According to a McKinsey survey of CEOs, “86 percent of executives say their organizations have been at best only somewhat effective at meeting the primary objective of their data and analytics programs, including more than one-quarter who say they’ve been ineffective.” And 38 percent of these CEOs believe that they are primarily responsible for the data and analytics agenda at their organizations.
Given these statistics, it is high time that CEOs grok data and the issues associated with making data useful.
Why Should You Read Further?
If one thing has become clear in the digital age, it is that “AI is transforming business operating models.”1 An amazing example is digital native, Ant Financial. Ant has more digital users than the biggest US banks yet only a fraction of their staff. Ant Financial uniquely has “the ability to leverage data to learn about users’ needs and respond with digital services to address them.”2 For example, the company has completely automated its loan processes to be digital services with no human involvement.
So why do CEOs today need to understand data and analytics? As this example illustrates, it’s because they have become a universal engine for business execution. In fact, in their book “Competing in the Age of AI,” which I quote several times in this blog, Marco Iansiti and Karim Lakhani suggest that artificial intelligence (AI) — effectively, data + analytics — is reshaping the operational foundations of enterprises for this very reason.
Some Good News
According to CNBC’s Abigail Hess, 40 percent of CEOs have an MBA. This means you should remember things like descriptive statistics including a mean, median, and mode, and maybe you even touched on things like regression analysis and multivariate analysis. This means you already get data and maybe even a little data science.
And while many of the terms have changed as operations research has become data science, the core concepts you learned are still there. The main things that have changed are the variety of data, the amount of data that you have, and the speed at which data is acquired.
Translating What Decision Scientists Are Talking About
To help you pick up the new terminology, here are some common terms and concepts that data scientists and decision scientists are talking about today.
Big Data means the amount of data is larger than traditional data process systems (data warehouses + business intelligence) and requires new technologies. According to analyst Judith Hurwitz, Big Data is determined by three key characteristics: the volume, or amount of data; the velocity (how fast the data is processed); and the variety, or types of data.
Clustering groups, for example, customers into segments by similarity.
Data Mining is about extracting useful knowledge from data to solve business problems.
Data Preparation is the process whereby data is manipulated and converted into a form that yields better results.
Modeling is a simplified representation of reality.
Analytical Modeling estimates or classifies data values by drawing a line through historical data points.
Machine Learning is a collection of predictive modeling approaches.
A Training Data Set is a set of data used to train a computer program how to process data for a decision model.
Structured vs. Unstructured Data — think name and address versus pictures and videos.
Supervised Learning uses a training data set to mine a broader set of data. For example, to determine which customers are likely to churn from customers than have churned.
Unsupervised Learning learns things on its own without the use of a training data set.
Data Problems You May Not Know About
So now we have established that you already know something analytics and data science. Even though the words above may be new to you, hopefully the definitions above helped. And we’re just getting started: This primer aims to get you up to speed on all the issues associated with data.
Data is not pristine and perfect like it probably was in your MBA classes. It’s messy, siloed, and difficult to maintain. But now is the right time to learn how to get the most from your data. As JoAnn Stonier, chief data officer of Mastercard, recently said at a conference, “this is an educational moment for business leaders to understand data quality, fake data, data bias, and how these issues impact decision making.”3
Know Where the Data Is
I know this is going to sound amazing, but most large legacy organizations do not know what data they have — or even who should govern the data and its usage within the organization.
The reasons for this are many. Systems were installed in pieces, and even active data marts were implemented over time to solve discrete problems. Tom Davenport’s research has found that 80 percent of data scientists’ time is spent simply discovering and preparing data. Additionally, less than half of an organization’s structured data is actively used in decision-making. The reality is, there is no card catalog or Google search for your data unless you invest in automation.
Getting Data Where It Needs to Be
Once your team discovers the data that you need for your analytics or digital innovation project, you typically need to move that data from where it lives to somewhere it can be analyzed.
Data tasks are typically intensive and, therefore, need to be done outside of transactional systems. Another reason for executing these tasks in a different location is because the data in these systems often is stored in a way that will not work for analytical analysis. Getting data to where it needs to be includes the functions of understanding data, gathering data, and moving data. This can also involve transforming data so it can be used for analytics.
Data Goodness
In your MBA statistics or Operations Research class, the data you were working with was ready to go. But in business, this is unfortunately never the case.
It could be the way your organization created its transactional systems, or M&A activity that has resulted in multiple disparate data sources. Different systems often describe simple things differently. Data fields and formats for things like region or country can differ between parts of the company.
And it can be even worse for data coming from manufacturing floor equipment. This equipment can produce a number like 70, but lack the descriptive data to say that 70 is a temperature and to which temperature scale it belongs. It lacks data about its own data, or metadata.
Another common business issue is having multiple unintegrated customer records across different systems. In addition to not knowing which system has the correct information, one application may contain certain fields that are missing in another, making it even more difficult to piece together accurate information.
For this reason, it is important to establish data owners and put in place a program to establish data governance and data quality.
Data Security and Privacy
A lot of what you have been exposed to with regards to data protection has to do with effectively creating moats and castles around your data. But just like in the Middle Ages, the castle mentality eventually came to an end. And the same can be said for protecting data.
In an increasingly open ecosystem world, the only way that you can secure data and protect the personally identifiable information it holds, is to protect the data itself and ensure that anyone with the keys to that data can see only a small portion of it. This demands data governance and architecture, which means creating data including privacy policies. It requires what Ann Cavoukian calls privacy by design.
To learn more about how privacy by design can help your organization comply with data privacy regulations while winning customers, read our ebook “Growing Profits and Achieving Compliance With Privacy by Design.“
Getting Data in Shape
To get data in shape, you should start by creating a data pipeline. “A data pipeline gathers, inputs, cleans, integrates, processes, and safeguards data in a systematic; sustainable, and scalable way.”4
As Iansiti and Lakhan explain, “the basic idea behind the data pipeline is to make (data) clean, consistent data available to applications.”5 Problems that get solved are getting data in shape, creating a single view of customer, employee, or supplier, and preparing data for analytics.
Analytical Modeling
Once data is in shape, it is time for modeling. AI and algorithms are about turning the data into value. Historically, that value was in a chart or graphic descriptive analytics. But today, value is generated by being able to generate a prediction from the data. And predictability is predicated on the largest amount of trusted data that a business can have reasonable confidence in and is available for analysis.
While AI can be applied to diverse use cases, the goal is to predict what a user wants, or know when a piece of equipment will need to be replaced. The power of supervised learning, unsupervised learning, reinforced learning, and other AI is to transform business models.
As Alibaba Group’s Ming Zeng says, “Our algorithms can look at transaction data to access how well a business is doing, how competitive its offerings are in a market, whether its partners have high credit ratings, and so on.”6 Simply put, AI can transform how we think of existing business models.
Becoming a Data-centric Leader
As with everything that you may want to accomplish as a CEO, a turning point for transforming your organization into a data leader “starts at the top with motivating and grooming a generation of leaders to do the hard work involved.”7
If you and your team get behind analytics, it is much likelier to bear fruit. Tom Davenport says, “if we had to choose a single factor to determine how analytical an organization will be, it would be leadership.”8
CEOs can also champion investments in technologies that support data democratization – ease of use, ease of access, and bringing in large number of employees across various function to use trusted data as part of their day-to-day jobs.
For the best results, you need to make analytics central to what you do and think. There is no better example than Brian Cornell at Target. A Fortune Magazine article described him as having a thirst for knowledge. For example, Cornell created an ad hoc focus group of Target moms and visited a Target store with them, and says he got great, genuine feedback from them. So, I recommend making this kind of thirst for knowledge part of your operating model as a CEO.
Parting Words
If you were awake during your MBA classes, you know something about analytics. Yes, there are data problems that you may not know about, but these are all learnable things. The important thing is to dig in and ask the right questions. Approach data-gathering like Brian Cornell, with a thirst for knowledge. That will arm you with the information and mindset you need to transform your organization into being data-oriented — the most critical step to winning at digital transformation.
Sources:
1 Competing in the Age of AI, Marco Iansiti and Karim Lakhani, page 12
2 ibid., page 25
3 MIT CDOIQ Conference, August 18, 2020
4 Competing in the Age of AI, Marco Iansiti and Karim Lakhani, page 58
5 ibid., page 72
6 ibid., page 37
7 ibid., page 58
8 Analytics at Work, Thomas Davenport, page 57