How ZOE built a golden source of truth for the entire organization with Holistics

Valued at $250M (as of Mar 2023), ZOE helps people transform their health through smarter food choices. By combining their world-leading science with personal test results based on the bodies’ responses to food, they deliver personalized nutrition and ongoing coaching. This approach has positioned ZOE as a leader in science and nutrition, improving lives across the UK and globally.

However, ZOE’s rapid expansion and its commitment to empirical, data-backed product development began to overstretch the capabilities of the existing BI tool, Looker Studio (formerly known as Google Data Studio). Despite being free, Looker Studio struggled to meet the growing demands of ZOE’s complex analytics environment. Critical data operations, such as centralizing data definitions, became umbersome and time-consuming, creating bottlenecks that impaired the data team’s ability to support business operations.

Craig Wilson, ZOE’s Staff Data Engineer, first used Holistics during this period of rapid growth for the company, as the company grew its team from 50 to 500 employees over the next 3 years.

Questions posed by business users kept pouring in - from “What does a metric mean?” to “How to calculate certain metrics". With 18 years of BI experience, Craig knew they needed a better way to address this pressing demand for data.

There was a reasonably high appetite for having access to data, but also not a lot of maturity in regards to what that means, or the best way to manage it is.

We don’t have enough engineers to create reports for everyone. They’ll see a jump in metrics or something and they’re like “What does that mean?”, and you’ll have to see it right, go off and find that for me, or do some analysis rather than go off and do that themselves.

– Craig W., Staff Data Engineer

More reports were created, intensifying the bottleneck as their previous BI tool lacked a central place to define and manage metrics. Each report has its definition, nothing is centralized, and no one knows what the right number is, because everybody defines it differently, with little reusability.

We needed to move off of that because each report had its own definitions and nothing was centralized, which made it hard to maintain.

That was a big problem for us, which was why our main goal was to find an approach to define analytics logic centrally and have people reuse it so you can get to a faster speed to insight.

– Craig W., Staff Data Engineer

Looker Studio’s rigidity further complicated the team’s efforts. Craig found his team struggled with joining and merging multiple large datasets, making it harder to curate accurate, consistent datasets for self-service.

It’s all about self-service. Google Data Studio (now Looker Studio) isn’t self-service. It shows you some reports and you can’t join tables together. It has to be on One Big Table. We’re joining three massive data sets together to create a single dataset to work on. One fell to another and they wanted to see the ones that hadn’t been filtered out. The other metrics change if you add them in because of the different joins. It just becomes very effortful to create these metrics correctly.

– Craig W., Staff Data Engineer

Being a veteran in the analytics industry, Craig immediately recognized the value of centralizing data models for accuracy and consistency in metrics definitions. This insight guided the team’s search for a new BI tool, emphasizing:

• A Kimball-based semantic modeling layer for centralized metrics management.
• A self-service exploration layer, with a good variety of visualization options.

So our main focus was to have a centralized place to define models and metrics and not have teams define it repeatedly. We wanted one central place. That was our main goal.

We also want our business users to be as self-sufficient as possible, to create week-on-week metrics, trend analysis, and conversion metrics by themselves rather than relying on our engineers to do that for them. We want our engineers to focus on creating high-quality trusted data for them.

So when we started looking for a BI tool, we wanted something with a self-service layer. We don’t want it to be too technical. We don’t want people to have to write SQL to do it. We want everyone to explore data.

– Craig W., Staff Data Engineer

Craig evaluated all industry-leading tools with modeling capabilities, including giants, including PowerBI, Looker, Qlik, Thoughtspot, and Tableau.

However, most of those tools only have modeling capabilities that touch the surface area and do not provide the best-in-class depth. He eventually narrowed down his options to Looker and Holistics - both are architecturally similar, with self-service exploration on top of a code-based semantic layer. The choice of Holistics was swayed by its attentive customer support for mid-market customers, contrasting with their experience with the other vendors.

Additionally, Craig found Holistics’ analytics as code philosophy, which embraces DevOps best practices for analytics, particularly resonating with ZOE’s software engineering ethos.

So our main focus was to have a centralized place to define models and metrics and not have teams define it repeatedly. We wanted one central place. That was our main goal.

Everything about AML (Analytics Modeling Language) and dbt integration resonated very very well with our software engineers as they can write code to describe BI logic, check them into Git version control, and make the whole process more maintainable.

– Craig W., Staff Data Engineer

Holistics’ Semantic Modeling Layer enabled the team to define models once and reuse them anywhere, anytime. This approach helped reduce repetitive SQL queries and ensured that metric definitions were consistent and accurate across different reports.

Such a structured environment led to clearer insights and more confident decision-making, which is critical in the nuanced field of biology and nutrition where precision is everything.

When it comes to other BI tools, you have to do things over and over again, it is quite messy. When you get more reports and data sources, especially if your analysts do their own thing, they would come up with different numbers because they were taking the data from different places. So I felt it was less controlled. [..]

With Holistics, we can have it all under control, we can have correct measurements and correct metrics going out there. I think that’s where Holistics’ modeling layer comes in because we can have things predefined, we can take that from the correct data sources that we build in dbt and have better quality metrics in there.

– Craig W., Staff Data Engineer

Holistics grew to be the definitive source of truth at ZOE. Analytics workflow processes are set up to promote ad-hoc SQL and Python analysis to reside in Holistics after their long-term relevance to the business is established.

Once an analysis proves its worth for long-term application, it is imperative that we integrate it into our central data system. We consolidate these insights in Holistics, establishing a routine of accuracy and oversight. This not only streamlines our operations but also fortifies our data processes, ensuring that every piece of analysis is thoroughly vetted and reliably executed.

– Craig W., Staff Data Engineer

By centralizing these analyses, Holistics facilitates a seamless transition from exploratory data handling to a more dependable framework. This approach allows for faster feedback loops and agile decision-making, with the added benefit of creating a comprehensive, centralized repository for reusable data assets.

Craig’s data team also found Holistics’ as-code approach useful, as they understand the value of applying engineering practices to analytics development. One of their favorite features is Git Integration, which helps the team easily track who changes what when, and roll back when mistakes happen.

We’re now more confident when changing something. Because if you make a mistake, you can go back. Being able to keep track of that sort of stuff is great. You can understand the changes and you can track how things are going and you can apply good software development practice to analytics. Without it, you have no idea what’s going on.

– Craig W., Staff Data Engineer

With Holistics, dashboards have become integral to daily operations, encouraging more report creation, data exploration, and data utilization across all departments.

Teams have created a lot of reports and found it easy to do so. Marketing report gets sent to our Slack channel every morning whereas our engineering team builds dashboards that help them monitor their key metrics.

– Craig W., Staff Data Engineer

Holistics also made it easier for everyone at ZOE to create, explore, and share interactive dashboards. This opened up more dynamic discussions about users’ well-being, cultivating a more collaborative approach to nutrition coaching.

Our Period-over-Period (PoP) dashboards show our users their progress in various health metrics, such as energy levels, adherence to plans, and overall well-being. We’ve created our daily biology reports Holistic and our coaches can just log in to Holistics, search for that user, show the analytical breakdown, talk them through their progress, and offer personalized advice.

– Craig W., Staff Data Engineer

PoP (Period-over-Period) comparison is particularly favored by the Product and Marketing teams. Being able to compare any metric across different periods with just a couple of clicks is a game changer for business users. Without this feature, everyone would have to endure the cumbersome process of exporting data to external tools for comparison.

Holistics makes it easy to do QTD/QoQ and YTD/YoY reporting without custom logic and formulas. Just point-n-click!

PoP is big with the product and marketing team. We’ve got a data scientist who had to write a big piece of SQL in Looker Studio just to do that. That is frustrating because that meant the business didn’t have this number for a while

– Craig W., Staff Data Engineer

With Holistics’ Usage Monitoring Dashboard, the data team has a bird-eye view of the data interactions at ZOE. They can identify under-used dashboards, prune redundant reports, discover data champions, support struggling data novices, and share best practices, improving the overall data proficiency across the entire organization.

With Usage Monitoring, we can see who uses what and how. If there’s one person who’s using Holistics every day that we didn’t expect them to, we can go to them and ask what’s good, and what’s bad, and share their practices across companies. Or, if there’s someone who’s supposed to use the tool often but didn’t use it much, we can reach out to them and see if they need more training.

Because if you bring a tool into a company, you want people to use it. So if you don’t know if they’re using it or not, you can then step in and help them understand why it’s beneficial to them.

– Craig W., Staff Data Engineer

At Zoe, certified nutritionists guide users in understanding their microbiome health. They advise and coach users on how diets affect their body, sleep quality and energy level. To do so, they need data, and they need it fast. However, before Holistics, accessing comprehensive user data required navigating multiple systems, a time-consuming process that could damage customer support quality.

With Shareable Links, Holistics enabled ZOE to maintain a more proactive stance on sharing data. Nutrition specialists are now able to get relevant insights faster to support their coaching efforts. Holistics also ensures that critical data points are monitored regularly, and updates are delivered automatically, keeping all stakeholders informed and enabling timely business decisions.

It’s all in one dashboard.

Previously they had to go to eight different systems. So if a user contacted them, they’d have to go into different systems to check all this. They are now brought together into one single report in Holistics which they can go and view. That’s been a really big benefit of the company.

– Craig W., Staff Data Engineer