Case study

How Wolt drove a data culture revolution with Avo and Mixpanel

Avo and Mixpanel are a lifesaver combo because they allow us to process a large amount of data and quickly get an understanding of what's going on with user flows or check data quality.
— Elmira Gatina, Analytics Engineer

Over the past decade, Wolt has grown from a small startup to a household name for food delivery across Europe. A foundation to this success has been a culture of continuous improvement around their products and technologies, supported by high quality data.

Wolt is a leading food delivery platform that makes it incredibly easy to discover and get great food at affordable prices. Founded in 2014 in Helsinki, Wolt Operates in 30 countries all over Europe and is continuously growing into new markets. 

Underpinning Wolt’s product development is a commitment to high quality data. This didn’t happen overnight—it developed out of a data culture revolution spearheaded by Analytics Engineer Elmira Gatina and Head of Data, Jacopo Chiapparino. 

Going from broken data and data dictionaries stuck in YAML, to robust data quality and seamless implementation, Wolt underwent an internal culture shift to put event data at the heart of its product development strategy. With Avo and Mixpanel together, Wolt has its “lifesaver combo” that empowers analysts and product leads to get great insights, quickly. 

The Challenge: Overcoming broken data and wasted resources 

As Wolt’s platform evolved, its product team was keen to run experiments and discover new ways to delight users. The team relied on data based on a single YAML repository where groups of events were locked together in a single file. While this was straightforward for event owners to update, there was no clear visibility for other stakeholders and it was impossible for colleagues to decipher what the events referred to. 

We had a YAML repository with some groups of events that were locked together as one file, and it was super difficult to explain each particular use case when you had folks just adding values to a property.
— Elmira Gatina

To make matters worse, developers struggled to understand these event descriptions, leading to mistakes in implementation and ultimately broken or missing data. Only months later would analysts uncover these data quality issues, at which point it was already too late. Time, resources, and energy had been invested in fruitless experiments. On the data team, Elmira knew things had to change. 

Imagine you already started calculating experiments that ran for months. And then you see that you have no data to tell the truth and end up with wasted resources. 
— Elmira Gatina

As a Project Manager for Analytics, Elmira spoke up about the situation, advising colleagues to refrain from making “any decisions based on broken data”. It wasn’t easy, and admittingly “it took some guts” to face these challenges head on. But Elmira was determined to turn the situation around. 

The Solution: building a data quality culture around Avo and Mixpanel 

Led by Jacopo and Elmira, Wolt’s data team took it upon themselves to rebuild trust in their data. Finding tools to improve processes around data was a great place to start. 

Elmira discovered that her colleagues were exploring Avo, which enabled them to “quickly find events in the library, where they can just click on triggers and see what’s going on with the picture.” Getting a tracking plan in Avo was the first step towards a more mature approach to analytics, where defining, grouping, and interpreting events was “super, super convenient” for Elmira and her team. 

I came across Avo and tested things in a branch to see how it all works. I was super surprised it was finally possible to add pictures to events, because that was the biggest downside of YAML files. I found it was super easy to copy similar events and quickly create new ones. 
— Elmira Gatina

Elmira and her team were impressed with Avo’s functionality, and how easy it was to evolve new data structures. The next step was to drive wider adoption of Avo to facilitate a stronger data culture. Elmira led the charge, delivering “demos to engineers, to product teams, and to so many different parties that were not really involved in analytics”. 

Beyond Avo’s core Tracking Plan, Wolt’s team values Inspector for observability, and Codegen to ensure consistent implementation. The end result: Wolt can ensure consistent, expected data for data consumers throughout the organization. 

I really love the two sides of the coin, that we first have it as a data catalog. It's super amazing. And then also Inspector, where you can check if everything matches what we get also in Snowflake.
— Elmira Gatina

It’s one thing to deliver great data, but another challenge entirely to make data accessible. With Avo and Mixpanel in combination, Elmira says she has the best of both worlds: high quality data easily available. For her team, Avo and Mixpanel is a “lifesaver combo”.

The impact on the data team: time saved on analytics reduced by 40%

With a solid tracking plan in Avo, implementing analytics has become a much smoother process for the Wolt team. Elmira notes that the time analysts spend designing and deploying tracking has been “decreasing”  because “they learned the guidelines, they understand how to do it properly, and those who work with data quite a lot, they already do it without any changes almost from the first time.” 

Event data is now easily discoverable in Avo’s single source of truth, reducing the pain of exploring whether something was already being tracked or not. According to Elmira, this has been a huge time saver “they  now know where to search for it, they know their events, and I think we save lots of time on this”.

When it comes to uncovering insights downstream, data consumers can use Mixpanel to easily drill into user behaviors. In this sense, Elmira describes Avo and Mixpanel as “working together like brothers”, where one ensures great data quality, the other, fast insights. 

We try to lock our events, keeping in mind how they will be used in Mixpanel, especially by the stakeholders that are not very aware of the data itself and so on. It should be super easy for them to grasp the idea of the event name.
— Elmira Gatina

Although Elmira says there’s still refining work to be done “to fully finish the whole process”, there have been marked process improvements and significant time savings. According to Elmira, in terms of “the amount of time we work on telemetry” has sharply dropped, with the analysts already getting back almost half of their working time, and even more potential for the Wolt team to optimize their data processes with Avo.

The future of data at Wolt: evolving with optimism

Elmira is a realist. She knows that data culture is a constant evolution—not a one-time fix. That said, her recent wins have given her cause for optimism. 

This is slowly building. That's nice. So I wouldn't say in half a year we will be in a bright future yet. But for sure, if we continue like this, it will be just one of normal things for Product and Engineering to care about data and see it as a part of product development. So yeah, I'm optimistic.
— Elmira Gatina

As for advice she has for other data teams considering tools like Avo? It comes down to persistence.

If you want to improve your data quality, give Avo a chance and see. And never give up!
— Elmira Gatina


Food delivery

Company size



Helsinki, Finland

Tech stack

Mixpanel, Snowflake, Looker, and Avo

Key takeaways

  • Avo and Mixpanel: With Avo and Mixpanel together, Wolt has a "lifesaver combo" to empower team members with high quality data.
  • Better visibility and understanding: With pictures and event descriptions, Avo makes it easy for internal stakeholders to instantly interpret individual events.
  • Great data, fast: With seamless implementation through Avo and Mixpanel's intuitive UI, data consumers get reliable insights in a pinch.
  • Validation, made easy: Avo's Inspector makes it simple for Wolt's data team to ensure they're getting clean, expected data in Snowflake.
  • Significant time savings: Wolt now spends 30% less time implementing analytics.