Industry consultant Mark Konya weighs in on eight critical factors for analytics.
Mark J. Konya, P.E.
The contemporary view of deadly sins originated from the works of 4th-century monk Evagrius Ponticus; the eight “vices” he advanced include: gluttony, lust, avarice, pride, sorrow, wrath, vainglory, and sloth. These historical sins have parallels in modern analytics processes which are meant to deliver tangible value to businesses.
The acronym “DOWNTIME” illustrates the eight deadly sins of analytics: defects, overproduction, waiting, neglecting human potential, transportation, inventory, motion, and excess processing. These modern “sins” define types of waste we should all be concerned with.
Businesses conduct daily operations through processes, or sequences of activities arranged to deliver a product or service to a customer. Ideally, each step of the process is a value-added activity or one which directly transforms an input into a deliverable the customer will pay for. Most processes, however, contain non-value-added activities – wastes – which consume resources but do not directly benefit the customer.
Whenever possible we should strive to eliminate non-value-added activities from analytics processes to produce an efficient, smooth flow of actions and information which deliver results meeting customers’ demands.
Another way of thinking of this is to imagine an analytics process as a river. In a perfect world, we want the “analytics water” to flow smoothly from the river’s head to its mouth. Over its course, the river encounters barriers to flow such as direction changes, lakes, rapids, and dams. These are analogous to the wastes which also inhibit the smooth flow of analytics activities from data to discovery to deployment.
For clarification let’s illustrate the eight modern “sins”, or wastes, with simple examples:
Defects
A defect is a deliverable which fails to meet customer requirements. The most common example of an analytic defect is delivering an incorrect result. For example, if you’re performing a reliability analysis of transformers and fail to account for censoring then the model will be biased, returning defective results.
A thorough understanding of customer requirements and an operational definition of “defect” helps us know when this situation exists.
Overproduction
Overproduction means that an excess of product or service is produced by the process without regard for demand. For example, if you are generating many analytic insights without regard for their business demand then you are engaged in overproduction. By fulfilling only customer requests which add value to production processes this waste can be eliminated.
Waiting
Assume you receive a request to produce a maintenance prediction. The first steps in fulfilling the request are to identify, access, and set up the data to be analyzed. But after identifying the data sources you realize an access request, which requires three days for approval, must be processed by data security. In the meantime, you are left waiting to begin your analysis. In this instance pre-processing a security request with enough lead time would eliminate the wait.
Neglecting or not utilizing analytics talent
Underutilizing your analytic staff’s knowledge, skills, and abilities have a detrimental effect on deliverables and on the organization. I recall a case where a company invested significant resources in training a data scientist but then relegated the individual to a reporting role; he left the company after a short stint in this position.
By not tapping into this individual’s talents the company paid twice. The first was an opportunity cost incurred by foregoing the valuable insights this talented analyst could have produced; the second was the cost of refilling the position with a qualified data scientist.
Executives make a big mistake when they fail to fully tap the knowledge, skills, and abilities of front-line workers. These workers participate directly in processes and are best-positioned to contribute insights which improve the flow of “analytics water.”
Transportation
While not a major problem in most offices, moving people or “things” unnecessarily causes waste. In the world of analytics, these “things” are likely to involve data – and lots of it. For example, unnecessarily moving data from edge devices to permanent storage can be quite expensive; it’s far better to transform and filter at the edge, retaining only the data which is useful for analysis purposes.
Inventory excess
Storing excessive inventory is expensive. This waste is commonly encountered when excessive inventory masks production problems or anticipated high demand never materializes.
In analytics processes, this could occur by creating and maintaining data stores in obscure locations which are not registered or traceable. You may have an excess inventory problem with data if this is the case. Eliminate excess inventories to help “analytics water” flow smoothly.
Motion
Wasted time and effort through unnecessary movements is a non-value-added activity. In an analytics office, this might include visiting other offices to search for data, creating ad hoc reports which add little descriptive insights or entering data in multiple locations.
Wasted motion in an analytics office environment is not always easy to identify, but through a detailed understanding of activities and their value, it can be mostly eliminated.
Excess processing
This occurs when a task involves more work than necessary, or a process produces a higher quality deliverable than required by the customer.
If your customer requires you to generate a day-ahead hourly forecast of energy demand with a MAPE of 3%, but your analysts are working overtime to find a 2% model after they’ve generated a fully acceptable 3% model, then excess processing is involved.
Don’t let these eight modern “sins”, or wastes, creep into your analytics processes!
When they do you will be saddled with costly, non-value-added activities which harm your capability to deliver analytics value to your customers. To identify and eliminate these wastes one must first understand the current state, then follow a regimen of process improvement to achieve a new, more efficient state. Value stream mapping is one piece of this regimen; it maps out the process, helps separate value-added from non-value-added activities, and identifies waste.
In the end, knowledge is the first defense against the eight deadly “sins” of analytics! If you are alert to their symptoms and call for help from your continuous improvement experts when they first appear, you can restore good health to your analytics processes.
Mark Konya, Advisory Industry Consultant in the Global Energy Practice at SAS, earned graduate degrees at Washington University and University of Illinois in the US. With over 35 years of utility experience in power generation, distribution engineering, process improvement, and leading customer analytics initiatives, Mark shares his insights and expertise with customers on analytics strategies and deriving value from analytics initiatives. Mark is a member of IEEE/PES, a licensed professional engineer, a Six Sigma Black Belt and a certified Lean facilitator.
