Chapter 9

Developing Your Business Acumen

IN THIS CHAPTER

Differentiating business acumen from subject matter expertise

Mapping out how data science professionals increase business profits

Increasing your business acumen

Digging into best practices for data science documentation

Data science is somewhat of a cross between software engineering and business consulting: You have to know enough code to do ad hoc analysis and build accurate predictive models, and you also have to know enough about the business to understand the context in which your work creates value for the company. That combination requires you to have coding skills, statistics chops, and, last but not least, business savvy. In this chapter, you’ll see the extent to which you need business acumen to be successful working in the data science field — and what you can do to bolster your business knowledge in record time.

Bridging the Business Gap

Not all data scientists work inside businesses. Some are in academia and nonprofits, and some work in scientific companies where business know-how simply isn’t relevant to their work. Let me start this chapter by taking a look at what business acumen is and letting you know when it's especially important for data science professionals.

Contrasting business acumen with subject matter expertise

Over in Chapter 1, I say that — in order to practice data science effectively — you need the analytical know-how of math and statistics, the coding skills necessary to work with raw data, and an area of subject matter expertise. That’s actually the first definition of the data scientist role, which was espoused by Drew Conway back in 2013, in his infamous data science Venn diagram. (You can see a re-creation of this diagram in Figure 9-1.)

FIGURE 9-1: The data science Venn diagram.

So that’s data science in a nutshell, but this reference to subject matter expertise is rather vague, is it not? I mean, what if you work as a data scientist in the Greenpeace Tech Lab? Imagine how different your subject matter expertise would be if you were working for an environmental nonprofit rather than as a data scientist for an investment bank like Barclays. The difference in requisite subject matter expertise between these two positions is almost indescribable.

Consequently, I need to tighten up this language a bit by defining the type of subject matter expertise I’m talking about. Luckily for you, however, if I’m talking about data science that’s being used to support a business, the term subject matter expertise is just a placeholder for business acumen — the keen ability to understand business dealings in terms of risks and opportunities to a company when it comes to protecting and increasing its profits and adhering to its business mission.

I scope out the definition of business acumen in greater detail in the next section.

Defining business acumen

Short of a keen ability to understand business dealings, what does it really look like for a data professional to have strong business acumen? One clear way to answer this question is to define the characteristics that this data professional would exhibit, such as

• Executive mentality: When I talk about executive mentality, what I’m referring to is your ability to think on a business big-picture level. In other words, you're able to see how all integral processes and components in a business function together to generate business profits and fulfill your company's mission.
• Financial savvy: Financially literate professionals understand the core drivers of profit, growth, and cash flow across a business. They understand how to not only interpret these financial statements but also formulate strategy and take decisive action in order to boost a business's bottom line.
• Leadership skills: Leadership skills involve product management or project management skills, as well as an exceptional ability to work with teams and stakeholders to deliver profit-forming projects or products that are within scope, on time and, hopefully, under budget.

If you're like most data professionals. you're probably feeling like a fish out of water while reading about the characteristics of a truly business-savvy data professional. Not to worry, though: The extent to which you need business acumen depends heavily on your role and function within your company.

For example, if you’re a data science project manager or product manager, you can bet your paycheck that you need strong competency in the three areas I list in this section. If you’re a data entrepreneur and your company stays in business for any extended period, chances are good that you will be forced to develop these competencies as a byproduct of the work you do every day to grow your company.

On the other hand, if you find yourself called deeper and deeper into data implementation work — and you love it — you’re probably a pure data implementation person. One core benefit of data implementation roles is that they don’t require many people skills, leadership skills, business acumen, and the rest. If you’re a data implementation person and you can cover the documentation requirements for data implementers — discussed later in this chapter — that’s about all the business savvy you should need. Next, let’s take a closer look at how different roles within the data space function to support a business in reaching its goals.

Traversing the Business Landscape

The first thing you need to realize when it comes to the business world is that the core function of any business is to generate a profit. Within corporate environments, employees are heavily discouraged from discussing how much money they earn. At many companies, discussing your salary can be grounds for termination. So, it's no wonder that data professionals aren’t naturally attuned to concepts and mechanics related to making money for a business.

The goal and primary purpose of any business is to generate profits.

Data is used to generate profit for a company in one of two main ways:

• Revenue-generation: The idea here is to build and monetize new products and services.
• Cost savings: Here, the emphasis is on increasing profit margins by improving efficiency or decreasing risk.

In the next section, I show you exactly how various data roles support businesses in turning a profit.

Seeing how data roles support the business in making money

When discussing how data roles support a business to make money, I limit my discussion to a few primary roles. I discuss the data scientist role (of course) as well as the data analyst role, and I look at the role of data project manager and data product manager. And I need to help you consider the business analyst role as well as the role of data engineer and machine learning engineer. Figure 9-2 diagrams how each of these data roles supports the others and supports the business in generating a profit.

When considering each of these data roles, note that — although all these roles are data-intensive — they act on different levels within a business and support different types of actors within the business.

Data product managers manage products that directly generate revenues. Data project managers sometimes manage projects that result in cost savings, and at other times they manage data services that generate revenues directly. Data implementation people — like data analysts, data scientists, data engineers, and machine learning engineers — do the coding, building, and sophisticated machine learning work that’s required in order to create the products and complete the projects led and managed by these data project and product managers. Business analysts use data analysis skills and strong business acumen to define the business’s needs and support those internal projects that have been launched in order to increase profits by saving costs within the business’s internal operations.

FIGURE 9-2: The mechanics by which traditional data roles increase business profits.

Because the business analyst is fundamentally closest to the pure business roles — roles like those occupied by business managers and executives — I’ll ask you to consider this role first. Business analysts use existing business data and strong communication skills to fulfill a cross-collaborative role between teams. It’s a people-oriented role, and business analysts are expected to have solid skills in persuasion. Business analysts collaborate between the business and technology, but sit firmly on the business-side. They scope the internal needs of the business, define requirements, and work to ensure that the technology portion of the business is doing the things that are needed to support the business. Business analysts gather requirements for internal business projects — in other words, projects that support and improve the internal operations of the business. The role involves using existing data to analyze processes, systems, organizational units, and overall problems throughout a business in order to help create effective business solutions to those problems. Business analysts support project managers and product managers as well as business managers and executives.

Data project managers work to manage and support the delivery of data implementation projects, whereas data product managers act as mini-CEOs of data products that are owned and sold by the company. Both of these roles function on the business side, but they interface directly with the technology side. People in these roles work to ensure timely and accurate delivery of their projects or products, and it’s through those products and projects that people in these roles boost a business’s bottom line.

Although these roles have lots of similarities between them, they’re fundamentally different. In terms of similarities, both product managers and project managers support data implementation teams in getting their needs met as they build out the respective data science product or project. Both of these roles are responsible for working closely with business managers and executives. Data product managers, however, manage products that are for sale and consumption directly by the business’s customers. They generally don’t work to support the internal business needs of the company. In contrast, data science project managers often deliver projects that support the internal operations of a business as well as customer-facing data science services. In the context of my discussion of the data superhero archetype over in Chapter 1, both data science product managers and data science project managers would be considered data leaders.

Both data scientists and data analysts are on the technology side of the business, but they are far closer to the business than professionals who work in pure engineering or software development roles. Their work supports data product managers as well as data project managers, and they often provide ad hoc analytical support as needed for various personnel on the business side. These folks help a company generate profits by successfully implementing the requirements that are laid out for them, and that are managed by their data project or data product managers. Though both are data implementation roles, data analysts tend to do more business consulting work, whereas strict data scientists often spend most of their time focusing on the data. You need business acumen to work as either a data scientist or a data analyst, but you don’t need it to the same degree as a data project manager, product manager, or business analyst. That said, data analysts, and especially data scientists, need quite a bit more technical chops than their more business focused counterparts. Data scientists and data analysts are both data implementation roles.

One step deeper into the technology side of the business lie the machine learning engineer and data engineer. (I discuss both roles in greater detail in Chapter 2.) People in these roles are almost all purely technical data implementers. There’s little need for business acumen when you’re doing the in-depth, detailed work of coding up decision engines or building data systems. Those aspects of your product or project dealing with requirements gathering or consulting are generally handed to you by a data project manager, data product manager, or business analyst. Despite this low bar for business acumen, data and machine learning engineers are still required to produce documentation to support the systems and products they build for the business. I help you dig into those requirements later in this chapter.

Leveling up your business acumen

Chances are, if you’re an aspiring data science leader or entrepreneur, you’re curious about the actions you can take to beef up your business acumen. The good news is that you don’t need to go out and get an MBA to make this happen. The bad news is that, like all great things in life, it takes some work.

To increase your business acumen, you first need to start by understanding the levels that make up this thing called business acumen. Don’t worry! It's a piece of cake — a 3-layer cake, to be exact. Figure 9-3 shows those three layers of the business acumen cake.

FIGURE 9-3: The three layers of the business acumen cake.

On Level 1, at the base of the business acumen cake, you need to develop general business expertise. (This is on par with the executive-level mentality I talk about earlier in this chapter.) Basically, you’ll want to understand the mechanics by which various business units function to improve a business’s bottom line. Luckily, the purpose of this entire section of the book is to show you exactly how data science impacts business functions to improve business profits. In Chapter 10, you can see how data science projects are used to improve operations, and in Chapter 11, you can see how data science supports improved returns from a company’s marketing activities. In Chapter 12, I talk about the decision-support function, and Chapter 13 is all about finance improvements. (This last bit of expertise is transferable across industries and companies you support, so you should make it a priority.)

On Level 2, in the middle of the cake, you want to develop business expertise that’s relevant to your industry. Throughout the remainder of this section, you can see examples of how data science is improving profitability for businesses in a wide variety of industries — from food and beverage to software and everything in between. Industry-level expertise is extremely valuable because, as long as you stay working within the same industry, the expertise itself is transferable between companies. If you hop into a new industry, though, be forewarned that most of this expertise is longer relevant.

At the top of the cake, on Level 3, you have your company-specific expertise. This knowledge is critical to the success of data projects at the company, but has little value to you outside the company if you want to switch jobs. In fact, this company-specific knowledge is usually covered under a nondisclosure agreement (NDA) as proprietary intellectual property that cannot be disclosed to people outside of the company. In Chapter 15, I talk more about how to collect information about your company. That information you collect can go a long way in satisfying the company-specific expertise you need to lead successful data science projects.

Fortifying your leadership skills

Truth be told, I could write an entire book on the topic of leadership in data science, but if I had to summarize my tips in one short bullet-point list, I’d recommend the following:

• Invest in relationship: If you’re a data leader, you’re in the business of relationships. You’re responsible for cultivating and maintaining meaningful relationships with business leaders, project stakeholders, and data team members. Take it seriously. Additionally, don’t neglect the broader data science community. Join a local data professional organization and network with other data science leaders by lending your expertise.

  Over on https://businessgrowth.ai/, I give you some communication best practices as well as an email template to help you manage your relationships with project stakeholders. Also, I show you my Mini-Black Book of Data Professional Organizations, which lists some groups you may want to consider joining.

• Stay educated on new data use cases and case studies: Never jump into data implementation head-first. Before each new project, take the time to get strategic by exploring and evaluating the latest relevant data science use cases and case studies. (I discuss how to do this in Chapter 16.)
• Cultivate a data-enthusiastic culture: Even if your data science project is managed and implemented to perfection, if your company’s culture doesn’t inspire and excite employees to want to do their jobs on a more data-informed basis, your project is likely to face user-adoption challenges. Though you will want to make sure you’ve created a top-down enforcement approach, you can supplement it with a bottom-up corporate culture that helps inspire employees to want to become more data-empowered by using your data science solution.
• Evangelize for your company’s data project and teams: Similar to the bottom-up approach I just mentioned, when you evangelize your company’s data projects, you help spread data literacy and enthusiasm across its workforce. If you’re serious about your data projects having a big impact on the company's bottom line, be the evangelist for your team and its data science projects.

Surveying Use Cases and Case Studies

It’s time that I introduce you to my 4-step framework for initiating and maintaining profit-forming data science projects. I came up with this framework, the STAR framework, after many years traveling the globe and helping companies plan and kick off their own successful, profit-forming data science initiatives. The STAR framework is shown in Figure 9-4.

FIGURE 9-4: My STAR framework, for managing profit-forming data science projects.

The STAR acronym is created from these terms:

• Survey: You need to know what’s out there in terms of successful data science use cases and case studies. Part 3 is your opportunities to practice surveying the data science use case landscape.
• Take stock: The second phase of the STAR framework is on gathering important information about your company. (You find out how to do that in Chapter 15.)
• Assess: After you take stock of your company, you enter the third phase of the STAR framework, where you access your company’s current state. (That topic is covered in Chapters 16.)
• Recommend: The final phase of the STAR framework is where you recommend a plan for using data science to generate a new or improved profit for your company. (I cover that aspect in Chapter 17.)

This is the part of the chapter where I sometimes mention Big Ideas, so let me throw another conceptual linchpin your way — it may prove helpful as you move forward. It’s the five main routes by which data science impacts business, as illustrated in Figure 9-5.

FIGURE 9-5: Five routes by which data science impacts business.

Building on the cake metaphor in Figure 9-3, Figure 9-5 summarizes the five main routes through which data science impacts (and improves) a business's bottom line. The following list points you to the chapters of this book that cover each topic:

• Operational improvements: Chapter 10
• Marketing improvements: Chapter 11
• Decision-support improvements: Chapter 12
• Finance improvements: Chapter 13
• Direct data monetization: Chapter 14

To boost a profit-forming data science project off the ground, you need these three key ingredients:

• Data science skills and expertise: You need data science professionals to both manage and implement your project.
• Data technologies: You need data storage and processing tools for your data professionals to build out and maintain the solution.
• Data resources: Last but not least, you need actual data that can be used to build out your predictive solutions.

Before jumping full-throttle into the STAR framework, however, you definitely need to get a few more business basics under your belt so that the rest of the book makes sense to you. Simply put, no matter what type of data science professional you are, when contemplating a new data science project, you should always start by reviewing the latest data use cases and case studies. And, if you're a data implementation person, go the extra mile to make sure you’re baking business value into your data science code by documenting it properly. In this section, you dive deeper into data science use cases, case studies, and coding documentation best practices.

Though the documentation requirements for data leaders are different from those for data implementers, the documentation itself supports the same goal — to protect and preserve your company’s return on investment (ROI) into its data operations. Let’s take a look at what type of documentation you need to prepare and how that documentation works to protect your company’s ROI.

The goal of documentation is to protect and preserve your company’s return on investment into its data operations.

Documentation for data leaders

Familiarity with data use cases and case studies is imperative for a data science leader, but if you want to be a well-rounded data science implementer, you simply have to take the time to stay up-to-speed with data use cases and case studies as well. The importance of use cases and case study evaluation comes down to feasibility. Alas, the data science landscape is awash with shiny objects that may or may not work well for your company. The key to your success as a data leader is to select the most promising data science use case for your company, given its current capabilities with respect to data resources, technology, and skill sets. Chapter 16 goes into great detail about how to go about doing that. For now, what you need to know is that, if you choose a data science use case that turns out not to be feasible, your company is highly likely to invest at least tens of thousands of dollars in deploying a use case that fails and ends up generating nothing but a massive waste of time and money. Step 1: Make sure the data science use case you choose is feasible, by taking the time to research current data science use cases and case studies. This research can help shed light on which use cases may offer promising results for your company.

If you attempt to execute a data science use case that's not feasible for your specific situation, your project will probably fail and generate nothing but costly losses for your company.

What is a use case? Think of it as a recipe or set of instructions on how to build something — a set of instructions designed to benefit the business in a clear and specific way. A use case includes a list of sequential actions that need to take place in order for a system to operate properly. Technical use cases also should come with basic technology specifications: details regarding which types of technologies are required in order to implement the use case. Because use cases are written in plain language that anyone can understand, they can be extremely helpful when it comes to establishing clear communications with stakeholders as well as developers and managers.

Let's take a deeper look into the elements that can be included in a thorough data use case:

• Use case title: The title should be descriptive of exactly how the use case benefits the business.
• Description: The description should consist of one or two sentences that describe what the use case does and the benefit it renders.
• Actor: An actor can be a person interacting with the system or the system itself. An actor is essentially an agent who takes action to support an outcome from the system. With respect to actors, you should document who the primary actors are as well as any supporting actors and offstage actors.
• Pre- and post-conditions: Preconditions document the things that must be true in order for the system to work, and post conditions summarize the output of the system after it's built and running successfully.
• Main success scenario: The main success scenario should include actor intentions as well as a clear statement of the success scenario. (Actor intentions are a series of sequential steps that the actors take in order to operate the system; the success scenario is simply the output capability of the system after it's built and running properly.)
• Industries and functions: A comprehensive use case needs to document both the industries and business functions for which they are relevant.
• Use case diagram: The business use case diagram is a visual depiction of how the actors interact with this system in a series of sequential steps in order for the system to work properly and achieve its outcome goal. (Check out Figure 9-6 for an example of a use case diagram.)
• Technology specifications: For data use cases, also include information about the technology that's required in order to make these systems work as well as the data science methodologies of any relevant vendors and any integrations that this system offers.

FIGURE 9-6: A simple business use case diagram.

I’ve created a reusable data use case template to help you build out your own data use case collection. It’s over on the companion website, at https://businessgrowth.ai/, if you’d like to use it.

Case studies, the companion piece to use cases, are designed to function as written narratives that describe the use case in story format. A case study should document the problem the business was facing, how it went about solving that problem, the solution that was implemented, and the outcome the business saw. An important aspect of case studies is that they should be engaging narratives that demonstrate the power of the use case without requiring readers to have any technical expertise. Case studies should be powerful and persuasive communication tools when you're trying to sell your data science project idea to executives and business leaders. (I provide you with tons of examples of case studies and data science use cases in Part 3.)

Documentation for data implementers

Earlier in this chapter, I talk about the documentation that’s most relevant to data science leaders and consultants. I still have lots to discuss, however, about how data implementation professionals can use documentation to create more value for the business.

All great data implementation professionals know the value of coding documentation: written and illustrative documentation that describes how your code works and what needs to be true in order for it to work successfully. Let's start by discussing why coding documentation is necessary in terms of business value. First off, you might need to pick back up on what you're working on in a year from now and, if you have good documentation, you won't need to spend tons of time retracing your steps to figure out why you made the decisions you made. That time is money for your company, so in this way, good documentation can save you a ton of money.

Additionally, when you use data science code to produce a product, that product is actually owned by your company. When you eventually leave the company, you need to be able to hand it over to others in a useful way and without a lot of downtime. Your coding documentation helps the next person in line understand how the product works, and how to use it properly. This protects the investment of resources that the company made in having you build that product, because it ensures that you’re leaving behind a working prototype that can be used with or without you. And it ensures that the knowledge required to use that prototype doesn’t get lost. Now let's look at the two types of details you should include in your data science documentation:

• Details regarding the data itself: Document details about the data’s source, its variable types, any outliers and missing value treatments, the data reformatting and clean-up that’s required, and variable relationships within the data.
• Details regarding the data science: Describe the business question that your model solves and how it goes about answering that question. Describe the models as well as their performance metrics. Include details on the hyperparameters (the parameters you set to tune model performance) and how they perform, as well as any feature engineering you did when building the model. Lastly, describe the final model you selected, and why you selected it.

No need to get too fancy with your documentation tool. You can easily document your data science code using a Google Doc or Word document. Other robust and easy options include Notion and GitHub. If you’d like to test out my favorite Notion templates, see the ones I’ve left for you over on http://businessgrowth.ai.