After you release a product, you have to continuously (and carefully) track its performance. Who is using it? How are those users behaving? What are your product’s strengths? Weaknesses?
You track all this by gathering factual data spread across different platforms, found in various formats. It’s fragmented and not at all optimized for structured analysis; trying to compare metrics across platforms leads to misleading results and errors.
One way to organize and navigate the whirlwind is Data Intelligence, a combination of a Data Warehouse and Business Intelligence. A similar standardization and report automation solution was used by Luftansa, which reported a 30% increase in organizational efficiency.
How It Works
A process created by STRV’s Data Science (DS) team, Data Intelligence collects and neatly stores all data in a single system called the Data Warehouse, which is pretty much a single source of truth that reflects reality with maximum accuracy. There, it optimizes everything for accurate data analysis and usage.
Once all data is organized in the Data Warehouse, we use Business Intelligence to build the necessary dashboards and reports for tracking. It’s also great for ad-hoc data analysis which is usually done (more painstakingly) in Excel.
The whole thing is set up and customized by our engineers — in cooperation with you, to make sure it’s all fully adapted to your company’s needs.
What It Means for You
With Data Intelligence in place, you can make business-critical decisions faster because you have advanced historical analytics, product analytics and operational KPIs unified in one place, telling you exactly what you need to know.
Specifically, you’re able to identify exact ways to increase revenue, profit and operational efficiency; analyze and understand user behavior; compare data with competitors; track performance and predict success; discover issues and spot market trends.
All reporting becomes faster and more flexible, and it’s a lot cheaper than other solutions.
A typical Business Intelligence license costs around $25,000 a year, and all that gets you is a tool for data visualization. There’s no Data Warehouse, which means it’s not really solving your problem of properly understanding and tracking data. You still have to store the data somewhere, consolidate it or create some sort of model… which requires more tools, more investments, etc.
How It Prepares You for the Future
Taking good care of your data straight away creates a sturdy foundation for all future DS initiatives, including AI and Machine Learning (ML) integrations — which you will need at some point. Technology is running the show in today’s world, and all businesses have to keep pace.
Sourcing data for such projects is about 80% of the effort. Our Data Warehouse can reduce this effort by 2/3.
Use Case: Cosmic Latte Dating Apps
Cosmic Latte is a spin-off company under STRV Labs that operates a portfolio of LGBTQ+ dating apps — Surge, Zoe and Grizzly — which have been downloaded more than 10 million times on iOS and Android. While successful, there were two areas in need of optimization: Data compilation and the matching algorithm. Both were tackled with Data Intelligence.
Head of STRV Labs Jan Pacek and data engineer Jozef Reginac explain.
Challenge #1 - Data Compilation
Our first goal was to better understand user behavior. Google Analytics is a very powerful tool, but it has its limits. With a Data Warehouse, we could combine detailed user activity data and precise location data (down to a city level) using subscription data from the App Store and Google Play — things that can’t be done well with Google Analytics.
To be specific, our focus was on financial KPIs and activity KPIs. From the financial side, we needed an accessible general overview for tracking monthly revenue per app, MoM revenue growth and subscription rates. A key issue was that Apple’s Fiscal Calendar sends payments in time frames that don’t always coincide with the standard calendar month, which distorts performance results.
From the activity side, we were looking for a clear rundown of retention and swiping activity — swipes, likes and matches.
Solution
We started by establishing which data sources we were working with. In our case, they were our own database, App Store Connect, Google Play Console, Firebase and Google Analytics. We then mapped out where we have which kinds of data.
The next step was building a Data Warehouse. We did a sample load of raw data from the backend, app stores and Firebase, and then got into data modeling for the activity and finance use cases.
Once we were happy with our data pipelines, we hit the initial load. Then, one billion swipes later and with the rest loaded, we were able to provide the Cosmic team with everything they need and wouldn’t otherwise have access to — such as an overview of which location is making the most swipes and matches, and how much revenue it creates.
Our Business Intelligence dashboards show relevant KPIs for all stakeholders (with various levels of granularity) and are updated daily. Thanks to the solution, we’re also able to react to new requirements quickly and can build new reports in a matter of hours.
Challenge #2 - Improved Matching Algorithm
Our second goal was to create a better user experience by enabling users to match with the right people as quickly as possible. We wanted to put more emphasis on profile quality and user activity — by rewarding active people with high-quality profiles through greater exposure and more matches.
Ultimately, we want a healthy mix of high-quality and new profiles while quickly blocking bad actors.
Hand in hand with that is improving the matching algorithm for specific cities. Dating apps are location-based; every city is an ecosystem of people that have a certain quality of photos, a certain percentage of fake profiles, etc. An algorithm that suits Barcelona may not suit Los Angeles. We needed to find out what works where while compiling all data in one easy-to-understand place.
Solution
Using Data Warehouse mechanics, we can now easily separate users in one geographic location into multiple categories and watch how various algorithms influence the number of swipes and matches and the percentage of subscribed users.
By leaning on our activity data model, we are able to analyze groups in different locations, create hypotheses about their behavior and apply A/B testing to challenge it.
One of our first experiments was to check if providing a free subscription makes users more likely to swipe. As it turns out, it's not that much of a deal-breaker. We’ve learned from this result and have created more hypotheses that are going to be tested next.
In Conclusion: Being Data-Driven Is a Must
Our answer to understanding a business is (in many cases) Data Intelligence, and for good reason — which we hope we’ve made clear. This STRV solution is the result of countless products built, problems solved, gaps filled. And it’s not expensive. It really isn’t.
There’s a lot of data business owners have to track and understand nowadays. Of course it can feel overwhelming. But if you’ve been around the block, it’s not that hard anymore. Just find the right partner (or your own team) that’s able to build a customized, future-proof solution, with the potential to expand that solution easily, quickly and at a low cost… and you’re golden.
