Self-Service Shouldn’t Mean Self-Struggle or How to Build Analytics People Actually Use

Data-driven decision-making isn’t just a business trend that will fade into obscurity once the next technological advancement or cultural inflection point shifts leaders thinking about what to prioritize. Good data delivers results. Consequently, many of those leaders are requiring their teams to inform their organizations’ strategy with solid data analysis and business intelligence.

That obviously has many benefits for companies, but it can also create serious challenges for data teams fielding an avalanche of requests. At MDS Fest 3.0, Netflix’s Bhargavi Reddy Dokuru and Shruti Khanna spoke about their work with the company’s games team, which has significantly expanded since its inception to now include more than 100 games on the platform.  

The team relies heavily on information it collects related to metrics like installs and retention rates to improve the performance of their growing portfolio. And for some time, that meant constantly tapping the data team for information.

“My data team on average was receiving like 50 to 75 ad hoc data requests from cross-functional teams trying to troubleshoot or understand what’s happening with the performance of the game or the platform or crashes or stability,” said Khanna, a manager of data science and engineering at Netflix. “All of these requests, depending on the complexity, would take anywhere from a few hours to a few days or weeks to service.”

This unsustainable rhythm of requests needed some sort of solution, so the data team started exploring self-service tools that could increase the rate teams adopted analytics solutions (analytics adoption) while also making information available and accessible for everyone (data democratization).  

Spoiler alert—it’s been a huge success.

“Once we piloted the self-service analytics platform and rolled this out, on average, we now get five to maybe ten requests, but these are less on data and more on strategy,” Khanna said.

How did they empower their less technical people to start working with the data themselves? Let’s explore some of the important steps in their journey.  

How to make self-service BI usable

A pretty basic tenet of business is that you can’t sell a product without a market. And you won’t have a market if your product doesn’t solve a specific need. This logic isn’t limited to audiences outside your company; it also applies to your internal teams.  

Now, you may have buy-in from the different departments in your company about the importance of data. But when thinking about how to sell a self-service analytics platform, you should home in on the “self-service” aspect in order to figure out what will work best for potential adopters.

Khanna shared the steps her team took when deciding what could work:

Step 1: Identify a small group of power users

Think about the teams that are submitting the most requests for data. They are probably more familiar with the key metrics that will be useful to the organization and will have already identified some of their pain points with the existing data systems.  

Step 2: Understand their use cases

Again, creating a product that sells is all about understanding what the specific requirements are then solving for them. Spend time conducting interviews and focus groups with your power users and actually watch them work. It’s one thing to hear about pain points; it’s another thing to see them in action, which could provide you with telling insights.  

Step 3: Map out the requirements for a platform or tool

Once you have a good understanding of what you’re trying to solve for, you can actually start to identify what key elements or features are essential for making a tool that works for your teams. Think about what it would take to onboard nontechnical people and whether you can build something internally or use an external platform.  

Step 4: Build and test your tool

Whether you choose to create an internal tool or seek out a product elsewhere, you’ll want to demo that tool with your power users and gather their feedback on performance. That way, you can make the necessary tweaks before you start to roll out the self-service tools more broadly.  

Comparing potential self-service analytics tools

Returning to step three, you might be wondering how to actually define those parameters. The truth is that there’s no set list of requirements that will work for every use case or for every company. But to give you an idea of how you might think about this, the team at Netflix shared their five must-haves for a self-service analytics platform:

1. The ability to define a semantic layer that could combine complex data sources from different business units into one data model.  
2. A user-friendly interface that would allow nontechnical users to easily join different dimensions and measures to get the insights they’re looking for.  
3. The ability to easily create visualizations, simple charts, and grids to understand or tell a story with the data.  
4. The chance to share views of their transformed data while collaborating with their immediate teams or larger groups.
5. Some levels of data security, governance, and compliance to establish trusted data access across the company.  

The Netflix data team chose to create a tool internally, but there are terrific platforms like Domo available that can meet your company’s requirements.

Building a semantic layer for self-service BI  

Now, if you’re one of the less technical readers, you might be wondering: What is a semantic layer in analytics, and why is it so important?

“When a business consumer wants to know how a metric is trending or they need access to the data, they don’t really need to understand where the data is stored or how it's being generated and how the metric is being computed,” explained Dokuru, a Netflix senior data engineer.  

She said to think of a semantic layer as a logical layer that sits on top of your underlying data and acts as a bridge between your data sources and the people using the data. “It abstracts your business definitions and the computation logic away so the end users can focus on accessing the metrics that they need to help with decision-making,” Dokuru added.

To put it more simply, the semantic layer is what allows all the raw data stored in different sources to appear as a unified, consistent, and understandable data set within your business intelligence platforms. This makes it much easier to analyze and visualize data for everyone in your organization, even those with nontechnical backgrounds and roles.  

For example, all of those disparate sources may label certain metrics differently despite the fact that they're referring to the same things. The semantic layer can use metadata to create a simple and clear metrics’ vocabulary for everyone. This ensures that metrics are calculated the same way and defines the relationships between data sets or points, while also keeping access to the data secure.  

Ultimately, it’s not enough to give people access to data. If they don’t understand it or can’t easily use it, they won’t. Semantic layers are critical to building usable, trusted self-service analytics. Employing tools or platforms that have built-in semantic layers can keep everyone across the organization working from the same standardized data and can be especially helpful for companies that are scaling up and growing a more complex web of data sources.  

Winning with self-service analytics

Investing in self-service tools has paid off for the team at Netflix. The intuitive and user-friendly interface of their tool has helped team members make decisions faster and solve problems almost immediately.

“For my team, it’s really helped in saving costs and resources,” Khanna said. “And as a whole, it’s helped in promoting a data-driven culture across the organization because when we piloted with an initial set of users, we got a lot of… questions from a broader business group showing interest in trying to use these kinds of tools for their business use cases as well.”

But it takes a village to build and create a good semantic layer. This will include business data stewards who define metrics and own the business logic; data teams that build foundational logical warehouses to maintain and share this logic; and a company culture that cultivates the use of standardized metric definitions as well as the teamwork to maintain it.  

If you don’t want to build from scratch, consider how Domo can help. Domo enables teams to deliver self-service analytics that scale without sacrificing usability. With Domo, you can bring all your data together into one unified view, empower business users to visualize and analyze data, and ensure teams have access to the right data so they can self-serve without risk.  

Are you ready to discover the power of self-service data analytics? Connect with Domo today to learn more about how we can help change your data reporting game for good.