Maor Frid leads data and analytics for Acid Labs— and lets Claude Code, powered by Mixpanel Headless, do most of the building for him.

Acid Labs is a gaming startup of about 20 people, building games on the Telegram platform and, more recently, releasing titles on Android and iOS.

Maor Frid is Head of Data and Analytics at Acid Labs, and he’s also a team of one. He’s responsible for making sure the rest of Acid Labs has access to trusted reporting. The data and reports they consume help them assess the success of past releases and decide what to tweak in their strategy next. That makes him the guardian of what’s in Mixpanel: If the data isn’t there, or isn’t right, everyone suffers from it.

Games at Acid Labs move fast. A central element of Acid Labs’ games is the sticker album: a limited-time collectible event, similar to a physical sticker book, where players collect a set number of stickers to complete a themed album before it expires. Acid Labs launches a new one every 3 weeks, and each launch is its own live event, with its own start date, difficulty, and sticker count logged as a database record. Maor creates and maintains the reporting to help game designers and user acquisition managers assess how a new album or a new campaign perform.

Maor is a whiz at Mixpanel; in fact he initiated the platform’s rollout at a previous company. For each event launch, he’d build advanced dashboards comprising 15 to 20 reports, 5 custom event properties (the extra details tagged onto a player action, like which album they’re playing), and 3 new cohorts (groups of players who share something in common, like everyone who finished the album) — all within a few hours. But with launches accelerating, Maor set his sights on moving even faster, turning a process he already knew by heart into something more automatic.

After finding efficiencies by automating some of his tasks, Maor wanted to experiment and take it even further. Enter Mixpanel Headless. 

Mixpanel Headless lets you create and edit things in Mixpanel — reports, cohorts, calculations — through code instead of clicking through the product. Claude Code can read his data and write directly into Mixpanel on his behalf, and, unlike a one-off request to the built-in agent, carry the context from one request into the next. That combination is why Maor now spends less time inside Mixpanel than the colleagues consuming his dashboards do. He describes what he needs in plain language, and Claude builds it using Claude Skills he refined over time.

• Taught Claude his own playbook first:  before delegating any real work, Maor gave Claude his own naming conventions and formula style, so every calculation it wrote looked like something Maor would have written himself. A single album’s naming convention alone can cover nine separate custom properties — things like how many hours since the album went live, or how many hours are left before it closes.
• Then let it write directly into Mixpanel: at first, Maor started generating formulas and calculations through Claude. He would then have to copy it into Mixpanel. Soon enough, with Headless, Claude could create and update reports, cohorts, and properties in Mixpanel on its own, closing that last manual step.
• Automated the full dashboard build: for a recurring workflow like a new sticker-album launch, Maor now points Claude at the previous album’s dashboard and asks for the same setup. Claude fetches the relevant data, applies the naming conventions, and returns a finished, linked dashboard for the new album.

The same approach solved a second, unrelated headache: keeping campaign tracking current for the User Acquisition team, who understood exactly which campaigns mattered but not how to build or edit Mixpanel properties themselves.

Maor built a two-way sync between a shared Google Sheet and Mixpanel: Claude writes the current campaign list into the sheet daily, the User Acquisition team flags what they want tracked, and Claude reads those flags back and updates the Mixpanel calculations — no Mixpanel expertise required on their end.

1. Hours of manual dashboard work down to 10 minutes

Setting up a new dashboard for a live event used to mean manually going report-by-report through the Mixpanel UI, a job that took a couple of hours every time a new album launched. Maor now gets a finished, correctly named dashboard back in about 10 minutes, and still checks it himself before sharing it widely. Because Claude rebuilds everything from the same defined process every time, it also prevents manual errors, such as forgetting a filter or using an outdated cohort carried over from previous launches.

2. Self-serve data for all, without having to touch Mixpanel’s back end

Every calculation and cohort Claude builds follows Maor’s own naming convention. That means colleagues across Acid Labs can find the right data point without knowing how it was built — they just search for the album by name.

With Mixpanel Headless, Maor built scalable, predictable systems, using Claude to act directly on Mixpanel data—and without Maor standing in the middle of every request. Beyond the time savings, this gave everyone trustworthy, self-serve data — a strategic lever to improve the quality of the decisions taken.