Introduction
Tech leaders often face the same question: should we build our own tools or integrate external solutions (a.k.a. “build vs. buy”)? The answer isn’t always binary — sometimes the most effective approach is a blend of both.
This article tells the story of one of our clients: a major social media platform. Their engineering team had built and maintained an in-house snapshot testing system for Android, but as the app grew, complexity grew with it, and weaknesses started to appear.
Here’s how they worked with the Happo team to enhance their setup without completely replacing it, and what other tech leaders can learn from their experience.
1. The context: a mature but heavy in-house system
By the time Happo entered the picture, the client’s engineering team had already developed a robust internal snapshot testing system.
They used Roborazzi to generate screenshots.
Their process was battle-tested and deeply embedded in their workflows.
The system worked, but it wasn’t scaling efficiently.
As the number of developers and tests grew, so did the pain points: slow comparisons, slow build times from bloated repositories, and maintenance overhead that took focus away from shipping the things that really mattered for their company.
2. The challenge: scaling without replacing everything
The client wasn’t looking to throw away years of investment. Their question was more nuanced:
How do we keep the parts that work from our internal system?
How do we add efficiency without starting from scratch?
How do we do this at the scale of hundreds of developers working in parallel?
This is the classic build vs. buy dilemma — except in this case, the solution was to combine the two.
3. The solution: a Gradle plugin that bridges in-house infrastructure and Happo
The Happo team collaborated with the client’s engineers to create a Gradle plugin designed for integration, not replacement.
Here’s how it works:
Input: Screenshots are generated internally by Paparazzi and Roborazzi.
Processing: The plugin inspects and uploads these screenshot images directly to Happo’s servers.
Output: Happo handles visual reports, diffs, and comparisons against previous versions.
The plugin acts as a bridge — allowing the client to keep their custom generation pipeline, while offloading the heavy lifting of storage, comparison, and reporting to Happo.
4. The added value: optimization, not disruption
This hybrid setup gave the client the best of both worlds:
Speed: Happo’s diffing engine processes only what actually changed, cutting down comparison times.
Efficiency: Repositories stay clean and lightweight, free from thousands of PNG files.
Scalability: Hundreds of developers can run tests in parallel without slowing each other down. Merging code becomes easier because there are no conflicts between PNG snapshots.
Less flake: Happo’s powerful diff thresholds avoid flagging unimportant diffs for developers to review, helping them to stay focused on shipping. More signals, less noise.
Happo didn’t replace their system — it optimized it, making the whole pipeline faster and easier to maintain.
5. Lessons for tech leaders: build vs. buy isn’t binary
For CTOs, tech leads, and engineering managers, the takeaway is clear: the choice between “build” and “buy” doesn’t have to be absolute.
If you’ve already built: explore how external solutions like Happo can enhance, rather than replace, your system.
If you’re starting fresh: weigh the opportunity cost of building against the speed and scalability of ready-to-use platforms.
In this case, integration was the winning formula: preserving the client’s investments while solving their scaling challenges.
Conclusion
As engineering teams grow, so does the complexity of their tools. The story of this major social media platform shows that “build vs. buy” isn’t always about choosing sides. Sometimes, the smartest move is to build where it matters, and buy where it scales.
For this client, Happo became the backbone of snapshot management.Comparisons are made faster, there is less maintenance, allowing their in-house system to thrive at scale.
