Being in "charge" of data—whether you're the Head of Data Science or the Chief Data Officer of your organization—is no easy task. It's not always easy to know what's expected from all of the data your company collects, or how you can own your role.
At the end of the day, there's no perfect playbook for running a data team. You need to have a variety of skills and tools at your disposal to be able to take control of your job—and your company's data.
From understanding probabilities to communicating with your team, we'll go over some of the basic skills that any data person needs to have to succeed in working with data in the workplace.
1. Understanding when to buy vs. build
Building or buying analytics technology is a question that comes up with some regularity for companies—and whether that's backend, onboarding flows, or data, there's pros and cons to both.
Upfront, it may seem appealing to take advantage of all the open source big data tools out there, and build your own analytics solution. It gives you total control of what you collect, how you collect it, and what you can do with it—right? Compared to many SaaS analytics solutions, this may be true, yes.
However, if you can outsource and still get what you need from your data, you probably should. Developing and managing data tools can quickly turn into big projects that pull engineering resources away from their core tasks. Many of them are also difficult to use for the non-data scientist end user.
If you're going to buy, you can explore options to decide if you want a full, robust platform, or if you might be just as well off taking a more a la carte approach, collecting data from internal tools and add-ons and stitching them together. Or, you can opt for a full stack platform, which can offer more thorough consolidation of your data, and quickly kickstart your work. You'll also get support integrating a team onto your new platform, which can be a great way to start getting different people comfortable with using data. Interana's unique full stack technology is an example of a solution that provides flexible and powerful behavioral analytics for anyone in the company.
2. Strike early
If you're going to work with data, the sooner you can start collecting it, the better. You need to be able to get a pipeline in place, even if it's in the early days at your company. If you've already got things running at pretty much full speed, you need to be able to set up data collection as fast as possible, while maintaining accuracy and immediately operating at scale.
You need to be able to assess the needs of your company and work to fit a solution to your data needs. This can mean coordinating integration across teams and working with those above and below you to get a realized solution in place.
3. Understand business context of your data
If you could just gather data and call it a day, running a data team would be a pretty easy job. But everyone that's tried their hand at managing data knows that isn't the case. You need to use your data in the context of the goals and vision of your business.
What changes do you want to see for your business in a month, a quarter, a year? What are the metrics that will help you measure your progress to that growth? Being able to figure these questions out is what takes data from a pile of numbers to actionable insights.
This can get even more granular by team. For example, Customer.io, an email campaign tool, has seen that key points in a customer lifecycle respond well to email (onboarding, new feature releases). These are the moments that are contextually relevant to your sales team building emails. Understanding the specific data segments that will answer a team's question is key to cutting down what they are expected to sift through.
4. Selective attention
Another challenge that has come with our newfound ability to track pretty much anything we want is discerning what deserves our attention. Especially if you have limited resources, picking and choosing what to analyze is a crucial part of building your company.
This builds on being able to put your data and metrics in your business context. Figure out where you want to go, and figure out the success metrics that will help you get there—even if you have a million ideas, keep your energy focused only on what you can realistically work off of. Andrew Chen, Head of Rider Growth at Uber sums it up nicely:
"If you’re not going to do something about it, it may not be worth measuring. (Similarly, if you want to act to improve something, you’ll want to measure it.)"
Don't skimp on the data you want to collect though. When in doubt, it's always better to collect more; just make sure your analytics stack can handle the load.
5. Know how to ask analytical questions
In a way, our previous skills lead directly to one of the biggest skills that any analytics person needs: knowing how to ask the right questions.
Data alone isn't necessarily what is going to set your company apart, or help you build a great product and thriving business. Analytics is about breaking apart that data to see what's going on under the hood. You need to be able to ask questions that will allow you to transform your information into hypotheses and theories.
This doesn't mean being able to ask the perfect question in one go—the one question that will hit upon an insight that will transform your retention or conversion; this means being able to ask analytically-minded questions of your data iteratively until you hit upon the insight you're looking for.
While this isn't exactly a skill you can pick up at the drop of a hat, you can help yourself build intelligent, analytically-minded questions by:
- understanding your product
- understanding the basics of data analysis
- being willing to think outside of the box
- working collaboratively
These will build the foundation for great questions.
6. Build models to communicate in your organization
Data is a tool that should be helping anyone and everyone in your company think more deeply about your product and processes. For that to be true, everyone needs to be able to access and understand that data, and what you're doing with it.
That doesn't mean that you simply distribute the final result of your analysis, or that you dump a bunch of raw data onto the company and expect people to know how to handle it. It means taking the information you have and making it accessible by simplifying it so that everyone can see, understand, and manipulate what you're collecting.
If you're building internal tools, this is something that you'll need to figure out. If you're picking a tool to buy, make sure it is something with an interface or options that don't alienate non-data specialists at your company from using it.
7. Basic mathematical probabilities
It seems crazy, but very simple mathematical concepts can confound even the brightest of mathematicians when presented in counterintuitive situations. That's because our gut feeling and the data are often pointing in different directions, and that makes it surprisingly hard to discern what is actually correct.
Part of combatting this is continually refreshing your basics—and going back to them when you feel like the data is telling you something that doesn't match up with your intuition.
This can help you avoid biases like Twyman's Law, for instance. Twyman's Law is that any statistic that looks unusual is wrong. That is, if you see something that's much more interesting than you expected, there's something behind it—for example, websites with no traffic from 2-3 AM on March 9, 2014 are simply seeing the effects of daylight savings time.
Being aware of the way that data is sometimes tricky for our brains to handle can stop you from making a foolish mistake.
8. Gauging success
For all the quantitative work you're doing in analytics, it can be easy to forget that you should be measuring your own success, too—how well what you're doing actually impacts a set of metrics or outcomes.
This is going to look different for different tools and different departments. Your sales team might need to evaluate how their CRM tool tracks customers, and whether that is working to help them meaningfully follow up. Your data team might need to figure out whether their analytics on new user retention are having an effect.
Encourage teams to take a critical eye to their own processes and set their own metrics for success—for example, it might be useful to explore the objectives and key results (OKRs) framework to focus in on how tools are helping team goals. (If you're new to OKRs, management tool Perdoo has a good checklist for getting started.)
Don't put into place a tool or a new process based off of your data without having a plan in place to figure out how well it's working for you, and don't hesitate to replace or iterate on tools or metrics that aren't working. It will save you time and build your long-term success.
9. Look at the “why” and not just the “what”
Your data is generated by actual people (or devices or other things). If you just look at data points you are missing a pretty big chunk of the story: the why.
Behavioral analytics can help you craft powerful hypotheses about why users of your product are doing what they're doing. Sometimes, it can be just as valuable to go straight to the source. To keep a holistic view of your data and a comprehensive process for your analysis, you can use:
- Customer research. Whether you bring in and talk to the customers using your product, ask them questions via webchat or conduct comprehensive surveys, hearing unfiltered customer experiences is one of the best and most direct ways to get insight as to why they behave the way they do.
- Session replay. Tools that allow you to see what a customer is actually doing in your product are invaluable for figuring out the 'why'. Session replay tool, FullStory, found that their own product helps them make leaps in their understanding—for example, they noticed that they were getting a lot of referrals from Evernote. By watching the beginning of the sessions of those users, they saw people were heading straight into their FullStory accounts, suggesting that current users were using the two tools in tandem, not that they were getting referred by Evernote.
Build your commitment to understanding how your customers use your product by going straight to the source instead of just relying on numbers.
10. Tell a story
Companies are full of stories. They build narratives around their internal values, they tell their customers a story to shape their brand, and their data should tell stories, too.
Stories all around us are informed by data—sports, news, politics—and data can be a great foundation for explaining things. But when you're using your data to tell stories about your customer and your company, you aren't just using data as a foundation. You're using narrative to hook people into what your data is telling them.
It's human nature to listen to stories. If you want people to pay attention to your metrics and insights, you need to be able to tap into that to make it happen.
11. Encourage adoption internally
Many data efforts fail because members of the data science team cannot get others inside the organization to adopt tools. Part of this comes with making data accessible, as we discussed earlier. The other part of this is also making sure that people understand what they have to gain from a deeper dive into analytics (and are empowered to dig into them).
We saw this clearly with BloomBoard, a SaaS tool for education. With Interana, they were able to solve problems that their users had with their previous analytics software. Because there was now a much easier path to getting the value from their data, BloomBoard saw people at their company using the tool across the board.
Never stop developing
This list is chock-full of interconnected skills that take effort to keep sharp. Mastering them—and mastering data—is an ongoing pursuit. There are always going to be places where you need to improve. And, as the field continues to develop rapidly alongside our technology, the people in charge of data will need to improve their processes and skills, too.