What to Read Next
In my career, I’ve spent more than 25 years helping companies with their data and data quality programs. Beginning in 2018, I undertook a broad-based research effort to understand why, so many years into the digital revolution, progress in the data space is so slow. This article synthesizes a review of my clients’ and others’ successes and failures, discussions with dozens of experts in data and analytics, and study groups that dived deeply into various aspects of the topic.
The headline result: Today’s organizations are unfit for data. Until companies address the underlying issues, progress will remain halting and uncertain.
Email Updates on AI, Data, & Machine Learning
Get monthly email updates on how artificial intelligence and big data are affecting the development and execution of strategy in organizations.
Please enter a valid email address
Thank you for signing up
To be clear, the many successes I’ve observed — in data science, analytics, artificial intelligence, data quality, and other ways to put data to work — confirm that the approaches, methods, and technologies around data work just fine. But most companies still struggle. Organizational structures, people issues, a lack of accountability, and other traps get in the way. This conclusion has far-reaching consequences, and leaders should not expect faster progress until they make some real changes. To do so, they must first understand and address the underlying issues that hinder progress.
To better understand both driving forces (those pushing progress forward) and restraining forces (those holding back progress) that are impacting data science progress, I’ve used force field analysis (FFA), a powerful visual tool derived from Kurt Lewin’s change management model. A companion toolkit provides information on how to build out your own FFA. (See “Visualizing Change With Force Field Analysis.”)
Five Areas Key to Success
“Data” is a very broad space, so for this analysis, we will look at five areas of critical importance to companies: data quality, putting data to work, organizational capability, technology, and defense. Failure to deliver in any of these areas can scuttle an otherwise terrific data program:
- Data quality: Poor-quality data adds incredible cost and friction.
- Putting data to work: Unless companies put data to work in ways that return value, there is little business benefit. Ways to do so include data science (including AI and machine learning), exploiting proprietary data, creating a data-driven culture, monetizing data by selling it or building it into products and services, and treating data as an asset.
- Organizational capability: This refers to the people, structure, and culture within the organization that support data programs. For example, silos can get in the way of data sharing.
- Technology: Technological infrastructure will be different for each company, but it will be difficult for companies to scale their data programs without the right tools and technologies in place.
- Defense: This category encompasses all of the organizational tasks related to minimizing risk, including security, privacy, and ethics.
In the following sections, we will explore each category in more depth using FFA.
Based on my work with Tadhg Nagle and David Sammon, only 3% of companies’ data meets even the most basic standards. The cost is enormous, not just from a financial perspective — think about throwing out 20% of company revenue per year — but also from a culture perspective, as about only 1 in 6 managers trusts the data they use every day. As the analysis shows, there are solid driving forces across external, organizational, and technical categories that can help companies improve data quality, but unfortunately, many senior leaders are unaware of which restraining forces have the greatest effect. (See “Forces Impacting Data Quality in Organizations.”) An effective way to begin to turn this around is to seek out the root causes of the restraining forces for data quality in one’s company.
As the chart depicts, there are many issues at the organizational level that require focus. Quality improves rapidly when all those who use data assume responsibilities as data creators (that is, they create data that will be used by others) and data customers (that is, they use data created by others). Unfortunately, most people don’t know what they must do to fulfill these roles properly, so educating them is essential.
The root causes of many data quality issues experienced in one business unit may begin in another, but silos make them difficult to address. Other issues include confusion between the data and tech domains, as leaders in the company may think data quality is the province of their IT department.
As these forces build up and compound, they put other strategic priorities for the company at risk. As we’ll discuss later in this article, when it comes to the issue of putting data to work, failure to address restraining forces that result in bad data quality can severely hamper AI programs and digital transformation.
If companies hope to build a future in data, they must attack data quality as a priority.
Putting Data to Work
There are plenty of compelling data science success stories, and data science is enjoying a certain amount of momentum: The number of qualified data scientists is growing, major tech companies are demonstrating what is possible, and other companies are excited by the hype and promise of AI. But the failures vastly outnumber the successes. (See “Forces Impacting Putting Data to Work in Organizations.”) There is a built-in structural animosity between data science teams (“the lab”), which are invested in driving change, and the rest of the business (“the factory”), which is interested in promoting stability. And the hype makes data science appear to be far easier than it is. As Thomas H. Davenport and I have observed, if it costs a dollar to develop a model, then it costs approximately $100 to deploy it — and companies simply have not made that investment.
Further, data science is only one way to put data to work. “Small data” and proprietary data both offer exciting opportunities, but most organizations do not even entertain them. Exacerbating this, data is not on corporate balance sheets or income statements, closing off a potential avenue to quantify results.
A company’s organization is supposed to make it easier for people to do their work. But organizational capability can be a major issue when it comes to data, as the sheer number of restraining forces in this category underscores. (See “Forces Impacting Organizational Capability of Data Programs.”)
I’ve already cited three restraining forces: management’s confusion and conflation of data and technology, data silos, and the lack of understanding around roles. Four further restraining forces merit special mention. First, the lack of skilled data architects, engineers, and quality professionals makes it more difficult to address data quality issues.
Second, although many companies make statements about valuing data, the actions within companies often reveal that this purported value is fairly hollow. Many leaders opine that data is the new oil, preach that data is an asset, or urge their people to make data-driven decisions. But the reality on the ground is quite different — for most people, data is just another thing they need in order to do their jobs. Companies don’t sort out ways to profit from their data or take care of it, and they don’t teach people how to use it to make better decisions.
Exacerbating these restraining forces is fear. Even as individuals don’t value data per se, they see and hear all of the hype, and they can ascertain that massive change is on the way. People are smart to be scared — that they will lose their jobs, be left behind, lose status, and so forth. But fear freezes people, and companies run the risk of losing talented employees to companies with stronger data cultures. Smart managers don’t let fear fester. They acknowledge it and then help their teams develop skills that will better position them to benefit as changes occur.
The changes needed to address the people, structure, culture, and fear issues cited throughout this article are massive, and they are the responsibility of a company’s most senior leadership. Yet most leaders appear to be sitting on the sidelines, perhaps fearful themselves or unsure of what to do. Senior leaders simply must begin to get a feel for these issues, move all things data related up on their list of priorities, and step up for their teams.
Many new technologies, including AI, cloud computing, and connected devices (the internet of things) have proved that they can be valuable. Still, implementing new technologies is not easy for most companies, because powerful restraining forces get in the way. (See “Forces Impacting Technology Implementation in Organizations.”) Technical debt in various guises (such as difficulty replacing existing infrastructure) is a major issue. So are the low levels of data quality cited above. Poor data quality is particularly worrisome when it comes to implementing AI, given that its quality standards are extremely high and the damage done by bad data may go unnoticed.
Another serious issue in the technology category involves poor relations between technology and business teams. A lack of common language and trust between these groups can impede progress significantly. Many tech professionals report that they feel overworked and underappreciated, making it difficult for companies to get the most out of new technology initiatives.
Perhaps the most serious issue involves poor relations between business and technology teams. Many businesspeople readily admit that they don’t trust their IT counterparts. It is difficult to see how companies can take full advantage of the technologies available to them under these circumstances.
The past several years have seen a spate of security and privacy legislation. While the European Union’s General Data Protection Regulation (GDPR) and other regulations appear threatening, and some rather spectacular fines have made the news (such as Citigroup’s $400 million penalty levied by the U.S. Office of Comptroller of the Currency, and Google’s $56 million fine under the GDPR), by and large enforcement has been weak. Capital markets and consumers are proving forgiving. For example, Facebook appeared to be in deep trouble after the 2016 Cambridge Analytica scandal, and likewise Equifax after its 2017 data breach. Facebook paid a small fine, but its share prices continue to grow unabated. Equifax paid a $1.38 billion fine in 2020, but $1 billion of that was applied to security upgrades that it probably had to make anyway. In the U.S., unless your offenses are egregious, you probably have little to worry about in the near term. The situation may be more nuanced in Europe and other jurisdictions, where citizens place a higher value on privacy.
Still, companies should not sit idle. Data piracy and malware are on the rise. And consider the restraining force “Customers are beginning to assert themselves.” A growing segment of desirable consumers appears to be linking a company’s privacy practices with its brand, and they are taking their business elsewhere if those policies offend. While not massive forces currently, both could grow, creating opportunities for organizations that take heed, and risk for those that do not.
Issues Extend Beyond Culture to People and Structure
The implications for all those interested in advancing data are profound. Culture is often cited as the biggest barrier to progress with data. My analysis confirms that culture is a leading factor holding companies back, but issues related to people and organizational structure are also quite powerful.
When it comes to people and talent, while many companies have made the connection between AI and hiring data scientists, there is still an alarming lack of data talent at all levels. Elsewhere in the organization, few people and teams are aware of their role in data quality and lack the ability to participate in smaller data projects. This breeds data illiteracy and fear of change in organizations. But if data is to be truly transformational, companies need to get everyone involved.
To summarize, the most important takeaways include the following:
- Progress in the data space continues to be held back because organizations are unfit for data.
- There is plenty of good news: Most companies and leaders see the value of data science and data quality, and there are useful related technologies to adopt. The number of solid data scientists is growing.
- Regular people are missing from data programs, slowing everything down.
- Poor-quality data has a suffocating effect on day-to-day work, data science and monetization, and new-technology implementation.
- Organizational silos get in the way. They hinder data quality and interfere with data sharing at all levels. There are considerable tensions between data science and business teams. Technology and business teams can’t talk to each other due to the lack of common language.
- Most individuals and companies confuse the management of data and the management of technology, hindering proper management of both.
- Cultures don’t value data and data science (even as many say they do). Instead, there is considerable fear of both.
- While companies lack needed talent at all levels, the most important gap is at senior levels. Senior business managers have yet to engage, possibly because they don’t know how to be effective leaders in this space.
- Even as there are a growing number of data breaches and concern for privacy grows, the investment community and general public have yet to seriously punish companies for breaches and privacy violations. Still, uncertainty rules, and companies should remain vigilant for changing customer sentiment and regulation.
The steps needed to build better organizations for data are likely very different in each company. Leaders can benefit from conducting their own FFAs to help determine the best areas of focus. (See “Visualizing Change With Force Field Analysis.”) Keep in mind that details matter. Be specific in capturing the different factors at play — perhaps a couple of very visible failures have made managers overly cautious, or a new chief data officer is better connected to senior leadership. Capturing these details will help ensure trust in the analyses and build support to tackle the areas that need the most help.