The Human MDM
I experience Deja-Vu every single time I visit a CIO, VP IT or Analytics at a manufacturing organization. Then again, it does not really matter what industry the client operates in. The Deja-Vu has to do with the appearance of one or two staffers, often organizationally situated in the Line-of-Business IT function of a core business unit, as part of my workshop with said company.
These unsung heroes are typically analyst-level folks who have the dubious pleasure of being the, what I call, Human Master Data Management (MDM). (my copyright) Note: At this point, you should imagine hearing the “MDM” with diminishing echoes.
Human MDM is an existence, not a function. Nobody really knows where to put it in the organizational structure for guaranteed and scalable success. More importantly, most of the data consumers or creators up or downstream really understand the importance and inherent futility of this existence role. I ran into such a “data cyborg” at a very large insurance company a couple of years ago. He had no idea who, and under what guidelines and motivations, customer and policy records were created and what analytics were run based on his set of tasks. Even worse, he had no idea what the KPIs of these analytics running on his data wrangling work were. Why do you ask?
The Human MDM job scope is to basically acquire data sets from a multitude of systems; ERP, CRM, supply chain, EDW and so on; clean them, transform and load them using Microsoft Excel into one or more analytical applications being used, which often are Microsoft Access databases. Our insurance example pretty much knew based on what key to create joins and how to clean up addresses and SIC codes.
Is Lean Investing Keeping You Starving For Data?
Just like Homo Habilis, Human MDM people are just another offshoot from the evolutionary tree. They did not discover fire or start walking upright. They have become purveyors of special-project data feeds. While this function is extremely important for an organization to tweak its operations; think to reduce stocking levels, adjust lead times or market a new product to a customer segment; it is dreadfully under-engineered.
These folks – bless their heart – are trying to do enterprise-level work with Microsoft productivity tools, freeware loader apps and a lot of hand-coding. Two customer anecdotes come to mind here. One ERP consultant once told me that a pharmaceutical company basically runs on Excel instances housed on thousands of employees’ laptops. The surprising thing is that this company was extremely profitable – assuming they knew their customer profitability for now – and at the forefront of innovation. Imagine what its profitability would be if they would use a more governed, enterprise-strength data management strategy.
The second anecdote was from a recent conversation with the controller of a process manufacturer. He stated that his company had humans do a machine’s job for anything having to do with data management. He literally told me that “Jessica” (name changed to protect the innocent) was basically an obsolete resource as she spent 100% of her time chasing and correcting customer and product data for him. This firm had somewhere between 5-10 “data operators” whose time was 100% consumed with editing and fixing data. Most of them had no idea about the up- and downstream impact of their work. Moreover, their workload was gigantic, the backlog according and the quality questionable. So now I ask, “When it comes to quality: How lean is too lean?” Aside from throwing bodies at this problem, many industries, especially manufacturing, have reduced this work to a level barely sustainable.
So how do we fix this problem? How can we ensure “Jessica” not only keeps her job long-term but also gets redeployed to a higher value-added task set, such as actually doing some analytics, impact studies around chart-of-account changes or alike? Can we turn her into a data analyst? What about the data science function such an organization may have on the R&D side? Is that just another parallel team doing the report bidding of executives on a more academically-challenging level but ultimately still largely dealing in data acquisition, profiling, cleaning, prepping, etc.? As we all know, cluster analysis, machine learning and RPA all require large, clean and well-understood data sets and 60% of such work is centered on the data underlying the insight.
What are we to do?
Well, let’s start with giving these poor folks the right tools. How about some business-rules driven data profiling and automated standardization and deduplication capabilities? How about machine learning-suggested match keys? I could go on and on. The point is that packaged software exists for a reason. Companies like Informatica have built their entire business to help ensure their customers are not inundated with extensive, mundane and often repetitive data management tasks. Let’s graduate these stakeholders into the realm of Robert Reich’s “thought workers”.
Think of it this way. A peer-reviewed, academically published study I ran in 2015 based on 54 organizations indicated that the median productivity gain for a company in this data set was between 6 and 16 FTEs. This means that the total excess cost of data management in an organization was worth the fully-loaded employment cost of 6 to 16 people. Note that for this result, the respective sample included organizations of which 41% had fewer than 10,000 employees, 43% generated less than $3 billion in annual revenue and had fewer than 500,000 consolidated records. Adjusted to my updated methodology and accounting for the current industry-average data quality, we would still end up with at least 1.6 to 11 FTEs. If you add other cost factors to this, such as customer acquisition, order or service cost; supplier spend; equipment maintenance; regulatory risk and so on, the total amount data management drives is quickly going into the millions, if not tens of millions. A staff of a couple of Human MDMers cannot capture this opportunity with their current charter and toolset. After all, Homo Habilis used stone tools to thrive. Walking upright in the Savannah to avoid getting ambushed by a predator wasn’t doing it anymore either.