The Leader’s Guide to Being Data-Driven in 2021

75% of Fortune 1000 companies are struggling to forge a data culture. If you’re one of them, read on for lessons learned from industry leaders at top data-driven companies like Netflix & Amazon, plus tips to help companies of any size on their own data journeys.

And if you’re part of the 25%? Let us know what’s worked well for you or what you’d like us to write about next. 😉

“How can we be more like Netflix?”

It’s the question I hear the most. I get asked after keynotes. In interviews. Over email, Twitter, or coffee. On a plane. At the gym. Twice in public bathrooms.

Before launching Noteable, I led the Big Data Tools and Data Core teams at Netflix. My role was to add “high impact through strategic innovation.” It was a dream job at a dream company, and I loved every minute of it. To execute on my vision at Netflix, I worked closely with ridiculously talented colleagues at the world’s most data-driven companies. And I learned a lot along the way.

Few companies have reached Netflix’s level of analytics maturity. To briefly illustrate what I mean by that, take a moment to add up the number of employees you have in data-focused roles. Include anyone who works directly with data (e.g. data scientists, analysts) as well as anyone in related support functions (e.g. data infra engineer, data product manager).

What percentage of your employees work directly with data? If you’re like the vast majority of established companies, it’s less than 1%. When I first joined Netflix in 2016, it was over 17%.

Take a moment to think about what that really means: 17% of all employees were responsible for helping Netflix use data to inform decisions and execute on strategies. As counterintuitive as it might seem to invest so heavily in data, the strategy paid off. Despite having only ~4,500 employees at the time (it’s tripled since), revenues hit $8.8 billion USD that year — making it one of the highest revenue-per-employee companies in the world.

That’s the power of well-informed decision-making.

Netflix isn’t successful because of one or two key things they’ve done that can be easily emulated. Their success is the culmination of a strong & strategic executive team, a uniquely powerful corporate culture, and mind-boggling investments in data–both people and infrastructure–over the past decade.

Here’s the good news: You don’t need Netflix-level investments to be good at data.

Through our new blog, we’ll help you cultivate your own data culture and evolve into a world-class, data-driven organization. My team and I have worked on some of the most sophisticated data infrastructures in the world, both at leading data-driven enterprises and on groundbreaking, open-source data technologies like Jupyter and Python. We know what it takes to succeed with data, and we’re openly sharing that knowledge with the world.

In this first post, we’ll share three high-impact initiatives for getting more out of your existing data investments.

The best part? You already have everything you need.

1. Cultivate a “one team” mindset.

Breaking down organizational silos has been a perennial challenge for executives. Unfortunately, it’s also a prerequisite for forging a healthy data culture.

Whether running a full-stack army-of-one unicorn or a sizable decentralized organization with specialized data experts, leaders often fail to identify and engage their entire data team. This is a recipe for disaster. How can you support your teams when you don’t even know who all the players are? How can your teams prioritize and effectively collaborate if they don’t even realize they’re part of the same team?

Fix that.

Data is a team sport, and you need to function as one team in order to win — a single extended team who is collectively responsible for making your data work for you.

Teamwork is arguably the most important prerequisite for being data-driven. That’s because it takes everyone working together for a single project to succeed. Look at the complexity of interaction in this diagram.

You may not have some of the specialized roles in your organization, but you have someone doing all of those functions. It’s your responsibility to break down those silos and normalize communication across teams and functions.

The more you champion this “one team” mindset, the more your people will self-organize their work in a collaborative fashion, leading to greater consistency, greater quality, and faster delivery on their projects. We’ve seen these results firsthand.

How to get started?

Identify your core data team. Begin with the obvious: your data scientists & ML engineers, data & platform engineers, BI developers & database admins, and any analysts who connect to your data sources directly. Essentially, anyone whose primary job relates to processing or consuming data. These are your data experts, the folks who make up the core of your data team.

Build your “one team.” Everyone upstream and downstream of your core data team should be represented here. This includes any software engineers, DBAs, TPMs, or anyone else directly involved in the creation or definition of data. If they can affect data structures, they’re part of the team. If they can materially affect data quality, they’re part of the team. Stakeholders are generally captured in a separate group, although exceptions are common — for example, product or engineering managers representing their stakeholders’ technical needs.

Break down silos and improve communication. Your team needs to be aligned on what you’re doing. This challenge isn’t unique to data, so your existing people leadership tactics work well here. Get everyone in the same room, physically or virtually, whenever you can. Hold team building events, do a quick icebreaker at cross-team meetings, and be intentional about bringing folks together. Above all else, make it interactive and engaging.

Take action to facilitate collaboration. One especially successful initiative at Netflix was Demo Day after the monthly All Hands. Teams were invited to set up “demo booths” (in reality, tables) in a style vaguely reminiscent of high school science fairs — only with kombucha and gluten-free snacks. Demos came in a variety of formats. Data engineers would excite us with new datasets, data viz engineers would dazzle us with new dashboards, and data infrastructure engineers would wow us with new APIs. Even PMs and TPMs participated, holding impromptu “office hours.” It’s impossible to quantify the benefit this had on the organization, but it was significant and tangible.

2. Pick an audacious goal and establish a “Data Success” team.

Winning teams need a common goal to rally around, and your team is no different. Give them one. Identify a BHAG, or a big hairy audacious goal, for your entire data team. This should be something bold, cross-functional, and impactful. Make it abundantly clear that the entire team is on the hook to deliver.

Once the goal is set, put together a Data Success task force. Populate it with your leads or top performers, with a representative from each group or functional area. This dream team helps to drive the overall success of your initiative and should be given authority to act accordingly. Their primary function, at least initially, is to facilitate fluid communication and collaboration across teams. They should also track progress against goals and report to your company’s senior leadership team on a regular cadence.

Let’s walk through an example together. Imagine your analytics team has just set aggressive goals for completing new stakeholder requests. Of course, not all projects are created equal, so you’ve matrixed turnaround times by project type, complexity, and business criticality; your new goal for “small ad hoc data requests’’ is one week. Everything’s on track until your analyst discovers an invalid datetime format in a JSON blob that’ll take a couple of days to fix.

Don’t fix it.

Instead, have your analyst reach out to the Data Success team, who would find and connect the upstream owner with the analyst. The upstream owner, a software engineer in this case, would be responsible for correcting the bad format and overseeing downstream reprocessing. And if the upstream owner is unclear on priorities? The Data Success team would escalate to leadership for an answer.

This powerful combination of increased executive visibility, greater individual accountability, and deeper context of business impact will naturally lead to greater productivity, both individually and as a team. It also offers critical visibility into bottlenecks your team is facing, providing the opportunity to take swift action. Yes, there will be escalations and hard decisions. Yes, in the beginning it’ll be slower to have the owner fix it. Yes, your stakeholders will be affected. And yes, it’ll take time. Lots of valuable time. But if cultivating a data culture is truly a priority, you’ll need to allocate your time accordingly.

3. Establish success-focused standards and evangelize consistency.

The road to standardization can be long and daunting. It takes time to properly assess the current state of your entire data environment, establish new standards that meet the unique needs of your business, oversee the rollout of those standards, ensure new data assets adhere to the new standards, and address your newfound technical debt of existing assets newly out of compliance.

Having been through the process a time or two myself, even writing about it feels draining. Is it any wonder companies avoid these projects?!

While a standardization initiative may not sound exciting, the results certainly are. Standardization will have an outsized impact on your company if done properly. A few things to keep in mind:

Start with small, noncontroversial standards. For many, even the thought of this undertaking is enough to elicit cold sweats and heart palpitations. This is especially true the larger or more bureaucratic the organization. Take heart: you don’t need a “big bang” rollout to make impactful progress. If you have no standards today, start with what’s least controversial and easiest to roll out. Can you find agreement on datetime formats? Establishing even that one convention with upstream source teams will have a massive impact on life for your downstream analytics teams.

Match a standard to its business impact. Standardization should be done with purpose. One common mistake I’ve encountered is the “all or nothing” approach. That doesn’t work, because you shouldn’t require the same rigorous standards for a quick experiment that you would for a mission-critical dataset. Doing so only adds more work for your people and increases friction in your processes, all for no measurable gain. Instead, consider establishing a promotion path with clearly defined standards for each type of project at each phase of its lifecycle. For example, expectations for a dataset should evolve as the dataset is promoted from development, to beta, to production. Remember, each standard comes with an incremental cost — a tiny cognitive tax on your team that grows as the number of standards do — so make sure to always align your standards closely with business impact.

Invest in quality automation. The more you automate, the easier it becomes to change standards and deploy new ones. Your goal is to catch the violation as early in the workflow as possible. As the saying goes, an ounce of prevention is worth a pound of cure. This is definitely true for data. For example, running a nightly schema crawl to ensure new datasets are adhering to naming and datatype conventions? Great job! You’ll catch errors quickly before downstream teams become dependent on those assets. But validating the code at runtime or upon commit to prevent the exception from making it into your schema in the first place? Even better.

Actively promote new standards with the *entire* team. Make sure folks outside of the Data Success team understand the new standards and motivations behind them. You want their buy-in, so focus on the benefits to them, such as the countless hours they’ll save by enabling their users to self-serve their own data needs. Remember to also deep-dive into business impact. The consistency these standards provide makes it significantly easier for new users to self-serve and onboard — another key prerequisite for a data-driven culture.

Remember: the greater the consistency, the better the scalability!

Adopting a data-driven culture has become an organizational imperative for most companies, especially industry incumbents. The main blocker today is no longer technology, but rather people and processes. These three strategies — cultivating a “one team” mindset, picking an audacious goal for your “Data Success” team, and establishing success-focused standards — get you moving in the right direction.

The journey to becoming data-driven is long and arduous, but well worth the effort. By the time you’re done, data should be informing decisions across every aspect of your business. And better informed decisions generally lead to better outcomes. But it doesn’t happen without effort, and it doesn’t happen overnight.

This post is the first in a new series for data leaders focused on enabling the data-driven enterprise. It’s also the first from our team’s new blog. If you’re serious about becoming a data-driven organization, consider subscribing. Topics will include tips, best practices, and lessons learned from industry leaders on everything from data-driven leadership to up-leveling data literacy, data engineering to data visualization, and every topic in between.

Thanks for taking the time to read. As always, we welcome your questions, comments, and feedback. Also, feel free to leave a comment with ideas for what we should write about next. :)

Curious about what we’re up to at Noteable? Check out CTO Matthew Seal Noteable: The Interactive Notebook Document for Modern Data Teams.

Interested in learning about our approach to data visualization in our product? Noteable Chief Visualization Officer Elijah Meeks shares his thoughts on Designing for the Data Visualization Lifecycle.