We recently caught up with Guillem Pons, Head of Data Science at Badi, about why they decided to switch to Adjust. Since making the switch they’ve been able to build some pretty incredible machine learning capabilities in-house, so we were keen to find out what they’re building, why they built it, and how things have changed since working with Adjust.
Could you give us a quick background about Badi and what the app does?
Guillem: Badi was born from necessity. Searching for a place to live can be quite a nightmare, especially if you’re looking to share your place. Maybe the apartment seems nice but the roommates are weird, and vice versa. So we created Badi - an app for matching the perfect apartment and roommates simultaneously.
We developed advanced algorithms to match profiles according to common interests and geolocations. With live chat functionality, users are able to connect instantly. Our app is highly visual, so apartment seekers feel like they see the room without the need to spend hours viewing actual physical spaces.
Our solution benefits both sides of the apartment equation - the ‘seeker’ and the ‘lister’. When we first started it would take 22 days on average for someone to find a new apartment - now we’ve reduced this to just 10. There are no middlemen needed either, which saves time, so we have a lot of happy users.
Badi started as a mobile first solution, but because of the great success we’ve had in helping people quickly find their dream apartment and roommate, we also developed a website and we’re continuously iterating on our algorithms to develop even higher quality profile matching with the aim of reducing search time even further.
Being mobile first means you really understand the need for having a solid attribution solution in place. What made you switch to Adjust?
We experienced rapid growth of our app user base and when we launched our website, we realized we needed to build a larger and more sophisticated data infrastructure.
We needed to see the full user journey and tie our mobile and desktop data together. On top of that, we needed to make sure we were collecting every relevant data point in a privacy compliant way. We were working with another attribution provider before, but their APIs didn’t give us the data we needed, so we were really happy when we started talking to Adjust and realized we’d be able to track everything without limits and this could all pass back to our own database. We’ve been using this data to build better algorithms, and the marketing team uses it to optimize.
So how exactly are you leveraging Adjust?
The thing that makes our solution unique is that we connect both ‘seeker’ and ‘lister’ profiles. To be able to make a match that will work long term, we need as much data as possible. This goes beyond just having someone’s name, gender, and location, but actually looking at the types of hobbies someone has, the activities their active in and more.
Thanks to Adjust we track more than just the marketing campaign or user acquisition source that brought the user to us, but also every single in-app event down to the user level.
We are receiving granular information like a custom login ID from Adjust, as well as information about the types of searches users are doing, the creatives they engage with, and more. We merge all data on our backend to track user activity over time, so we can shape user profiles and influence our own product development.
When we target new users to use our app, we rely on our predictive models we built with Adjust data, guaranteeing the quality of the targeted audiences.
Facebook or Google Adwords is a great place to advertise, but unless you’re using some sort of analytics tool alongside it, it can be a black box. You create a lookalike audience and hope for the best. Now I am capable of performing my own in-depth post campaign analysis. We know with absolute certainty if the users we acquired are exactly the ones we paid for.
Analyzing the data and updating the marketing campaigns allowed us to decrease our CPA by around 30 percent month on month.
Have you had a chance to try out our Audience Builder?
Yes! The idea of creating targeted user audiences with our own data and just uploading it to Facebook to make lookalikes is incredible. I also want to try intersecting the data we pull from Audience Builder against the data collected in our backend to see the efficacy of my predictive models.
Knowing the audience size helps the marketing team to double check our predictive model suggestions. It could be that the user profile I had in mind might have an extremely small sample size.
What are the next steps for Badi?
We have big plans for our app users. We're developing new technology that will allow us to track user relationships and enrich them with data from 3rd parties. For example, you and I are friends, and a mutual friend updates a room on Badi. Because, well, you know me, the probability that our mutual friend will be matched with you is higher.