How Moonpig built a product data culture that runs on curiosity, not dashboards

Most product teams have more data than they know what to do with. The problem isn't access to the numbers. It's building an org where people trust the data, know how to use it, and reach for it without being asked.

That's a harder problem than it sounds. And it's one Moonpig has spent real time solving.

At MXP London, Tom Mann, Director of Product at Moonpig, sat down with Mixpanel CRO Damian for a fireside chat about how the company operates. No slides, no frameworks, just a direct conversation about what it takes to run a product org where data is genuinely part of how decisions get made. Moonpig's Mixpanel usage is disproportionate to its size, and Tom was candid about why.

Here's what he said.

Ten years ago, Tom said, the PM role didn't look the way it does today. His title at the time was "sales, marketing and operations manager," essentially everything and nothing, because product as a discipline hadn't fully arrived yet.

Back then, the work was largely about managing systems: aligning teams, scoping builds, waiting weeks for data to come back, then deciding what to do with it. Now, AI agents are shipping code. The time from idea to live feature has compressed to hours. That shift changes what a product leader is there to do.

When shipping velocity stops being the bottleneck, understanding becomes the bottleneck. A product team that can build anything in a day still has to answer the right question first: what should we build, and how will we know if it worked? Data can't be something you go check after the fact. It has to be part of how the work happens.

Related read: The PM playbook is broken for AI. Here's how to fix it.

Damian noted that Mixpanel usage at Moonpig is disproportionate to the company's size. Tom's explanation wasn't a single program or initiative. It came down to three things.

Easy access. Tom put it plainly: "There's nothing worse than the data looking like a screen from the Matrix." When data looks impenetrable, or sits with one person or one team, people stop asking questions. Moonpig focused on removing those barriers so anyone curious enough to look could get to something useful.

Momentum. Insight without action is just information. Moonpig runs weekly experimentation clinics where teams share what they learned from tests, not just what won, but what the data showed regardless of outcome. The act of sharing creates a feedback loop that keeps curiosity alive.

Celebrating failure as much as wins. "If you just celebrate the wins, people become risk averse and then your experimentation becomes experimenting to win, not learn." When the only visible reward is a successful outcome, people stop taking the swings that lead to real discovery.

Related: Mixpanel teams share their tips on building a company data culture

The thread running through everything Tom described is access. When someone has to ask another person, file a ticket, or wait for a scheduled report just to answer a basic question, most people stop asking.

Moonpig's product data culture didn't happen despite everyone having access to data. It happened because of it. Self-serve isn't a nice-to-have. It's what makes the other things (the momentum, the experimentation clinics, the psychological safety to fail) possible in practice.

Moonpig was among the first customers to start using Mixpanel's Model Context Protocol (MCP) server, and Tom said the team was itching for it the moment it became available. He described two specific ways they use it.

Dashboards in hours, not weeks. Building dashboards used to require knowing which events to pull, how they were named, and how to configure reports around them. For someone new to the business or without a deep data background, that was a real barrier. With MCP, a team member can describe what they want to understand in plain language and get a working dashboard without the manual setup. What used to take days or weeks now takes hours.

Session replay at scale, in Slack. Moonpig primarily lives in Slack, and they run Mixpanel's MCP server through it so the team gets answers without switching tools. The session replay use case is the more notable one. 

There's a practical ceiling on how many recordings any person can watch. So Moonpig ran MCP across all their replays to surface behavioral trends first, then let the team focus their attention on the sessions that actually warranted a closer look. Tom's qualifier: "It's not perfect but it's really the direction we're going."

When asked what from the MXP keynote got him most excited, Tom didn't hesitate. It was the shift from reactive to proactive insights, the idea that instead of spending time hunting for a signal in your data, the signal comes to you.

His image was specific: arriving on Monday morning and finding a problem already surfaced, with a suggested next step ready. "I think that we don't want (I certainly don't want) the team spending all of their time trying to find data to find an insight for a problem to solve. We want them solving the hard problems and spending time on that."

That's the direction Mixpanel AI is headed. The vision Tom described at the keynote, insight that comes to you rather than waiting to be asked, is exactly what it's designed to deliver.

The keynote also introduced Mixpanel Headless, a Python SDK built for technical teams who want to go further. With Headless, AI agents can read and drive Mixpanel programmatically, querying funnels, managing cohorts, rolling back feature flags, without a human at the keyboard. It's built for teams that want an agent working on its own, not just answering questions when prompted.

Together, these represent a meaningful shift in how product intelligence gets delivered: less time hunting for answers, more time acting on them.

Tom's closing advice was simple: stay curious, and don't wait for certainty. "You cannot put your head down and go, I'll just think about that AI thing later. By the time your head's up, the whole world around you has changed."

He wasn't arguing for adopting every new tool. He was arguing for being in it, using the technology, learning where it works and where it doesn't, experimenting without waiting for a perfect answer. "Don't wait for certainty because the next model will be tomorrow or the next day and you won't know. It's impossible to keep up. So you just have to try it, give it a go and see what works."

The teams pulling ahead right now aren't necessarily the ones shipping the most. They're the ones learning the fastest, and building the habits and access to make that learning repeatable.

To see what Mixpanel AI can do for your team, learn more here or start building agentic workflows with Headless and Mixpanel as the data layer.