CHAPTER 7: Essential Roles of Senior Management in Getting to the Next Level

This chapter is aimed at senior departmental and corporate managers. It is motivated by my earlier observation that data quality programs go as far and as fast as the senior manager (or management team) perceived to be leading the effort insists. This doesn’t mean that provocateurs cannot achieve a real result or two, or that data quality teams without strong mandates can’t make progress. They can, and they do, all of the time.

But the overall program reaches a plateau and after that progress is too slow and uncertain. Further, you’re lucky if one person in 500 has the insight, courage, wisdom, and persistence to be an effective provocateur–far too few for data quality to permeate the entire organization.

If you want your entire department or company to enjoy the benefits of high-quality data, then you must actively engage. Some 20 years ago, Dr. Joseph Juran, captured this cold, brutal reality brilliantly, making the case that leadership for quality cannot be delegated.33 While Juran was talking about manufacturing quality, his words ring true for data as well. If anything the data quality challenge is tougher.

You must insist that your organization get in front on data quality and, in so doing, close those wasteful hidden data factories; you must put the right people and organizational structure in place; and over time, you must make clear that contributing to the effort is not optional. Your efforts need not take a lot of your time, but they must be genuine, consistent, long-term, and forceful.

Instructions:

1. Develop a feel for the issues as they impact you personally. Understand the case for data quality and what you must do. Then decide: go for it or not.
2. Put a data-friendly structure and the right people in place. Adopt a federated model for managing data quality and get the right people in corporate and departmental lead roles. Don’t look to IT to lead the DQ effort.
3. Build the culture. Insist that your department or company get in front on data quality. Set high targets for the DQ team and your organization.
4. Actively engage.

Understand the Business Case

I started this book recounting a discussion I had with a senior media executive as he tried to understand what data quality meant. He resonated when I asked whether he trusted the data the last time he made an important decision (he couldn’t recall ever fully trusting the data).

This example underscores an important point—it is easy to think of data quality as some esoteric problem buried in computers. Many data managers unwittingly reinforce the misconception! So senior managers should first consider how data quality affects them personally. Perhaps you too have been frustrated when you couldn’t trust data needed for a big decision, perhaps you grew angry at a direct report when he or she “bent the truth” in describing a problem in his or her work, perhaps you were confused when three department heads gave you different numbers for last month’s results, perhaps you felt enfeebled when they had no data whatsoever to address a new issue, perhaps you worried whether a similar error could happen to you after a colleague was called out in the news for a big mistake rooted in bad data.

This exercise can be tricky—we’ve all grown accustomed to excusing little issues and viewing more major ones as “just one of those things.” Think through colleagues you work with and the sources in your personal network, making mental notes of those you simply don’t trust. Revisit important decisions, asking yourself how much of the data you considered turned out to be false, misleading, or otherwise not up to snuff. Look too at hidden data factories in your own work, things you and your staff do to accommodate bad data. Recall yourself as the rising star of Chapter 2. Now develop your own mental picture.

In parallel, develop the larger case for data quality. Include three components:

• Some deeper knowledge about how good the data really is.
• “So what?”—a better understanding of everything from day-in, day-out operations to the company’s long-term strategies to the plans to “compete with data.”
• Addressing the tough question, “do I want to put in the time and energy this will require?”

How good is the data?

You can get the deeper facts by conducting a few Friday Afternoon Measurements (Chapter 2) and talking to colleagues. Interestingly, I find that (when in a non-threatening environment and encouraged to think it through) most senior managers readily admit that data causes plenty of grief and recall many specifics.

So what?

You can estimate the hard costs associated with bad data by sorting out the costs of a hidden data factory or two and by employing the rule of ten (also Chapter 2). Expect numbers in the range of 20 to 50 percent or more of the total costs of operations.

Unfortunately, it is impossible (so far anyway) to associate hard costs with impacts such as:

• Missed opportunities.
• Bad decisions.
• Angered customers.
• Turf wars because people can’t trust each other’s data.
• Difficulties in executing longer-term strategies.

But estimable or not, these costs are real, and you must factor them into your business case.

To continue the so-what analysis, think more opportunistically. Think not in terms of reducing costs, but in terms of creating benefits. Ask specifically, “How would the company be better if the data was better?” Perhaps the company would enjoy less friction between departments; perhaps you could put some distance between yourself and your nearest competitor; perhaps you could better use the data to innovate. It may be trite to observe that data is inseparable from operations, day-in, day-out management, and strategy, but it is certainly true. When thinking this way, many see leverage in high-quality data. And the answer to the so-what question is not one big reason, but dozens of little ones.

Note that I did not pose this question as, “Can we do anything about it?” None of the instructions discussed herein are all particularly difficult, so you certainly can. Now to be fair, taken together, the new roles represent a big culture change, though certainly not beyond reach of any good management team. Even in the harshest light, you certainly can!

Competing with data

Whether you buy into the hype associated with big data, advanced analytics, “data is an asset,” and the Internet of Things, data is growing more important. So sooner or later, every company must have a data strategy.34 While a full discussion is beyond scope here, I do wish to synthesize points made earlier.

First, while you may not think about it explicitly, you’re already competing with data. Consider the enormous quantities of data you expose externally, to their prospects, customers, suppliers, competitors, financial markets, and regulators. Bad data can leave bad impressions, and long-lasting ones at that.

Second, recall the four basic strategies35 for competing with data (Chapter 2):

1. Becoming data-driven.
2. Using Big Data and Advanced Analytics to innovate.
3. Providing new and better content.
4. Becoming the low-cost provider (by reducing costs associated with hidden data factories).

The first three depend on high-quality data and the fourth is nothing but data quality.

Third, in the short-term, at the very least, you must protect yourself from the likes of Uber, which is transforming an entire industry simply by capturing “I’m looking for a ride” with “I’m looking for a fare.” Think an upstart can’t threaten you? Be circumspect in this regard. Ignore data and data quality at your peril.

Fourth, having a measure of proprietary data, data that you have and no one else does, is key to long-term strategic advantage (discussed more fully in Chapter 5). Even if you’re ready for a full-fledged data strategy, you should identify and focus first on proprietary data.

Two instructions are clear:

1. You must factor the longer-term strategy, even if not-fully formed, into the near-term case for data quality.
2. Over the longer-term, you must sort out a data strategy.

Are you up for it?

The final question, “Do I want to put in the time and energy this will require?” is tougher, especially in light of Dr. Juran’s observation that leadership cannot be delegated. There are plenty of reasons to answer, “No.” After all, you may have other priorities, wish to leave a legacy in a different way, or lack the energy to take on anything new, never mind data quality.

There are also plenty of reasons to say “Yes.” Data quality is the most effective way to cut costs, and it can help your company be a tougher competitor. On a personal level, advancing data quality may help you distinguish yourself from your peers. Better data will certainly help you make better decisions. Finally, you may use data quality as a means to explore the joys and perils of data, as emerging assets.

You may need some minimalist data quality program to respond to regulators, mollify customers, or keep costs from exploding. Just don’t confuse such a program with serious effort.

I have no easy prescription for answering this question. Just take a hard look in the mirror. If you’re up for it, great. If not, so be it.

Then synthesize the above. If the business case intrigues and you’re up for it, then engage, following the instructions that follow. If you don’t believe the business case, have a better opportunity, or are simply not up for it, then leave data quality for your successor.

Put the Right People and Structure in Place

This book explores the roles of people, as data creators and customers, process owners, embedded data managers, data quality team leads, and leaders. This section builds on earlier chapters, proposing a federated structure for managing data quality, advising on how many people a full-fledged data program will take, exploring where DQ teams should report, and describing traits you should look for in DQ team leads.

Adopt a federated model for data quality management

Take a look at Table 7.1. It proposes a federated model for managing data, along the lines of a federated model for managing people.

Table 7.1 A federated approach for managing data assets parallels the approach for managing other assets, such as people.

	People Management	Data Assets
Senior Executive	Usually one of the top few executives in the company.	In time a Chief Data Officer will become one of the top few executives.
Corporate HQ	Corporate succession; policy and administration.	Corporate metadata management; provisions for unique data; strategy; policy.
Departmental Staff	Help their units find the talent they need; on-board people.	Help their units become good data customers and creators; drive departmental data program; home for embedded data managers.
Everyone	Day-in, day-out people management; follow established HR processes.	Good data customers and creators; develop novel ways to compete with data.

The left-hand column summarizes pertinent features of “people management,” including a high-level leader, a headquarters group that sets and administers policy, and HR people seeded throughout and reporting into departments. Still, the most interesting feature is that HR does relatively little actual people management. That happens “in the line,” as managers build teams, provide feedback, negotiate performance contracts, advance (or not) the organization’s culture, and hundreds of smaller actions every day. And not just managers. Everyone contributes, as they seek to improve their skills, interact with each other in professional (and not-so-professional ways), provide 360° feedback, and again, in hundreds of smaller actions every day.

In my opinion, companies should manage data as professionally and aggressively as other things they view as assets. People and capital almost always qualify. While I could base Table 7.1 on either, I chose people because I see important analogies between “data” and “what’s in people’s heads.” See Data-Driven for a discussion of the larger implications of treating data as an asset.

Over time, you need to build something similar for data, as the right-hand column proposes. Analogous features for managing data, include a high-level leader (e.g., a Chief Data Officer), a data quality team at headquarters, and data quality teams in departments. The analogy extends to people and managers as well. After all, everyone touches data every day and so must take on their responsibilities as both data customers and data creators. And managers must take on special responsibilities to ensure that processes work well, and to build needed communications channels with data suppliers and customers.

Of course different companies employ different management styles. Some take a more “command and control” approach, others are more highly decentralized and so adopt a more decentralized approach. Match the degree of centralization for data to the degree of centralization your company uses for people.

How many people?

Continuing this discussion, see Figure 7.1, which provides “stake in the ground” estimates of how many people a solid data quality program will require. Not surprisingly, the more critical data is to creating and maintaining strategic advantage and the greater the intensity of interesting data, the more people required. These estimates stem from three sources:

• The sizes of quality staffs at AT&T and Motorola when they were in their quality heydays and won Baldrige Awards.
• Discussions with companies on the sizes of staffs associated with other assets, capital (e.g., Finance) and people (HR). While my study is hardly random, I’ve received numbers as low as 1.5 percent and as high as 6 percent, though most answers cluster around “about 2 percent.” To be clear, these estimates reflect the number of people in full-time staff roles (including embedded data managers). Of course, as Figure 7.1 notes, everyone bears some responsibility for both HR and money/physical property, just as all who touch data are both creators and customers.
• Experiences with those enjoying success with data quality.

Depending on the criticality of data to strategy and a company’s “quantity of interesting data,” numbers can range from as low as 1 percent to as high as 4 percent. For most, I recommend 2 percent as “getting started” number, with two-thirds to three-quarters of those in embedded data manager roles and the remainder in staff roles. If those shares seem high, compare it to the 50 percent of time people devote to the hidden data factory!

Figure 7.1 Percentage of people in staff and embedded data manager roles.

Where should data teams report?

I’ve also noted herein that principle responsibility for data quality rests with data creators and data customers and that data quality teams should not report into IT. Rationale for this conclusion comes in two forms. Liz Kirscher, past president of the data business at Morningstar, explains it this way: “We would no more have Tech run data then we would have Research run Tech. They are different kinds of assets.”36

There was a time when the notion that data teams should report anywhere other than IT provoked some controversy. Today most IT departments readily admit that “the business owns the data.”

Second is the fact that DQ teams reporting to Tech have not enjoyed the same levels of success as have teams reporting elsewhere. If your current data teams report into IT, find a better spot as quickly as possible.

Of course, “not in IT!” does not address the question of where the DQ Teams should report. For the corporate data team, my order of preference is as follows:

1. If you have a true Chief Data Officer (a lot of companies have CDOs, but relatively few are truly C-level jobs), then the DQ Team should report into him or her.
2. If you have no CDO, then to the CEO. I don’t know of any companies doing this right now. But during the heyday of the quality movement in manufacturing, heads of Quality departments reported to CEOs, an arrangement that worked exceptionally well. In and of itself, data quality does not demand a C-level job. Depending on the size of the company, one or two levels down.
3. To either the Chief Operating Officer or Chief Financial Officer, whichever is better positioned to nurture the new area.

This same logic applies for Departmental data teams. Get them as close to the department’s most important data as you can, one or two levels below department head.

Hire quick learners who aren’t afraid to make mistakes

I’ve worked with many solid data quality team leads and I have three excellent ones in mind for this instruction on what to look for: Nikki Chang at Chevron, Karl Fleischmann at Shell, and Lwanga Yonke, at Aera Energy. First, look for someone who thinks independently and learns quickly. Fleischmann and Yonke, in particular are veracious readers. Chang questions everything. Look for someone who is already respected or can earn respect quickly. A deep understanding of your business is a real plus. Chang spent twenty years working on the information systems that support her unit, Fleischmann is a Ph.D. geologist, and Yonke has an advanced degree in petroleum engineering. All have huge networks and enjoy deep respect within their companies.

Most companies, big ones anyway, need at least one world-class data quality expert. The practical reality is that you’ll have to seek outside help until you develop or hire one.

Look for someone with the courage to take on sacred cows and the confidence to make mistakes. It is important to have some experience in data or quality, but fast learners pick it up quickly enough. A passion for data, and a dogged persistence are more important. Finally, look for extroverts. I find that the hard work of data quality is done in other people’s offices, as one builds relationships, explains roles to people, helps solve problems and the like. While I’ve known some introverted data quality team leads, extroverts enjoy the role more.

Insist that the Organization Get in Front on Data Quality

I find discussions about culture incredibly frustrating. People say things like “that’s just a cultural issue” or “we have to change the culture first” as if those comments somehow clarified the matter. They don’t, at least for me.

First, great companies already have great cultures, cultures of innovation, of service to customers, of citizenship, or commitment to people, and the like. Improved data can enhance those cultures. So attach data quality to values your company already holds dear.

Second, I’ve also noted several times that organizational momentum tends to relieve data creators of any responsibility for data quality. Data customers are pressed for time and they get immediate relief by making corrections, in time instantiating the effort in hidden data factories. Your biggest challenge lies in deflecting that momentum. No one will misunderstand the logic. But old habits die hard! You and your team leads must become “evangelists in chief” for the new approach to data quality.

Third, cultures are built on deeds, not words. Of course big things matter, but it’s the dozens of little, unscripted things that leaders do that advance, or retard, the culture.

Fourth, there is a temptation to expand any cultural issue into a larger one. For example, many companies are struggling with what it means to be data-driven and you certainly can’t become data-driven when you can’t trust the data. Becoming data-driven sounds (to many) like a better rallying cry than “Let’s fix the data,” so sometimes this is a good idea.

Other times, not so much. People may find data-driven less tangible, while the data quality program provides specific to-dos and yields results more quickly. One solution is to look for a middle ground, along the following lines: “We want to become data-driven because doing so will enhance our existing culture. That’s going to take a long time, as we learn exactly what that means and how we each contribute. In the short-term, let’s focus on data quality. It’s a bit more tangible and we’ve already had some big wins in the area.” And in getting in front on data quality, we’ll learn what it means to be data-driven.

Focus the effort, setting stretch targets

Besides getting the right structure and people in place, the most helpful thing senior leaders can do for a data quality program is to create and sustain an urgency around the effort. This translates into narrowing the focus, measuring the quality of newly created data, and on setting demanding goals. As previously mentioned, I find that targets of the form, “Cut the error rate of XYZ in half every six months” highly effective. Depending on your circumstances, you may wish to adjust the time frame. In Liz Kirscher’s case, at Morningstar, this meant halving the error rate every year, not six months. Conversely, halving the error rate every three months may be more appropriate for high-velocity data creation.

Engage Visibly

All senior leaders know that to lead, you must be visible. This can feel uncomfortable when it comes to data, just as talking about anything new and unfamiliar can be.

Perhaps nothing is more damaging to a data quality effort than a senior executive claiming publicly to be a whole-hearted supporter, then doing nothing.

In the last chapter, I advised DQ team leads to push their leaders to engage and provided them a list of ten opportunities. Obviously you need not do all ten. Pick a couple. As your confidence grows, I’m sure you’ll think of plenty more.

The real purpose of this section is to propose that you take on two specifically. The first is this: Call for, then participate on, an improvement project. The next time someone gives you a report with an error in it, don’t dismiss it. Rather, say, “Let’s sort out what happened here, get to the root cause, and see that it doesn’t happen again.” Then be a full participant in the improvement team set up to do so. Two things will happen: First, you’ll learn a lot about data, about becoming a good data customer, and about organizational issues that hinder data quality. Second, (as I’m sure you know) word will get out. Others will want to lead improvement projects also.

A good deal of the benefit in a data policy stems from discussing the issues with senior management teams, sorting out where data and data quality fit, and, in some cases, clarifying the high-level flow of data across the company. The policy itself need do no more than clarify people’s roles as data creators and data customers, e.g., “Don’t take junk data from the last person. And don’t send junk data on to the next person.”

The second is a high-level policy that clarifies the management responsibilities described herein. The policy should be short, perhaps a page, and simply summarize what’s expected of departmental leaders in advancing data quality and of everyone else in their roles as data customers and data creators.

In Summary

Your department and company need high-quality data and it is always in your interest to make sure customers, potential customers, regulators, and others receive high-quality data. Further the data space is exploding in dozens of different directions. What today may be a strategic opportunity could well be a strategic imperative in just a few years and an all-out fight for survival a few more after that.

While provocateurs and DQ teams can achieve real results, data quality efforts plateau unless senior managers (at department and company levels) take responsibility for spreading the effort broadly and deeply. The steps you must take are quite obvious and, in any fair accounting, simpler than the alternatives of dealing with bad data. Still data quality requires a different mindset, new organizational constructs, and a culture shift. This is why you must engage.