How Flip built an AI-powered Insights Engine with AppsFlyer MCP

The challenge

Flip is one of Indonesia’s leading fintech apps, helping millions of people move money every day. What started as a domestic bank transfer product has grown into a platform of 10-plus financial services, including international remittances, shopping cashbacks, and more. But growth in product breadth brought a new problem for the performance marketing team: how do you keep up?

Three people. Ten-plus products. Strategy, execution, optimization, and reporting, all of it, end to end. The team ran periodic reports through semi-automated Google Sheets dashboards, but generating the actual insights and walking stakeholders through results still required manual effort. Every week, product marketers and business owners would come to the team with ad hoc performance questions. Between three and five requests would land each week. Each one took up to an hour to answer. That is up to five hours a week spent on reactive reporting instead of proactive strategy.

“The team’s bandwidth is limited. When we needed to handle ad hoc requests, we couldn’t do other important things like campaign optimizations or drafting strategy for the next campaign.”

Asriana Septari, Digital Marketing Manager, Flip

The depth of insight also varied depending on how much time the team had and who was writing it. Flip needed a way to give stakeholders reliable, on-demand answers without burning out a lean team.

The approach

Flip connected AppsFlyer’s MCP to Claude, the AI tool already used across their organization, and built a self-serve insights engine that works as an extension of the performance marketing team.

The setup was structured around Claude Projects: one per product, each loaded with the context the team had spent years building. Project instructions defined how to analyze performance data. Uploaded knowledge files captured benchmarks, best practices, and product-specific baselines, what a good CAC looks like for one product versus another, what CTR thresholds matter, how to read attribution data correctly. Through MCP, each project connects directly to AppsFlyer and pulls live attribution data, install data, activity events, and cost data on demand.

Business owners and product marketers across Flip now have direct access to their product’s Claude Project. They type a question. They get a report built to the performance team’s standard, in language designed for a non-technical audience, with action items included.

The solution

Before MCP, the team was fielding three to five ad hoc insight requests every week. Each one took up to an hour to research, write up, and deliver. That added up to as much as five hours a week spent on reactive reporting, for a team of three with no spare capacity.

The same request now takes under 10 minutes, handled directly by the stakeholder without involving the performance team at all. The result: up to four hours saved every week, and up to 16 hours reclaimed every month. For a three-person team, that is the equivalent of two full working days given back each month to focus on what actually drives growth.

“AppsFlyer’s MCP has helped us create an extension of our team, one that handles the routine requests and gives the rest of us time to focus on what actually moves the needle.”

Asriana Septari, Digital Marketing Manager, Flip

The consistency of insight quality has also improved. Every report generated through the Claude Projects draws from the same knowledge base, the same benchmarks, and the same analytical framework the performance team built. Insights no longer vary based on who wrote them or how much time they had.

Why it matters

Flip’s story is not about replacing the marketing team. It is about amplifying what a lean team can do. By connecting AppsFlyer’s data to an AI that understands their standards, their products, and their audience, they have turned a resource constraint into a competitive advantage.

“Performance marketers do not need to do the heavy lifting of crunching data anymore. They can focus on strategy and actionables, and make sure the AI’s answers are right.”

Asriana Septari, Digital Marketing Manager, Flip

That last point matters. Tari is clear that AI works because the team knows their data deeply enough to verify it. The knowledge base they built is not just instructions for Claude. It is a codification of everything the team has learned. MCP is the bridge that connects that expertise to live data, at the speed stakeholders need.