Business Strategy For Data Scientists

Data science is exciting. From the moment we learn our first machine learning algorithm, there’s an urge to apply it to data, make forecasts, and uncover deep insights.

But the Business side is really important and interesting too. I promise you that investing some time into understanding how businesses work and what separates good companies from bad ones will pay significant dividends as you progress in your data science career.

This will probably be the first of several posts as it’s impossible to do the subject justice in just one post. Also, please don’t think of these posts as an MBA-lite. Rather, I plan to write about the concepts and ideas that have most influenced how I view the business world. Thus, there might be some relevant business strategy topics that I skip entirely. Of course, if there is anything you really want to hear about from me, please let me know in the comments (and I will do my best!).

Company Growth Phase vs. Data Science Needs

Silicon Valley start ups not withstanding, the point of a business is to earn profits. Whether and how this happens depends primarily on the phase that the company is currently in:

Hyper-growth – At this stage, sales growth matters more than profits so the company is likely absorbing substantial losses in order to gain more and more customers. The fuel for this growth is provided by venture capitalists, with the objective being to scale revenues to the point where the company is financially attractive enough for an IPO (initial public offering).

Since growth is what’s needed, data scientists at hyper-growth firms focus on customer acquisition and product development. Problems that they might work on include:

Quantitatively identifying the most attractive prospective customers.
Designing and optimizing marketing strategies such as couponing, referral campaigns, social media advertising, etc.
Researching and developing the algorithms that form the core of a company’s product or service.

Public, growing (less quickly) and profitable – Some private companies definitely fall in this category too but we will focus on the public ones for now (because Wall Street is such a driver of how firms in this phase act). Once a company is public, earnings per share (net income divided by shares outstanding) becomes the key metric. Public investors reward firms that can consistently increase profits and punish firms that cannot (Uber is a recent example of this). Thus, data scientists at these firms must focus on more than just growth. Oftentimes, these firms are large enough that optimizing the expense side of the income statement matters as much to the bottom line (profits) as increasing sales. Some challenges data scientists might tackle include:

Designing customer loyalty programs or other ways of keeping customers engaged and spending. The objective of this and most of the next few bullets is to maximize customer lifetime value.
Quantitatively identifying customers that are the most likely to churn.
Building recommendation systems to effectively cross-sell new goods and services to existing customers.
Using data and analytics to help identify new markets to enter or even companies to acquire.

Declining – While no one aspires to work for a declining company, these companies often have as much or more data needs than growing ones. Usually a company is in decline either because it is mismanaged or because its industry is in a structural decline.

If the company is mismanaged, there might be opportunities for data savvy folks to use analytics to right the ship. Frequently with these firms, there are so many areas for improvement that even simple analytics (a few Excel pivot tables) can make a significant difference to the business. Given the abundance of improvement opportunities, the more important question is whether the company’s board and management have the appetite, courage, and willingness to change. If not, then stay away.

Companies in structurally declining industries pose a tougher test to data scientists. And it might be a test where there is truly no right answer, just ones that are less wrong. I would definitely recommend against hooking your career to an industry in decline. But if you must, then as a data scientist, you might look into identifying quantitatively why your industry is in decline and whether there are any levers to lessen the damage to your company relative to its competitors.

Also, are there any opportunities to consolidate? Consolidation (buying up your competitors) is often the one lifeline that industries in severe decline have as it removes supply and restores some semblance of pricing power. Granted, most of these decisions are probably above the pay grade of the data scientist – but I do believe that an objective, data-driven approach could prove quite helpful in these types of situations (also important is the ability to effectively communicate quantitative findings).

A Note on Cyclical vs. Structural Declines

Just now I focused on structural declines, which are declines caused by a seemingly irreversible trend – some examples include what Uber did to the taxi industry, Netflix to cable television, digital photos to film, and Amazon to traditional retail. In most of these cases, the new product or service was both better (either in terms of quality or convenience) and lower cost and the company offering it was better run and hungrier. That’s an incredibly tough combination for the incumbent to overcome.

On the other end of the spectrum are cyclical declines (visually depicted below). These usually involve commodity producing industries such as crude oil, semiconductors, flat-screen TVs, and even smartphones (Apple excepted thanks to its iconic brand). These industries are characterized by intense competition, minimal product differentiation (Exxon Mobil’s oil is the same as Chevron’s), and an inability to set prices. Cyclical declines are the bust part of boom and bust cycles that affect all commodities – the cycle usually goes something like this:

High prices attract both new entrants into the industry and higher investment from existing players – which massively increases supply.
The flood of new supply pushes down prices and everyone’s profits. If it’s accompanied by an economic downturn, shrinking demand forces prices even lower. Bankruptcies abound.
Bankruptcies and consolidation eventually restore the balance between supply and demand. The companies that survived the downturn now get to reap the rewards as the cycle begins anew (we never learn).

The key is that cyclical declines are eventually followed by a recovery while structural ones do not conclude until the incumbent industry is irreparably damaged. In the moment, it’s often very difficult to figure out what type of decline your company is in. I guarantee you that many cable company execs assumed that the decline in cable subscribers was merely a temporary one – so they dutifully invested in new shows and channels all the while waiting futilely for the trend to right itself (it never did).

That’s another area where Data Science might help. It starts with the hypothesis that the decline your company is experiencing is cyclical. To test this, you might study economic time-series data or the financial statements of your firm’s customers (you could do a sentiment analysis of their quarterly earnings call transcripts). And after all that, if you deem the downturn to be temporary, then it’s time to turn your data science skills towards identifying cheap acquisitions (to spur consolidation) and optimizing your firm’s operations (to cut costs and build cash so that your firm can make it to the recovery phase).

Metrics to Measure Success as Your Company Attempts to Achieve Lift Off

But most of us will hopefully work for growing, high potential companies (less stress of being laid off). Thus, let’s take a look at the two primary deciders of whether our firm will achieve liftoff and ultimately become profitable and successful – customer lifetime value and customer acquisition cost.

Photo by Chevanon Photography from Pexels

Customer lifetime value measures in today’s dollars (inflation adjusted terms) how much a customer is worth to the business. The exact definition differs from company to company, but you can imagine it this way. Let’s say you run a cafe and I am your reasonably loyal customer. I visit approximately once a week (so 50 times a year) and spend around $7.00 each time (a coffee and a cookie) – so each year I spend $350 at your cafe. There are a lot of other competing cafes nearby and you want to keep my business, so you give me a few gift cards each year and the occasional free cookie – the cost of these freebies amount to $50. So net of my retention costs, I am worth $300 to you each year. Finally, based on your analysis of previous customer cohorts (and their attrition patterns), you estimate that your typical customer sticks around for 4 years. *So my customer lifetime value to your cafe business is 4 $300 = $1,200.** Keep gifting me those cookies!

Keep in mind that there are 3 primary levers for increasing customer lifetime value:

Sell more to each customer.
Charge each customer a higher price.
Reduce customer churn rate (which increases average customer life).

An increase in any one of these will increase the worth of a customer to the business.

Customer acquisition cost measures how much it costs on average to acquire each customer. Bear in mind that this metric includes the cost of both your failures and successes – that is when calculating it, you need to spread the total cost of acquisition across just the customers you ended up successfully acquiring. For example, in the past year, you spent $60,000 on posters, Facebook/Google ads, and email campaigns. All in all, you were able to acquire 200 new regular customers (including me as I’m a sucker for a nice poster). So your customer acquisition cost is $60,000/200 = $300.

Good news, assuming that I am representative of your typical customer, your cafe’s customer lifetime value of $1,200 is substantially higher than your per customer acquisition cost of $300.

It’s important to remember that while the $1,200 is inflation adjusted (and often cost of capital adjusted as well), there is a lot of uncertainty attached to it – your estimates of customer attrition could be too low, there could be a fire, annual retention costs could rise, etc. Also, it is earned over the next 4 years, while the $300 customer acquisition cost is paid out right now (during the current year). So the $900 difference between your customer lifetime value and acquisition cost is by no means a done deal. Rather, it should be a thoughtful and conservative estimate of how much profit the average customer might bring to your business over the duration of your relationship with him or her.

How Should These Metrics Ideally Trend Over Time? How Do They Actually Trend?

Lastly, let’s think a bit about what we want to see in regards to these metrics as well as what we are likely to see.

Obviously, it would be great if customer lifetime value continuously trended up over time while customer acquisition cost declined.

This is the case for the most successful companies where by virtue of their competitive advantage (we will explore each source of competitive advantage in detail in future posts), they are able to continuously coax more and more money out of their customers. All the while, as these firms become more successful and well-known, their cost of acquiring new customers drops substantially. For example, think about how much cheaper it must be for Airbnb to acquire a customer now that the company is synonymous with travel and hospitality versus 10 years ago when it was still trying to prove itself as a business. And as it continues to increase its scale and offerings, each new customer will probably stay loyal to Airbnb for a longer amount of time (few comparable alternatives) while spending more on its platform each year.

But what if our company is not the next Airbnb? Then what should we expect to see? The plot below depicts a less ideal but probably more frequently encountered scenario.

Customer lifetime value still trends up but much more slowly. Recall the 3 levers for increasing customer lifetime value: more transactions, higher prices, or reduced churn. All 3 of these suffer when there is significant competition. And unless our company owns some magic sauce like a powerful brand, a patent, etc., you can bet that there will be competition (causing our customer monetization to grow much more slowly than we would like it to grow).

Customer acquisition cost will probably go through 2 phases. In the first phase, there are either fewer competitors or enough customers for all. So our firm (along with its competitors) is able to reduce its customer acquisition cost as its products and services become more familiar and accepted. But eventually, our industry will reach an inflection point where suddenly the market is over-crowded with competitors and the easy opportunities for growth are exhausted.

This second phase is characterized by price competition and customer poaching. Our firm will need to offer more coupons and loyalty points to our existing customers to retain them. Meanwhile, our marketing efforts bear much less fruit than they used to. Thus, both customer acquisition costs and retention costs start to increase and our company struggles just to maintain the current spread between customer lifetime value and acquisition cost. In these times, good management and execution is critical. For example, a firm with a lean cost structure and a culture of valuing its customers will probably be able to strengthen its industry position as it can differentiate its offerings through both lower prices and better service (while taking a smaller financial hit relative to its competitors).

Edit: One important thing to note is that in the early stages of a company (which can last years), customer lifetime value can be less (sometimes significantly so) than customer acquisition cost. This is not necessarily a bad thing – it just means that no one yet knows about your business, so it’s more expensive for you to advertise and promote. And those that do sign up as customers are still skeptical of your value proposition, so they are less likely to give you all their business. This is a natural challenge that pretty much all businesses must overcome as they seek to reach sustainability (in terms of scale and profits).