurate with that value? These become critical questions.
Of course, things aren’t quite so simple — data isn’t exactly a product like any other. The difference between durable goods production and data production is that the latter is a consequence of the former. In that sense, data can be considered not simply a product but a byproduct. In general, byproducts can provide great revenue streams, but only if their value exceeds any subsequent refining and distribution costs. Data has potential value as well, but that value often requires refinement and distribution, all of which require some resources. How many resources? And how much refinement? It all depends upon the data.
Let’s use a tangible example by comparing enterprise resource planning (ERP) data with data from a shop floor programmable logic controller (PLC). Information in the ERP system is typically centrally managed and vetted, as this is a critical mechanism for costing and for measuring financial performance. Due to its visibility across the entire enterprise and use in financial reporting, ERP data is typically consistent and well understood.
Contrast this with data being generated by a PLC that controls some piece of equipment in a plant. This data is understood by very few (in some cases only a couple of people) and is often vulnerable to service interruptions, measurement inaccuracy (due to inadequate maintenance), and dropped signals. This data is not centrally vetted and is only governed locally (i.e. by those who understand it), if at all.
Looking at these two data sources, which one requires more processing and refinement? Which one can generate more value with fewer resources? The answer may appear obvious — but appearances can be deceiving.
Assessing the Value and Cost of the Data Byproduct
Businesses will usually dispose of some portion of any given byproduct as scrap. The majority of automation data today is treated similarly, as a byproduct that is scrapped due to the lack of clarity around its value in relation to the costs required to leverage it. But its value is like that of all products — based on market demand.
So, is data in demand? In short, yes, and in several ways. Customers want more transparency. Supply chain leaders want more predictability. Manufacturing directors want more efficiency. In many cases, managers, directors, and leaders have their compensation tied to performance metrics like overall equipment effectiveness (OEE). Ensuring the transparency and consistency of the data used to calculate these metrics becomes a matter of fairness. The less reliant the data is on manual input, the more it will be trusted.
But what kind of effort does it take to turn data from scrap into a value-adding product? Many organizations allocate the lions’ share of resources into downstream BI and analytics processes, without taking into consideration the impacts of upstream variability on the end result. These efforts are doomed to fail, since variability is the hallmark of PLC and automation data. So what’s the solution? Just abandon analytics until the data is of perfect quality and perfectly transparent? Surely not — that’d be the equivalent of scrapping a byproduct because it’s too difficult to grade.
Grappling With Quality Variability
When a product is in demand but has variable quality, its market value can be tied to its grade. At lower grades, it may have lower value — but it has value nonetheless. Automation data of lower grades provide value in two ways. First, by revealing a lack of consistency and control. Second, by shining a light on waste.
Inconsistent data with high degrees of variation can suggest a number of things, including poor production control and maintenance ineffectiveness. Data is a reflection of the variation of equipment and processes. If you want to drive towards predictability, you will never know how far you are from it until you start analyzing the data, not with the intent to predict, but with the intent to rate the feasibility of prediction.
Just as poor data quality can reveal poor control, it can also illuminate wasted resources. You cannot grade the importance of a data set without scoring its quality first. As a controls engineer once told me, “If the sensor isn’t critical to the operation of the equipment, no one will care about it and the data will be useless.” He is 100% correct.
The answer however isn’t to abandon all attempts at adding sensors or capturing data. It’s to ensure the accuracy of the sensors that matter and to stop wasting time on those that don’t. If, for example, a flow meter has been out of calibration for six months, why invest any money in replacing it? If the ineffectual sensor is both irrelevant to the control of a process and the object of attention and labor, it is a source of waste, and any efforts at storing its data are wasted.
Like any byproduct, not all data has the same quality. But regardless of the data quality, the fact remains that it is collected for a reason. Its relative importance to production supervisors or supply chain managers can paint a very clear picture of the health of some processes as opposed to others. And it can be used to eliminate waste, not just through highlighting loss leaders on a production line, but by shining a light on wasted efforts.
There is a wealth of manufacturing information out there that contains value at various degrees. Only a fraction of it is being properly leveraged because it is often treated in a homogenous manner. All data was not created equal, but there is value to be mined from it — as long as you know where to look.