How DANA used AI to boost productivity and quality through Mixpanel MCP

DANA (PT Dana Digital Group) is Indonesia’s leading digital wallet, connecting more than 200 million users with secure, fast, and easy payment platforms and financial services since its launch in December 2018. Built by the country’s top fintech engineers, DANA serves consumers, merchants, and financial institutions with a shared mission: to accelerate financial literacy and inclusion through technology.

In 2019, DANA’s team was struggling to track business and product metrics with the objective to better understand user behavior. They turned to Mixpanel to provide the analytics that the team needed for critical insights and informed decision-making.

“Mixpanel provided the sweet spot for engineering, product, and business teams to visualize and understand what happens in the field in a timely manner,” explained Randi Waranugraha, VP of Engineering, DANA. “The hard numbers gave us insights into whether or not an initiative worked, whether or not our customers were getting the best experience.” 

With these valuable insights, the DANA organization began to rely on Mixpanel data in executive meetings and consistent reporting. This growing demand naturally increased the time spent by certain team members on pulling data into presentable reports. Then came the opportunity to join the beta group for Mixpanel Model Context Protocol (MCP), and DANA happily accepted. MCP was created to fulfill the need to connect large language models (LLM) – like Claude or Cursor – to Mixpanel and get the answers needed.

With the growing popularity of using Mixpanel throughout the organization, the DANA team identified an opportunity to streamline two areas with MCP server integration:

1. Freeing up domain expertise 

DANA wanted to boost data democratization even further to support additional employees who typically might not utilize these data insights. To broaden access, they looked to move beyond the domain experts who typically handled segmentation, funnels, and user timelines, freeing them up for higher-priority work while giving more employees the data they needed on their own.

2. Automated reporting

The engineering team relied heavily on Mixpanel for critical workflows including tracking company KPIs, improving metrics such as performance and error rates, monitoring feature observability, and troubleshooting production incidents. Turning that data into executive-ready summaries and operational insights required too much time and effort. 

With the MCP server integration, the DANA team has transformed their data access, workflow, and speed of execution, as well as uncovered some additional unexpected benefits. 

Democratizing Data Access

The MCP server expanded data access by shifting queries to Mixpanel AI Agents. “Before MCP, if a PM or support team wanted to know which feature is generating the most errors this week, they’d ping the domain owner who would reply when they were available,” explained Ricardo Pramana Suranta, Head of Engineering, DANA. “Now they just ask the agent directly in chat. Anyone on the team can query Mixpanel data in natural language, directly from chat — no waiting, no dependency on the domain owner.”

Two of DANA’s AI agents (an analytics assistant and an error triage bot) run entirely on top of Mixpanel’s MCP server, making data access self-serve for the whole team. The Mixpanel AI Agent is the go-to for anyone who wants to explore data in natural language by combining Mixpanel’s MCP server and Mixpanel’s own API to investigate user-reported issues from multiple angles, giving a much richer picture than either could alone. Then there’s the Error Master Agent, which focuses specifically on error analysis. It queries error trends through MCP and cross-references DANA’s Notion knowledge base to automatically surface error definitions, root causes, and ownership, all in a single message.

“All of this freed up our domain experts from being ‘Mixpanel query humans’ and let them focus on actually fixing the problems, not just finding them,” shared Pramana Suranta.

Comprehensive Data From Multiple Sources

One of the biggest unlocks for DANA is being able to combine multiple data sources into their AI agents. With MCP, they’re able to tie multiple sources of data together to produce insights, without having to understand where the data lives or how to handle it. 

“We can synthesize massive information from Mixpanel, crash analytics, Notion knowledge base, and more, to give us the whole picture,” said Waranugraha. “That is what is really unlocking us.” 

Streamlined Reporting

With the MCP server, the team has automated much of their workflow through AI. Instead of manually compiling reports, they can query Mixpanel data directly and generate summaries and analyses that are automatically stored in a centralized knowledge base. This also allows them to combine Mixpanel insights with other internal knowledge sources, creating more comprehensive analyses and faster decision-making. 

“Mixpanel’s MCP server transformed Mixpanel from a dashboard we periodically check into a system we can query and reason with,” said Waranugraha. 

Uncovering Hidden Insights

MCP has enabled the DANA team to quickly investigate spikes in negative ratings and reviews in the App Store, especially during potential incidents. In the past, when ratings suddenly dropped, it triggered a manual investigation. A team member familiar with the specific feature would explore multiple Mixpanel dashboards, correlate user behavior with the feedback, and estimate the impact—such as how many users were affected and which segments experienced the issue. This process could take several hours.

“With MCP integrated into our workflow, we can now ask our agents to immediately analyze the trend and generate a report explaining the scope of the issue,” said Waranugraha. “Within minutes, we can understand the breadth and depth of the impact—how many users are affected, what kind of users they are, and which part of the product journey is involved.”

The team has also extended this into their technical support workflow. When troubleshooting a user case, the agents can submit a user ID and instantly reconstruct the user’s journey before and after the issue. By combining Mixpanel insights with their internal documentation and codebase knowledge, the system can even suggest potential root causes and possible fixes.

When an incident related to merchants’ EDC machines occurred, the DANA team was able to manage it efficiently without having to visit each merchant across Indonesia. “We used Mixpanel to analyse user events and error codes, identified the affected merchants over the last 30 days without leaving a desk, shared the list with the merchant team, and they prioritized visits by severity,” said Pramana Suranta. “It took us five minutes to detect the issue in Mixpanel and five minutes for the merchants to fix it on-site.”

Faster Decision-Making Leading to Better Outcomes

Leveraging MCP is enabling the DANA team to not just identify issues faster but also make quicker decisions and prioritize evidence-based decisions. This has resulted in reducing incidents from having a significant impact. 

For example, during the Hari Raya campaign, the team encountered an incident. “Normally we’d need time to gather reports, identify domain owners, find the impacted metrics, and determine the breadth of impact,” explained Waranugraha. “Our AI agents immediately assessed the exact events we needed to look at, showed us what caused the incident, and told us how many users were impacted.” 

The result? “This helped us decide in real time whether to release a fix immediately or wait. Pre-MCP, we would’ve waited until the campaign was over and then released the changes later with consequences to that delayed timing. With MCP, we saved one week of the impact becoming more widespread,” said Waranugraha.

DANA’s engineering team has unlocked several benefits in a short amount of time using MCP.

Time Savings of Manual Tasks:

• Reduced data query time by 99%: From four hours per day of manual work per 20 FTEs (80 hours per day) to less than one minute to access real-time data analysis 
• “Using MCP, we have reduced the time spent on manual work so the team can focus on planning for improvements and better decision-making,” said Pramana Suranta. “It used to take half a day for the 20 people involved in moving data for a C-level report. Now it’s down to less than a minute, and we have been able to recoup that time for more value-add work.” 

Troubleshooting Speed:

• Reduced cycle time for troubleshooting: From hours to minutes
• “Instead of having a specialized engineer spend hours digging through dashboards to understand an incident, anyone on the team can now ask a question and get the full story from user impact to potential root cause in minutes versus hours,” cited Waranugraha.

Company-wide Adoption of Mixpanel:

• Increased employee adoption of Mixpanel by 500% 
• “Before MCP, approximately 200 employees actively used Mixpanel,” said Pramana Suranta. “By connecting the MCP agent, we’ve unlocked data for 100% of the company with all 900+ employees at DANA now having access.”

“We see MCP as an important step toward the next generation of analytics where data is not just visualized in dashboards but actively integrated into how teams think, investigate, and make decisions,” said Waranugraha. “MCP opens the door for analytics to move from passive reporting to an active intelligence layer that helps teams understand their systems, users, and product decisions much faster.”

The team sees the biggest shift with analytics becoming part of everyday workflows. Engineers can quickly validate the impact of a release, support teams can understand user issues in context, and leadership can get faster answers to strategic questions without waiting for manual reporting. Going forward, DANA plans to focus on building ‘virtual engineers’ to scale workflows beyond just human engineers, as well as implementing AI across the entire software development lifecycle. 

“As we look to the next phase, we envision AI embedded in processes from requirements gathering through testing and deployment,” said Pramana Suranta. “We recognize AI’s potential to streamline manual work and further enhance our focus on other areas, allowing our teams to accomplish more, especially in maintaining quality. We see AI helping with forecasting peak seasons, particularly during high traffic events like Christmas and Hari Raya, to better allocate resources like server instances without wasting budget on over-provisioning.” 

“We’re excited to continue exploring this direction with Mixpanel,” concluded Waranugraha.