How to be a data science unicorn: Developing your “business sense”

So many resources on “how to become a data scientist” focus on the technical side: Learn Python! Do machine learning! Study statistics! These technical skills are often framed as the biggest barriers to entry in data science: they’re the things that people *know* they don’t know. “Do you know Python?” feels like a yes or no question: either you can write the code, or you can’t (though there is, in practice, so much variety even in the answers to that question).

It’s harder to talk about the things that people often don’t know that they don’t know. Many “business sense” skills fall into this category. We all think that we are strong communicators, good strategists, intuitive product-builders. But these, too, are learned skills that require training and practice, just as much as any technical skill.

A technically-skilled data scientist may be effective, but a technically-skilled data scientist with a strong sense of business is unstoppable. To excel in this field, you need to refine your skills for both the technical and the business sides of the job.

Why business sense matters for data scientists

In any business, no matter your role, your job is to drive value for the company. Your technical skills are simply the tools you leverage in order to achieve that end. But unlike software engineers, data scientists often cannot take their cues from product managers on how they can best add business value. Data is unwieldy and surprising, and there are often opportunities lurking within it that only the person who has their hands in it all day will be able to see. As Eric Colson, Chief Algorithms Officer at Stitch Fix, writes:

“The goal [of data science] is to learn and develop profound new business capabilities. […] There are no blueprints to follow; these are novel capabilities with inherent uncertainty. Coefficients, models, model types, hyper parameters, all the elements you’ll need must be learned through experimentation, trial and error, and iteration. […] With data science, you learn as you go, not before you go.”

Eric Colson, Chief Algorithms Officer at Stitch Fix

That’s why my definition of a data scientist is simply: Someone who can spot (and leverage) opportunities in data.

There are plenty of technical skills that align with this definition: You will probably need to write SQL in order to dig into the data to find said opportunities. You will probably need to write Python to take advantage of those opportunities. It will be helpful to have a researcher’s mindset (so you are comfortable asking questions in the face of the unknown), a good grasp of statistics (so the data doesn’t lead you astray), a strong numerical intuition (so you can follow your nose to the right mental model of a problem), and a broad understanding of basic software engineering principles (so you can see the way from initial data-wrangling to solution-in-production).

But as the Chief Opportunity Spotter, you also need to be aware of broader business goals and how you can leverage your data to serve them. To be effective, you need to have what’s generally lumped under the nebulous heading of “business sense.”

Developing business sense

Business sense is what allows you to spot meaningful opportunities in your company’s data. Having business sense essentially amounts to having a strategic understanding of the company’s goals, and bringing that understanding to bear on how you define and prioritize data-based projects.

Primarily, business sense requires that you have a deep understanding of how your company works. Luckily for you, you are the best person to figure out the by-the-numbers story of the business, because you already have all the company data at your fingertips.

So the first step in developing your business sense is to understand your company of interest. Even if you’re not employed yet, you can use the frameworks below to contextualize your skills during the interview process, and carry this understanding with you once you get the job. I recommend picking a practice company (perhaps the one you dream of working at) and mapping out your answers to the questions below.

Understanding your company

To understand your company, explore the questions below. This is not an exhaustive list of business-related questions, but it should be enough to get you thinking in the right direction.

How does the company make money?

Get a good understanding of the company’s business model. Where does the revenue come from? With what frequency is revenue realized: is it recurring or one-time only, monthly or annually? What is the lifetime value of a customer, and what does that lifecycle look like between the time they start paying and the time they stop paying?

Having the actual numbers here is helpful: How much revenue does the company generate per year, per quarter, per month? Are the revenue numbers consistent over the course of the year, or are their seasonal trends? Many consumer products realize over half their revenue during the Christmas season, so the operations of the entire company are organized around the fourth quarter of the year. Understand when and how the money comes in and you’ll start to understand how you can be most effective.

You should be asking yourself: How can I help the company make more money?

Who is the customer?

Determine what is the target market your company is selling to: Is it upmarket (a premium product) or downmarket (a good deal)? Are you selling to businesses or consumers? If businesses, who are the decision makers you have to convince to buy? If consumers, what are their demographics? What geographic regions do you sell in, and why? What sections of the market are you missing, either because they aren’t your target audience or because you haven’t figured out how to reach them?

You should also figure out the distribution of paying customers. Does the company make 99% of its money from 1% of its customer base? What separates those high-value customers from the rest of the pack? Do you have power users (super fans or people who use your product with surprising frequency)? What differentiates the power user from the rest of your customers?

You should be asking yourself: How can I help the company attract more customers, or convert more existing customers into high-value customers?

How does the company lose money?

There are leaky points in every business: customers drop out in the acquisition funnel, customers “churn” (don’t retain), or stick on the non-premium version of the product. What is your customer retention? Where do customers tend to fall off? What does that say about the value of the product? What differentiates users who churn from your devoted users? What does the upgrade process look like from the basic version of the product to the premium version, and what percentage of users choose to go premium? Do you have a return or refund policy? What is the policy? How frequently is it used?

Learn the industry standards for customer retention, acquisition rates, upgrade rates, return rates, and other metrics to understand whether your company is doing well or poorly at these things. Understanding the biggest leaky points at the company will help you prioritize your work going forward.

You should be asking yourself: How can I help the company lose less money?

What is the company’s acquisition model?

How does the company acquire new customers? Is it through advertisements, field sales, SEO? If ads, what ad channels do they use (Online or offline? Facebook, Google, Instagram?). What are the conversion rates for each channel? Why do these rates differ?

Check out this fabulous talk by Rachel Hepworth, who heads up growth at Slack. In the video, she talks about how every company’s growth model needs to align with their business model, product, and market. Growth through online search ads works best for a product that aligns with what people are already searching for, but ads won’t work for a product that requires a lot of user education (for that, field sales is probably better). Know your company’s acquisition model and understand why it works with your particular product, business model, and market.

You should be asking yourself: How can I help the company acquire more customers, more efficiently?

How does the product work?

Map out the acquisition funnel from the customer’s first encounter with the product (for example, seeing an ad) to first dollar they spend. Where people drop off? At what percentage rates?

Go through the product onboarding process yourself. Look at your own onboarding data and understand how it maps to the onboarding process. Figure out how to query all this onboarding data so you can assess the onboarding success rates for incoming customers. Where do people drop off in onboarding? How does that impact overall retention?

Use the product a few cycles (for a video game: a few logins; for a smartwatch: a few days) and look at your own data. How did you like the experience? What about your experience can you see reflected in your own data? Learn your customer retention like the back of your hand. Do people drop out after a couple hours, a couple days, a couple months, a couple years? You’re learning whether your product is a fruit fly (short lifecycle) or an elephant (long lifecycle). There’s nothing inherently wrong with either of these, so long as it suits your business model.

Write down what you love and what you hate about the product—but don’t necessarily take your feedback straight to your Product Manager. Keep those thoughts in the back of your mind, so when an opportunity in the data comes along, you can think, You know, this would help solve Problem X, or This would help capitalize on Great Feature Y.

You should be asking yourself: How can I help make the product better?

Who are your competitors?

Make a list of your biggest competitors. How are you similar or different from them? Do you target the same customers (same parts of the market), or not? How does your product compare to theirs? Can you find any stats about their business: sales, retention, funding?

Learn about the people running your competitors’ companies: What kinds of businesses have they run in the past? What is their pitch for this business? How do you think these factors will impact their approach going forward?

What does your company do that your competitors don’t, and vice versa? What opportunities are available to you that aren’t available to them? For example: At Pebble, I studied Apple all the time once they announced they were building the Apple Watch. Apple is notorious for having a fairly closed-off developer environment (it’s part of their brand, after all), so I advocated for making Pebble’s as open as possible. We shared our step-tracking algorithms and built tools so developers could work easily with the data they collected through the apps they built on our platform. You should also be aware of what your competitors can do that you can’t (or won’t).

You should especially look at the data science teams of your competitors. Do they have a data science team? Do they split up data science and data analytics? What’s their engineering stack? What does their engineering support for data look like? Who’s in charge? Follow their leadership team’s blogs and see what you can learn from them. Hopefully you’ll run into them at networking events, and you can pick their brain about their process and assess how it’s different from your own.

You should be asking yourself: What can I learn from our competitors?

What is your brand?

Understand what your company is good at, what it’s weak at, and what identity it’s trying to put out in the world. Know the fuzzy stuff. You want to build products that align with the identity your company is working to create for itself. Are you fancy and sleek or scrappy and approachable? What you build with your data should be tailored to the company’s core identity.

Google your company and see what kind of press it’s getting. How is your company perceived by the outside world?

You should be asking yourself: How can I help define, enhance, or at least act in alignment with the company’s brand identity?

What is your pitch deck?

Some companies will show you the slides they’ve used to raise money (you may also eventually find yourself stuck in a conference room with the CEO making the slides alongside them). But even if they haven’t, you should assess the kinds of numbers that would get shared in an investor context. What is the total addressable market for your product? What is the company valuation? What has the growth curve looked like to date—in terms of monthly sales, growth in the customer base, retention?

When is the company planning to raise money next? How much would they like to raise? What milestones do they need to hit in order to have a successful round of fundraising? Your CEO may or may not be forthcoming with these details (which is fair: fundraising is a hairy topic to discuss with employees), but just knowing these are good questions to think about can help you frame your work more effectively. Boosting sales, expanding market reach, and improving retention are all the kinds of things that look great on a pitch deck—if you can help the company do that, you are improving their chances of raising cash in the future. Look at a few pitch decks in your industry to get a sense of what investors are looking for (there are, for instance, a bunch of examples here).

You should also know what about your company is going to look bad on a pitch deck, and seek out opportunities to mediate it. Weak retention may not have a huge impact on revenue in certain instances, but it’s a metric that may turn investors sour.

Research online to find out how much has the company raised to date. Crunchbase is a good resource for finding this information. Who are your investors, and what other companies have they invested in? Read your investors’ blogs and follow them on social media. In theory, these are the people giving your CEO advice, so it’s good to understand how they think.

You should be asking yourself: How can I help the company raise more money?

Learn your limitations

Despite what many data scientists (and often, CEOs) might think, data is not the solution to every problem. Sometimes a few good qualitative insights (like, say, running user surveys) are better than a slew of complex quantitative feedback. Understand that data is not always the silver bullet you’d like it to be, and interrogate whether the data-powered solution you’ve come up with is truly the best way of solving the problem at hand. Creating value for the company may sometimes mean telling the CEO, “You know, I really don’t think we need to build a machine learning algorithm for that.”

Similarly, some areas of the business are not optimally suited for traditional data science intervention. Many data scientists take an interest in the marketing side of the business, for example, and hope to crunch the numbers in order to improve ads and conversion rates. But the kind of data you get from Google Analytics or Facebook Ads Manager is not the kind of data that most data scientists are adept at working with: it’s summary statistics, not a giant relational database of clicks ready to be mined. So beware the seduction of problems where the data set isn’t suited to what you do (and be forthcoming in telling the marketing team that despite all appearances, you are not, in fact, a wizard).

Keep it simple

Many of the opportunities you spot (especially in a startup environment) will appear simply because nobody has ever looked at the data before. That means there will be a lot of low-hanging fruit: opportunities to create value for the company that don’t require you to flex your technical muscles to their fullest extent. Go for what is high-value, not what is technically complex or personally interesting. Keep your focus on how to best serve the company and your efforts will almost always be effective and appreciated.

Meanwhile, keep this in mind: Part of your value as a data scientist lies in your ability to help educate the company on how to think about its problems. Playing with data requires you to build mental models of the dynamics of customer acquisition, product usage, possible competitive advantages. These mental models are your most valuable deliverable. They not only underpin every data-powered product you want to build yourself, but they also have the potential to form the very foundation for how the company operates. As you develop your business sense and bring it to bear on your work, you might find the executive team slowly learning to speak your language. They’ll use your same phrasing when talking about retention. They’ll incorporate your thinking on acquisition channels into how they measure company KPIs. They’ll bring your experimental discipline to bear on how they run ad campaigns. Be strategic in your work and you can have a huge impact on practically every area of the business.

Liked this post? Follow us on Twitter to see when we post something new.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s