The nuance behind data

Who would ever think I would enjoy a book about data? I could think of nothing more boring—until I read “Data, Strategy, Culture & Power: Win with Data-Centric AI by making human nature work for you.” From my work with my clients, I understand there is a lot more to data than meets the eye, particularly now with the onslaught of AI into business models. The old rule “rubbish in, rubbish out” still applies. 

The question

Between 70-90% of all digital transformation initiatives, which include data, analytics, and artificial intelligence (AI) projects, fail. Why isn’t anyone getting it right?

It is a patchwork

Most companies rely on a data ecosystem that grew and evolved accidentally rather than intentionally. Instead of a cohesive, well-architected data infrastructure, this patchwork of databases, data warehouses, and data pipelines cobbled together over time leaves data teams with a fragmented and disorganized data ecosystem that is difficult to navigate, maintain, and scale.

You are broken

Ultimately every system, including the data ecosystem your company works within, is some degree of broken, all the time. You should anticipate fragility. This includes unavoidable failures, imperfect understanding, and fractured communication. The larger a system gets, the more broken stuff there will be at any given time.

Gebru’s law

The book starts with Gebru’s law: All data, analytics, and AI output will encapsulate the power structures and dynamics between the people and groups that originally developed them and the ones that use them, no matter how far in the future that use occurs. And you would be well advised to remember this. An organization with hierarchical or imbalanced power structures, not fully conscious of its own biases, will produce data-driven insights that perpetuate or amplify those power differentials. That means that:

  • Creating trustworthy data-driven systems first requires establishing trust between people.
  • Changes in organizational power dynamics will result in changes in how data, analytics, and AI are produced or used.
  • Understanding an organization’s power structures can help uncover limitations of the data, analytics, and AI it produces and the actions it takes as a result.
  • Power dynamics and organizational factors are just as critical as technical considerations while designing and developing data-driven systems.

Data

The word data, sometimes defined as “pieces of information”, comes from the Latin datum, which means “something that is given.” By giving form or character to data, we bring the given pieces together to generate information. Data does not have to be true or verifiable to be considered “data”.

Empowerment or control

While data, analytics, and artificial intelligence (AI) can be used to inform, empower, and liberate good ideas from the irrationality of power games. They can also just as easily become tools of abuse, coercion, and control. Data can be used to empower and liberate; it can also be used as a weapon for people to gain status, recognition, or domination over others. In these cases, it becomes the poison that slowly, gradually, and stealthily compromises the integrity of (otherwise functional) work environments.

3 data roles

There are typically three common roles: the data producer, the data consumer (which can be a person or a system), and the data user. For any given data, there may be a multitude of consumers and users, and even future users whose intentions we can’t guess.

Data starts with culture

It is the power of the people behind the AI that determines whether we collectively succeed or fail. Here is a checklist (17 questions):

  • What you say, and how often you say it.
  • What, when, and how do you celebrate?
  • The losses and missteps you acknowledge and how you respond to them
  • How you behave when the chips are down 
  • What you fight for at all costs The corners you cut 
  • Who you hire, promote, compensate, and fire
  • Who you “smoke out” until they leave the organization 
  • The worst behaviour you accept and the best behaviour you reject 
  • The voices you amplify, and the voices you suppress When you encourage conformity, and when you promote diversity 
  • How you handle disagreements and differences 
  • How and where you spend your time and money 
  • What gets discussed at the water cooler 
  • What gets discussed out in the open and what is shared behind closed doors 
  • The contradictions you allow vs. the ones you stamp out
  • The exceptions you make vs. things that never budge
  • Who sits together, and how you arrange your office 
  • Who gets access to the best tools, technology, and infrastructure and who has to wait 
  • What you say about your customers during challenging situations

Data is everywhere

It’s everywhere: people collect “impressions” of situations and others around them all the time, regardless of whether this “data” is formally gathered. Data provides the information (and confidence) that helps us formulate and select goal-directed strategies, both personally and professionally. Data is more like electricity: a utility that the business stakeholders who use data expect to turn on and just work.

Data is energy

Data encapsulates a universe of potential energy waiting to be unleashed. The key to awakening this dormant force lies not within the data itself but in the beliefs and inspirations it fuels that compel us to act. By mistakenly focusing on the mythical power of the data itself, it’s easy to lose sight of where the true power lies. In our ability to manage and harness the effects, data has on behaviour. The focus, then, should not be on accumulating data but on gathering, curating, and applying data in ways that will unleash action in the desired directions. 

Everyone needs power. The right kind of power (at the right time) can mean the difference between achievement and failure, between vibrance and despondency, and between supporting others fully – or letting them down.

Data is perception

The book uses time as an example. Building a source of truth for time zones is a continuous, iterative effort, more like the historical reconstruction an archaeologist might work on than a data management project. Who gets to decide exactly which time zone a locale adheres to? Who gets to share and disseminate the information?

Data is archiving

Archival institutions have tremendous power. The power to decide which information is captured and which is ignored (or “lost”). The power to grant or deny access to the data or to limit the time or type of interactions people and systems can have with it. The power to classify and tag information in ways that shape how others think about, use, or interpret it. The power to make finding information easy or difficult The power to effect political control over ideas, even long after the politics have faded

Data is interpretation

“Truth” originates from multiple sources and is shaped by those who interpret it. It emerges over time. When you interact with an archive of data (or any other critical information resource), you’re engaging with its entire political history. This includes how the data was created, the backgrounds and biases of the people who originally created it, the people who currently act as its guardians, and the power those people or institutions have been formally or informally incentivized to uphold. Truth is subjective, shaped by multiple sources and interpretations, and always influenced by political and social factors.

Clean data

Once we understand how much imperfection our organization is willing to accept based on our risk appetite, we can make better decisions about how much time and effort to invest in cleaning data and making meaning from it. Data can provide insight or be used to mislead. When looking at data, you need to think critically about how you could be misled.

Data is deviance

Collecting and tracking data can help keep you honest about when and where deviations occur. Selective enforcement is much more difficult when reliable data is available and visible. Second, recorded data can help to compel the actions that pull people back into compliance. To be data-driven means to take appropriate action based on the evidence presented in each situation.

Data and integrity

When a work environment is characterized by unreasonable demands, unquestioning loyalty, or the continual fear of punishment, data and data operations suffer. After all, just matching a leader’s expectation of what data should show is far less painful than taking a stand for accuracy and integrity. Data integrity suffers when people can be punished for what it reveals. Survival wins every time. To maintain the integrity of data and AI systems, it is crucial to have robust accountability measures in place, including independent oversight bodies, external audits, and clear paths for whistleblowing and reporting concerns.

The eight data delusions

  1. The Delusion of Linear Causation. This refers to the erroneous belief that risk factors and events align in a simple, deterministic sequence that results in a specific outcome.
  2. The Delusion of Compliance. This is the false belief that strict adherence to procedures and regulations inherently ensures quality, safety, or any other characteristic they are in place to promote.
  3. The Delusion of Consistency. This is the mistaken belief that uniformity and standardization in management practices will always lead to better control of risks and hazards, protecting against failures.
  4. The Delusion of Risk Control. This is the erroneous belief that comprehensive rules, procedures, and work systems can entirely eliminate or control risks. The remedy is to replace certainty with a healthy regard for complexity.
  5. The Delusion of Human Error. This represents a flawed perspective that attributes the majority of failures to human mistakes, and argues that these can be eliminated through behavioral modification (otherwise known as “more training”).
  6. The Delusion of Quantification. This is the belief that performance can be adequately represented through metrics at all. The remedy is to understand and embrace variability.
  7. The Delusion of Invulnerability. This is the mindset or culture in which a person or an organization is either ignorant of—or perceives itself to be immune to—accidents, errors, or failures.
  8. There’s one final delusion that’s rarely called out: that all you need is to “hire good people” and then “trust your team.” These are dangerous delusions that shift the burden of success away from the work system and its managers to the people who have to work within that system.

Rules are time travel

Every rule exists because something bad, unexpected, or undesirable happened in the past. Rather than feeling dismayed or overwhelmed by standards or regulations, think of them as time travel: they are messages containing critical data information from your predecessors. Every process failure shines the light on an opportunity for improvement around data or analytics, meaning that data and analytics can help you debug your business.

Why go digital

Digital technology has immense potential to enhance the performance of any business process, particularly safety and emergency preparedness. Immediate feedback like this can lead to improved performance and increased adherence to safety practices and principles while helping workers develop a sensitivity to know when to take an intelligent risk and deviate. Context-aware tools and artificial intelligence can offer real-time monitoring and control of workflows at remote sites, mimicking the advancements seen in onshore manufacturing facilities. Leveraging intelligent and automated systems not only enhances safety but can also boost productivity and efficiency.

The key question

An effective data strategy to support any critical process must ask one key question. How can we ensure that data and information of the appropriate quality get to the people and systems that need it in time for them to make the decisions that will protect the integrity of a process? Withholding data and information and failing to detect and act on early warnings can lead to missed opportunities for improvement that lead to injuries and death. Delusion emerges in the whitespace of how people relate to each other in groups and breeds confidence because it feels so certain.

Data integrity is process integrity

Data integrity is impossible without process integrity. Process integrity requires collaboration and alignment across multiple functional units of a business and between key people therein. Collaboration requires connection. Process integrity requires alignment. This means data integrity is a team sport. Lack of alignment negatively impacts outcomes, professional relationships, and, ultimately, the bottom line. It can emerge between any two people or process steps. Regardless of how skilled or experienced you personally might become, there is no workplace where everyone will be educated enough, competent enough, informed enough, and emotionally healthy enough – all the time – for interactions to be completely productive and frictionless. There is no utopian work environment.

Some other lessons

  • Having a little bit of data is not the same problem and is often not as significant a problem as having bad data in any volume.
  • Make sure people interpreting data use the same terminology to understand that data.
  • Trusting data too much can have disastrous consequences.
  • The organization is responsible for establishing standards and guidelines that align with the users’ particular needs and objectives and ensuring that data quality is evaluated from multiple perspectives and in accordance with criteria relevant to those perspectives.
  • Outliers are data elements that deviate significantly from other related elements. While outliers may seem like errors or noise that should be eliminated at first glance, they sometimes carry important information about the observed physical phenomena. Not all outliers are informative; some may indeed be due to instrument errors, human errors, system errors, or 
  • Not all data is created equally, and the quality of every data-driven decision depends on data being clean enough and meaningful enough.
  • People use software because they’re motivated to get their own work done rather than by a desire to store accurate, complete, consistent, and valid data that will grow in value as an asset for their organization over time.
  • Acquiring data is good. Verifying that your data is good is much better.
  • One of the most insidious root causes of bad data is people.

Being data-driven

You want to be data-driven. You want your organization to be data-driven. While it’s rare to encounter someone who will argue against becoming data-driven, it’s just as rare to find a well-articulated strategy that explains exactly what “data-driven” means (and how you’ll know when you achieve it). A truly data-driven organization values and honours evidence follows processes to interpret the evidence from multiple perspectives within a current context and, based on that interpretation, determines whether an action to realize change should be taken now, later, or never. The premise of data-driven is that better outcomes result when decisions are informed by reliable and relevant evidence rather than by assumptions, traditions, or anecdotal experiences.

Factors

There are many factors to consider: governance, management, strategy, third-party suppliers, chain of custody, the data journey, relevance, triangulation, variation, risk lens, power differentials, ambiguity, obfuscation, clarity, data lineage, semantics, jargon, data entropy, chaos, data collection design, completeness, and data literacy.

Proxy and synthetic data

At the end of the book, I got uncomfortable. Covering topics such as proxy data and synthetic data. One of the biggest threats to the validity of the information you collect is the risk of not actually measuring or assessing the thing you think you are measuring. 

Data is inspiration

Einstein told us that energy and matter are one and the same, and one can be converted into the other. Data catalyzes belief, and belief releases the energy of inspiration. Data (and, by extension, AI) can be powerful tools for growth.

Cracking book

Cracking book. I learned a lot, particularly about the nuances and factors that impact data. If data is of interest, you should also read:

sensemaking cover

WHY REINVENT THE WHEEL AND WHY NOT LEARN FROM THE BEST BUSINESS THINKERS? AND WHY NOT USE THAT AS A PLATFORM TO MAKE BETTER BUSINESS DECISIONS? ALONE OR AS A TEAM.

Sense making; morality, humanity, leadership and slow flow. A book about the 14 books about the impact and implications of technology on business and humanity.

Ron Immink

I help companies by developing an inspiring and clear future perspective, which creates better business models, higher productivity, more profit and a higher valuation. Best-selling author, speaker, writer.

Leave a Comment

Your email address will not be published. Required fields are marked *

Scroll to Top
× How can I help you?
0 Shares
Share
Share
WhatsApp
Email