In May 2017, The Economist’s cover depicted oil rigs that looked like office buildings and bore corporate names like Google, Facebook and Amazon. The article’s bold claim: “The world’s most valuable resource is no longer oil, but data.” Before and since, the oil-data analogy has been popular in the halls of Wall Street, Silicon Valley and Washington.
To extend the analogy, data became so valuable in the same way as oil. By itself, oil is worthless. It provides neither food nor shelter nor enjoyment. Yet it transformed the world once people figured out how to extract, refine and harness its energy. Human ingenuity made oil valuable. The same is true of data. To an unaided businessperson, massive piles of data are impossible to work with and less valuable than an anecdotal customer conversation. But with the aid of analytics technology and the right analytical approach, data provides incredible value.
Proper collection and analysis of data can reveal almost any information and insights that are important to a business. It can tell you about your customers and your employees, your industry and your operations, your past, present and — sometimes — your future. For growing businesses, data can play a vital role in making the leap to the next level. But to harness that power, you need to build the data analytics equivalent of an oil rig, a refinery and a combustion engine.
The good news: In the world of small business data analytics, oil rigs turn out to be easy to build. Some companies even build them by accident. As companies grow, they tend to use more software. That software — from basic accounting to enterprise resource planning (ERP) systems — generates “data exhaust,” digital records of business activities which you can organize and analyze. Data analytics becomes even more powerful with a plan: By slightly tweaking the information technology most businesses already use, you can collect data in a way that speaks to your important questions and helps you run your most important analyses quickly and easily. From there, companies can go the extra mile and collect additional data via tools like surveys and publicly-available sources, or they can purchase relevant datasets.
This installment of the Project $50 Million series distills advice on harnessing data to fuel growth, from the questions to ask to the data collection processes to the analytics — even how much to trust the results.
Among other benefits, launching a data analytics practice at a growing business can:
As a business grows larger, it generates and can collect more data while gaining resources with which to mine that data. Plus, the value of a small data analytics contribution multiplies as an operation grows.
Data can help companies cut through months of experimentation while yielding insights that may have been otherwise unattainable. At the same time, though, it’s easy for businesses new to data analytics to focus on the wrong patterns or draw erroneous conclusions. Said another way, putting faith in a specious data correlation or a misinterpreted significance can be worse than no data analysis at all.
Every smart business first wants to know more about its customers. For young businesses, customer conversations tend to generate the most useful insights. But as your company grows from $10 million to $50 million, business leaders’ customer conversations simply won’t be able to keep up. Thoughtful data collection and analytics, however, can keep pace as the customer base expands. If you’re selling a retail product, for example, you may collect data on customers who have set up online accounts or by encouraging reviews. Many companies randomly sample customers for surveys. Businesses in which salespeople speak directly to customers may have information in customer relationship management (CRM) systems and, with a little training, their teams can learn to collect that data in a consistent way that enables later analysis.
As a business grows from a small core team to a large enterprise, not everyone will know, trust or coordinate well with each other. Collecting data on operations and personnel can give you insights into how your company is handling the growth and improve managers’ ability to get teams coordinating properly. Some of that can happen automatically with good analytics: If various business functions are using specific metrics or KPIs as barometers for success, then transparency around those metrics will focus their efforts and deliver instant, objective feedback – as long as those metrics are chosen well. That transparency in priorities and data-informed decision-making can also engender trust among employees who, as the company grows, have fewer opportunities to interact with senior leadership.
Like so much in business, building a data analytics process that will steer you right involves first defining the goals of your program and the role you want it to play in your strategy. While that sounds awfully obvious, too many organizations start by taking whatever data they have and running regressions on it until insights pop out. Yes, early on, a business can mine existing data for useful insights. But over time, there’s a fast-growing risk of making a wrong turn by using data and models that don’t match the questions most important to your business.
1. Hire the right person or people who know statistics, as well as how to interpret statistics and when they can go wrong.
2. Determine which questions matter most to the success of your business.
3. Identify the data that is most relevant to those questions.
4. Collect or buy the required data.
5. Let the statistics maven(s) analyze that data until they find insights.
6. Automatically generate the most useful metrics, and distribute them to all relevant stakeholders.
7. Iterate steps 2 through 6 to refine your analysis and dig for more insights.
8. Accept that this process isn’t always linear. Sometimes a later step informs a prior one and you must iterate again.
Building the right data analytics process for any business, therefore, must start with identifying those most important questions. The next step is to figure out which data and analytical methods are required to answer those questions for that company. For example, there’s no single way to learn about customers: A $10 million company selling candy has vastly different customer interactions — in number and type — than a $10 million company selling software to businesses. But clarifying about what you want to know, and the sources of data you have or can create to speak to those questions, is the first step to designing an analytics process that will be successful in the long run. The candy company, for instance, might answer questions about how to best prepare for Halloween and Valentine’s Day by learning more about the seasonality of its sales. It would start internally, by collating and setting up automatic collection of sales data in useful formats. Meanwhile, if the software company has questions about a new target customer for the next version of its product, it must look outward for data. Leaders there might use some combination of interviews, focus groups, surveys, third-party data sources, social listening and more. Data from the company’s past and current operations won’t speak to those new-customer questions.
Here’s a rundown of seven key questions small, growing companies should think through and tailor to their specific needs, then gear their data analytics processes to answer. Each includes thought processes to help you move from question identification to data analysis, kicking off with three common customer-related questions:
1. Does our business attract the right types of buyers?
Good data analytics can spot mismatches between actual customers and the ones you expect, potentially helping you avoid missteps or exploit unexpected opportunities. But turning findings into insights can be tricky. Say you have a product for which one of the target markets is graduates starting their first job after college. Data collection shows that you captured the right segment but your customers are overwhelmingly males. Now what? Is there something about your product that appeals to men, not women? In so, the right response might be to target men more specifically in advertising. On the other hand, perhaps something about your messaging disproportionately appeals to men, and you need to make your ad campaigns more inclusive. Or perhaps the gender discrepancy isn’t the fault of your company at all — maybe a competing product is aggressively targeting female demographics. Your data analytics can pick up the issue, but you’ll need to do more research to craft an appropriate response.
2. What’s the yield of our customer pipeline?
A good analytics platform can diagnose the effectiveness of your sales efforts at specific pipeline stages — if your salespeople track each lead from initial outreach to purchase decision. You can regularly update numbers such as: X% of leads turn into real sales conversations; Y% of those turn into trials; and Z% of those become paying customers. If X starts to drop, perhaps you’re not targeting the right customers. But if Y starts to drop, maybe your sales team needs more training. If Z starts to plummet and you haven’t changed the way your trials work, there may be a problem with your core value proposition. You can glean this extremely valuable information with relatively little effort. If anything, it usually involves only slight modifications to and standardizations in the way your sales team tracks their progress. For instance, keeping very brief records of exactly when and why a sales team stopped pursuing a potential customer involves minimal effort today but will answer a lot of questions tomorrow.
3. Where and why are we losing customers?
Some companies do everything they can to collect detailed data here: exit interviews, follow-up surveys, investigations into service failures and more. If you have thousands of customers, it’s not feasible to investigate every departure, but it’s almost always worth looking into why customers leave. Do they tend to leave when you raise prices? At a particular time of year? In response to a competitor’s product launch? In response to your product launch (see: New Coke)? You can collect some relevant information — like the timing of canceled subscriptions – automatically, while you may need to go out and collect other information via interviews or surveys. Even a single question sent over email or asked during the cancellation process, if you have one, can serve this purpose. Collecting this information is harder with retail items for which customers have to opt in to buying again rather than opt out. For example, if you sell bedsheets, it’s very hard to know when customers switch to other brands, but if you sell a monthly subscription to a publication, it’s much easier. In either case, though, it’s almost always worth putting in the time to understand why customers leave.
4. Are we hiring the right people, and are our offers competitive?
Hiring the right people is tricky, but there are indicators that something might be amiss. High turnover, for example, can be a sign of several issues, but over the long term, the rate at which you fire employees and/or mutually separate is almost surely an indicator of hiring process effectiveness. Looking at retention and performance by cohort, or groups of employees hired relatively close together, can give you a sense of whether your hiring program is improving, worsening or staying the same. This is especially important as companies grow. Comparing your offers and salaries with those of competitors takes more work because you must seek external data. But it can be well worth the effort, especially during periods of fast growth. It may be counterintuitive to spend extra time on difficult data-gathering when you’re already short-staffed, but if non-competitive compensation results in a low-yield recruitment season, you’ll wish you had put your best people on the task.
5. How would various scenarios impact our bottom line?
Answering this question is a little more complicated than calculating a metric from existing data streams. But once a business tracks and runs enough of its operations with software, conducting “what if” analyses becomes more straightforward. In some cases, it’s as easy as typing in a few numbers and seeing how results from your spreadsheet, accounting software or ERP system change in response. Complicated scenarios may require a more detailed understanding of how moving parts fit together, but data analytics can help you estimate those relationships. “What happens to our bottom line if sales drop by 20%?” is a good example of a simple scenario to analyze. Questions like, “What happens if China puts a tariff on our industry?” require more sophisticated statistical modeling and assumptions — and probably external data.
6. How can we best take advantage of growth opportunities?
Answering this broad category of questions is one of the most popular and important uses of data analytics. Whether companies need to figure out how much excess capacity they can devote to promotional giveaways, measure the virality of a new innovation or A/B test referral incentives, data analytics can help. By casting a wider data-gathering net — e.g., fielding a survey, buying external data or leveraging free government data — you can take a data-informed approach to exploring new geographic markets, introducing new product lines and even improving the way you invest in potentially groundbreaking innovations. For these more far-reaching questions, you may need to do some one-off data gathering, seeking information that you don’t intend to collect regularly for tracking purposes. This is especially common when expanding into new geographic markets. Once you decide to enter a new market, however, you won’t need to constantly refresh your original analysis on whether doing so might be good for you — you’ll collect data on your actual results and experiences.
7. Where are our blind spots? Which other types of data do we need?
It’s almost universally true that as you go down the data analytics path, you’ll identify needs you didn’t plan for. Whether it’s new situations emerging or natural follow-ups to the analytics you’re already running, you’ll eventually find yourself wanting to answer questions that your data doesn’t speak to but which you could answer with some added effort.
Sometimes this can send you down a new line of inquiry — for example, if a competitor is targeting a particular demographic and you need to figure out why their product or messaging is luring away your customers. But sometimes, blind spots can exist as critical nuance within the questions you’re already asking. Many companies offer products to make life easier and safer for parents of infants, but some find that “parents of infants” isn’t specific enough. Some products are better targeted at “new parents” who just had their first infant, while others may appeal to parents managing an infant while also caring for older children. Companies that broadly target “parents of infants” may spend a lot of time and money communicating to parents who aren’t really in their target market. Asking the right questions and collecting the right data can transform the way a company views its customers and their problems — and, in turn, transform its return on marketing investment. Not every item on the wish list will be worth pursuing, and not every blind spot will be possible to illuminate, but staying humble, open-minded and curious throughout the process will yield continual improvements to your data analytics.
Having committed to a data-informed approach and thought through the key questions, it’s time to build your business’s data equivalent of the internal combustion engine: the machinery that turns data into insights and helps you get where you want to go. Here are the most important guidelines.
If you’re going to rely on data analytics to make important business decisions, you need to be sure the data is telling you what you think it’s telling you. You need someone who knows not only how to run the analytics but also how to interpret the results and put them in context — what you can and can’t say based on the data. You need someone who’s aware of the assumptions and limitations of the methods used and can, ideally, explain the implications to a variety of stakeholders. Midsized businesses often struggle with how to hire for these roles; some assume being good at math or computers makes a candidate good at data analytics, but those are different skill sets. You want someone with a background in analytics, statistics, data science, social science or similar. Businesses utilize social scientists — except for economists — perhaps the least, but the quantitatively-oriented professionals in those groups are specifically trained to think about how human factors and choices impact the reliability of data analysis and how to get around those challenges.
Some companies prefer to start by developing analytics talent from within. They often begin with an employee who is a good communicator with potential aptitude for and interest in these skills, providing them with training and the resources/courses needed to become an analyst. That works great for analytics programs that need to manage data, calculate KPIs and provide feedback and analysis on trends. When the questions become more complicated, however, a company may need team members with more formal training. Questions of causal inference — when you don’t just need to know whether X and Y move together, but whether X causes Y to move and to what extent — require more expertise, especially when a high degree of confidence is required for making business decisions. Companies that don’t need a full-time staffer for this engage consultants – or, more cost-effectively, a PhD student who has the skills and hasn’t finished their dissertation yet.
As data analytics teams grow, roles can become more specialized. Some companies have employees who specialize in data collection and management working together with those who specialize in using the data, plus some regular contractors or consultants on an as-needed basis. The right person, and later the right team, will look a little different for each company. But one truth is constant: Hire good people at the beginning, and the rest of the process becomes easier and far more reliable.
Sometimes this means literally implementing Agile, a framework common in software development that has been borrowed and applied to business processes. Agile methods(opens in new tab) put structure and discipline around an iterative approach to creating solutions to known problems for situations in which you don’t necessarily have a good sense of the answer. Agile iteration is similar to plan-do-check-act (PDCA) cycles, which are discussed more fully in the next section. Whether you’re literally implementing Agile or not, disciplined flexibility and an iterative, scientific approach is essential.
A good data analytics engine studies the KPIs that most meaningfully illuminate the questions discussed in the previous section. Determining which KPIs are most important to your business is an idiosyncratic exercise, as the question discussion shows. And once you have those KPIs, don’t rest on your laurels.
Specifically, periodically rethink your KPIs in three ways:
Think beyond the standard metrics.
This is not to disparage the common/standard metrics. If you’re running a software-as-a-service business, for example, a handful of standard metrics — such as customer acquisition cost, customer lifetime value and churn rate – will tell you almost everything you need to know about the business’s economic sustainability. Net Promoter Score (NPS) is a tried-and-true metric relevant to companies in many industries. But there are also factors unique to your business — particularities about your operations and challenges about your situation. It takes work and experimentation, but it’s worth figuring out which unusual or even unique KPIs make sense for your business, the way passenger revenue per available seat mile — i.e., PRASM — makes sense for an airline.
Consider the appropriateness of your
As your business changes and grows, KPIs that served you well in the past may not work as well going forward. Ask yourself: Are there important metrics that aren’t being captured? Do we have legacy KPIs that no longer impact the bottom line very much?
Examine how well your KPIs do their
Understand how much you can trust your numbers. Start with key measurement characteristics, such as:
A common mistake is prioritizing KPIs simply because they’re the easiest to measure. Remember, employees will respond to these KPIs, especially if you manage to and build incentives around them. The metrics that get the most attention should be those you care the most about, and those metrics need to reflect and summarize reality in a dependable, useful way.
If you’re planning to grow by leaps and bounds, get ready to manage data as a much bigger company. One smart person running analytics part-time on a spreadsheet can work well for a startup but not when companies hit tens of millions of dollars. You’d need teams of spreadsheet gurus at that point, and even then, you’d face serious risks in usability, portability and interoperability of your data.
The longer you wait to implement a sustainable solution to storing your data — ideally the clean, easily-accessible, usefully-formatted, central repository known as a data warehouse — the harder and more costly it’ll be to make the switch. To the spreadsheet aficionados out there, don’t worry. A good system will still export data extracts to Excel.
Like moving from spreadsheets to data warehouses, formalizing your data governance is another action item on which delays can be costly. As the company grows, so too will the teams contributing to and using your data sources. Once you reach a stage in which the people who need to coordinate aren’t always directly collaborating, you’re on a ticking clock. The sooner you create formal, standardized procedures, the better off you’ll be. Data governance is one of those terms that strikes fear of bureaucracy in the hearts of entrepreneurs, but, at its most basic, it is simply the policies and processes meant to guarantee good management of an organization’s data assets. Robust data governance(opens in new tab) ensures the quality of a business’s data, helps businesses extract maximum value from their data and mitigates risks associated with the data.
Over time, growing companies are better able to afford and take advantage of expanded data collection and analytics. These can help you grow by delivering information on the new potential customers, geographic areas and competitive landscapes you may want to enter. The key is to add resources in a manageable way. “Don’t bite off more than you can chew” may be a cliche, but it’s prudent here. Companies can spend a lot of money acquiring resources that they’ll wind up either not using or using very ineffectively as they attempt to digest too much too quickly.
For example, some companies buy expensive software and use only the basic capabilities they could’ve gotten for free elsewhere. Meanwhile, others buy lots of data from third-party sources to “cover their bases,” even though the analytics team doesn’t have the time or a specific use for it. Precisely target the questions you want to ask and the information you’ll need to answer them. Growing your capacities steadily and realistically will reflect in your performance.
More Resources From NetSuite
Understanding Data Warehouses
Get a further breakdown of data warehouses, including signs it’s time to start using one and real-world examples of their applications.
Making Data-Driven Decisions With Analytics(opens in new tab)
This free webinar further explains how technology aids in data analysis, especially in tracking KPIs and extracting actionable insights.
NetSuite Analytics Warehouse
Learn more about NetSuite’s prebuilt, cloud-based data warehouse and analytics solution that consolidates multiple sources of data to help teams build and run analyses.
The PDCA cycle(opens in new tab) is a four-step framework which many organizations use to make business decisions and manage change. PDCA stands for:
Entrepreneurs familiar with Lean startup practices may see substantial overlap with its build-measure-learn cycle(opens in new tab). This is, indeed, very similar. While Lean startup terminology frames the concept for the specific use case of creating new products and services, both frameworks derive from the scientific method — i.e., testing hypotheses and iterating until you gain some valid learning.
It’s important to realize how data can improve PDCA cycles and similar frameworks in all areas of business decision-making. For example:
Data analytics can help you recognize opportunities and even predict the results of various actions. With enough data, you can learn from similar situations and determine planning priorities.
Example: A food company sees that most of its fastest-growing products are cheeses. Looking deeper, it discovers a disproportionate share comes from its few vegan offerings — one in 20 products is vegan, but six of the 50 fastest-growing are vegan. The company decides to expand vegan offerings and marketing.
You have a plan and some outcomes you’re targeting. Now, you need to collect the right data in a way that will let you draw conclusions about the hypotheses involved in your plan. If you know which analyses you’ll run in the next step to test the assumptions you made in the previous step, then you’ll be able to back into which kind of data you need to collect.
Example: The food company realizes it’s making a lot of assumptions, one of which is that consumers want more and new vegan products. The company runs targeted ad campaigns for existing vegan products, as well as vegan products it hasn’t yet created but is considering. The landing page for the nonexistent products invites customers to join an email list for notifications about the launch, and it asks a few questions about which other products they might like to see.
Data analysis is often a large component of this step — in some cases, it’s the entirety. Be methodical and disciplined in focusing your analysis on the questions, hypothesis and assumptions from your planning phase. Make sure the questions you’re asking and the experiments you ran are well-aligned.
Example: The food company analyzes results from the experiment, looking at clickthrough rates on each ad and resulting customer feedback. It confirms that customers want more plant-based options. However, the team is surprised to learn that many of the people who filled out their short survey are not actually vegan.
This is when most people think, “Great, my data analysis work is done. Time to go implement.” But remember that at some point, you’ll have to make more decisions, implement more adjustments and go through more PDCA cycles. Do yourself a favor and only consider the job done when you’ve put in data collection that will inform your future efforts. The first task of this data-gathering is to see how well your testing predicted the outcome at scale; eventually, your data-gathering will inform the “plan” step in your next PDCA cycle.
Example: With its hypotheses mostly confirmed, the company can more confidently invest R&D dollars into new vegan products, with one important adjustment: It will market the products as “100% plant-based” to appeal to non-vegan customers who are trying to eat healthier or more sustainably.
This is a common question, and you might expect an article on using data analytics to say “a lot.” But truthfully, the best companies and business leaders are more data-informed than data-driven. They use data and analytics as important inputs into decisions, but they don’t always treat them as the only or even most important factors.
In some situations, data may give you everything you need. If a business is scaling and trying to do more of the same thing but more efficiently and for more customers, then data analysis can accelerate its timeline and improve performance. There will be times in most businesses' journeys, however, when it won’t even be clear which conclusions teams can draw from the available data. And in some business decisions, especially the boldest and most important ones in which companies enter uncharted territory, there will be factors that you just can’t capture with any kind of data collection. Making decisions about the future necessitates betting on the unknown. Good data analysis is extraordinarily valuable in placing those bets, but to go where it points always and without question is to assume that every valuable input can be captured in the data and properly understood with analytics.
Data may be the most valuable resource on Earth, but you can’t fully realize that value without human intelligence, innovation and inspiration. If you get the right team in place, keep your priorities straight and make a commitment to data-informed decision making, then data analytics can help take your business to the next level and beyond.