AppsFlyer builds a data-drive culture with Looker and Google

Profile picture for user ddpreez By Derek du Preez April 1, 2021
Summary:
AppsFlyer has gone from one general BI dashboard to everyone in the company using multiple dashboards on a daily basis, as it centralised its analytics on Looker.

Infographic display with hands data storytelling © Quality Stock Arts - shutterstock
(© Quality Stock Arts - shutterstock)

AppsFlyer provides some of the world's leading brands with marketing data that aims to offer a holistic view of every user journey across platforms, channels, and devices. It has raised hundreds of millions of dollars in funding, counts Walmart and Nike as some of its customers, and in recent years has grown from 200 to over 1,000 employees. 

However, approximately three years ago, the company realized that whilst it was providing cutting-edge data solutions to its customers, internally it was not adopting the principles of a data-driven culture. It was sitting on petabytes and petabytes of data, and yet the company simply had one general BI dashboard for use across all operations. 

AppsFlyer prides itself on putting customer success first and foremost, and so it realised that this needed to change. Enter Alexandra Sudilovski, Senior BI Expert at AppsFlyer, who joined the company to build a proof of concept that would kickstart a data-driven approach across the whole business, making use of Google-owned BI platform Looker. 

Sudilovski was speaking at Google Cloud's Born Digital Summit this week, which can be accessed here. She says: 

Just to give you some numbers, we are processing 120 billion events on a daily basis. We have about 90 terabytes of data in our AWS S3 buckets. We also use BigQuery for our row data events and we currently have about 40 petabytes of data in BigQuery. And we keep growing very very fast.

About three years ago the company started to grow and we saw that besides the great product that we provide, we need to measure ourselves as a company. We want to provide the best solution possible and we want to measure ourselves, we need to be able to see dashboards and see the full pictures of our data.

A change

The first challenge for AppsFlyer was finding a BI tool that could handle such large amounts of data, with the ability to process and analyse billions of events and petabytes of data on a daily basis. It also has a centralised data lake that makes use of different data sources, coming from AWS and Google Cloud - where it needed to be able to see all this data in one place. 

After assessing the options, Sudilovski says that Looker was the best tool that could meet all of AppsFlyer's needs and once chosen, a production environment was up and running in a couple of days, with the first dashboards being deployed. She adds: 

What we really loved about Looker is the ability to create data modelling and the business logic, between the data source and the dashboard, so we can create and measure KPIs. We can create many,many models, and basically provide self serve BI.

However, setting Looker up was only the first step towards changing how the business operates, using data. Shortly after going live Sudilovski realised that the scale of the project was more than she could handle on her own, given that she found herself answering user questions day in and day out on Slack channels, on email and in direct messenger. 

To tackle this, AppsFlyer decided to adopt a ‘Guild' approach, which is essentially a knowledge centre and centre of excellence for all data users across the company. Sudilovski explains: 

Let's create a team that will be cross-company that will help us in our governance, for everything that's happening in Looker. I took time to educate those people inside the Guild, created training materials to help them, and to give them all the support and all the tools necessary for them to be able to work efficiently on Looker and provide the analytics solution to the other stakeholders in the company. And basically they're now responsible for supporting other teams.

So we have a sync once a month.  Every new version that we upgrade in Looker, we go through all the readme and older release notes. I make sure they are aware of every new feature that is coming out of Looker. I create examples and I put it in our repository or in the deck we created or inside the training. 

We have a dedicated Slack channel, where we can share with each other the work we do. We can ask questions, we can help each other navigate the 1000s of views and models that we've created during the past three years. We share knowledge and we share data. We ask questions, we help each other.

Challenges

Sudilovski reflects that three years ago there was just one AppsFlyer dashboard that everyone at the company saw. Now there are more than 1,000 employees at AppsFlyer, all of them using Looker on a daily basis, getting push notifications and reports to their email or Slack. She says: 

We're now a data driven company. We don't do anything without looking into the data. We use this tool to help our customer success and sales teams to do the best that they can, and also we use Looker to measure ourselves in every possible way. All the business operations work and see numbers on a daily basis in Looker, they rely on those numbers. All our finance and collection and billing, they rely on the numbers in Looker. 

We use a looker to measure our operations department, our support department, our business department, IT department, legal department. We at each point in time how things are going, how we can be better, how we can maybe change, how we can make our processes better. Looker became one of the number one tools at AppsFlyer for use on a daily basis and the adoption was incredible.

Sudilovski says that the key to getting this right is to approach such a project with an ‘open mind approach' and to be willing to adapt quickly. She added that if mistakes happen, not to view these as ‘mistakes', but rather as milestones that you learn from. In addition, she urges people to rely on support from Looker and Google, which have many customers doing the same and will be willing to give advice. 

One such change that AppsFlyer had to make was centralising its data analysts into a new team, so as to create consistency across the business. Sudilovski explains: 

One of our biggest challenges during the last three was that we worked in a decentralised approach, where we have analysts in every business department, dedicated to that department. We used a self serve BI approach, meaning the BI team is providing all the data, making the data available and accessible to all the analysts, and to Looker. They were then creating the models, developing the dashboards for this, for the dedicated department. 

The interesting thing was we saw that we have multiple measures for the same KPIs, but with some different numbers. Every department was using their own logic and their own calculations for the same KPIs. So how can we make it aligned across the company? So we create a new department at AppsFlyer, the business analytics department. So the business analytics department now is the owner of all cross company KPIs, they're deciding what will be the right calculation, they're deciding on the logic, and then they're building the base in Looker. Then we extend it for other projects for other business departments and analysts to use, so we make sure the logic is in one place.