How Miinto embedded AI into data workflows using Secoda

‍Miinto, one of Europe’s fastest growing fashion marketplaces, connects customers with boutique products from over 3,000 different brands, ranging from luxury names like Balenciaga, Gucci, and Prada to hand-picked boutique labels. The platform relies on data from marketing, product, finance, and operations to manage everything from Google Ads spend to stock levels, pricing, and returns.

When Martin Brummerstedt joined to lead the Business Intelligence team, his mandate was to build a reliable foundation that could support consistent, trusted data across the organization. That involved making strategic decisions around tooling and data modeling to enable clear, aligned metrics across teams.

As Miinto’s ecosystem scaled, it became increasingly difficult to keep documentation accurate and definitions aligned. By bringing in Secoda early in the rebuild, the team established a centralized, searchable documentation layer and used automation and AI to keep it continuously updated.

The results

• Cut over a month of manual documentation work (20 min per table) with an AI-powered flow using Secoda’s Automations, APIs, and AI personas — eliminating the need for the BI team to ever manually document tables again, while significantly improving documentation quality.
• Reached 100% documentation coverage of business measures in Power BI and BigQuery, enabling consistent self-serve reporting across teams and reducing dependency on the BI team.
• Accelerated root cause analysis and data discovery by combining column-level lineage, natural language AI search, and targeted BigQuery models, helping users resolve questions in seconds, without BI intervention. 
• Improved customer support speed and consistency by integrating Secoda AI into Slack, reducing escalations and enabling agents to access accurate, policy-aligned answers on demand.

The goal

Miinto’s team set out to do more than just modernize their BI tooling. Their aim was to build a scalable, self-serve data environment that improved decision-making across the organization. To do that, they focused on four key outcomes:

1. Metric standardization: Align every team around consistent business definitions and remove the risk of conflicting reports

2. Faster root cause analysis and time to insight: Make it easier and faster to trace metric logic and understand changes without needing deep technical expertise

3. Data democratization and internal enablement: Reduce day-to-day blockers by making documentation easier to find, search, and trust

4. Scalable, sustainable documentation: Free up data team resources and keep documentation up to date without relying on manual processes

Miinto's Secoda implementation was guided by these priorities, encouraging company-wide data trust.

The solution

After selecting tools like BigQuery, dbt, and Power BI, the team brought in Secoda to serve as their self-serve intelligence layer. It gave teams a single place to explore available data, understand how it’s used, and trust that it’s accurate.

Step 1: Standardize business measures with a unified metric layer

One of the team’s first priorities was to align on shared definitions. “Each different department had their own definition of what the measure was and their own report,” Martin shared. “All of them told a slightly different story of what the value of the measure was. This is an issue because if you cannot agree about what this steering table shows, then it becomes very hard to make any decisions.”

Using Secoda’s Glossary, the team created a template for defining each measure. It included:

• A business description to explain the purpose
• SQL or logic in a code description
• Source details and lineage notes in a data description

This structure, along with Secoda’s automatic Related Resource metadata, helped uncover how measures connected to broader outcomes like sales or revenue. Every measure is now documented in Secoda and linked directly to relevant dashboards and datasets in Power BI and BigQuery.

“Now we’re actually at a point where we can look up the definition of different things and figure out much faster where specific elements are defined in our systems.”

This made metric logic more accessible to both developers and business users, strengthened trust in reporting, and helped eliminate guesswork by aligning the organization around a single source of truth.

Step 2: Accelerate root cause analysis with lineage and AI

As Miinto’s stack grew, questions like “Where is this column defined?” became increasingly common. To address this, the team turned to column-level lineage and Secoda AI.

Instead of tracing upstream logic manually, the team can now ask AI directly about a column’s origin. This speeds up debugging and highlights dependencies across models without tool-switching.

The team also started building AI personas trained on specific domains within their data stack. Each persona has access only to the resources relevant to a given topic, which helps keep responses accurate and focused. One example is their pricing benchmark process, which involves several intermediate steps in dbt and BigQuery. It was difficult to document in a way that made sense to everyone.

By creating a dedicated persona for that workflow, the team made it easier for users to ask questions and get reliable answers. It also reduced the need for repeated explanations and helped preserve institutional knowledge in a reusable format.

They extended this approach to natural language analysis by building a contribution model in BigQuery. It allows users to ask open-ended questions like “What changed in our product categories?” and receive structured summaries. That experiment led to the development of similar models across other parts of the organization.

This combination of lineage, Secoda AI, and targeted BigQuery models is helping Miinto unlock new use cases without needing to spin up new dashboards or run manual queries. Questions that used to require hands-on help can now be answered in seconds using context pulled directly from the data stack.

Step 3: Support customer-facing teams with Secoda AI in Slack

After seeing the impact of AI personas within the BI team, Miinto extended the same approach to customer support. The team created a Slack channel where support agents could ask operational questions and get responses powered by Secoda AI. The AI was trained on internal documentation stored in Secoda, including policies and standard procedures for handling customer issues.

This setup gave support team members quick access to reliable answers without needing to check multiple systems or escalate questions. In one example Martin shared, a customer received a damaged product and reached out for help. The Miinto support representative used the Slack integration to ask what the process should be—return, refund, or replacement—and the AI responded with the appropriate documented steps.

When team leads made updates to these responses, they were added back into Secoda, forming a feedback loop that improved both the AI and the documentation over time. The system became more accurate and more aligned with how the business actually operates. This led to faster support responses, fewer escalations, and more consistent customer experiences.

Step 4: Automate documentation maintenance with APIs, Automations, and AI

To reduce the manual effort involved in keeping documentation current, the BI team set out to build a system that could scale without needing constant oversight. Their approach combined several layers of automation, beginning with Secoda’s built-in Automations.

These no-code rules allowed the team to handle routine tasks like assigning owners, tagging assets, and verifying documentation. Martin described them as especially helpful for “maintaining tasks that are boring to do, but setting it up as a one-time thing and forgetting about it.” The Automations helped establish consistent standards without requiring regular manual work.

For more complex tasks, the team developed custom Python scripts powered by Secoda’s API. These scripts run on a schedule and begin by identifying a target table. They gather metadata, trace upstream and downstream lineage, review example rows, and analyze SQL logic. This context is passed to an internal AI persona trained to generate documentation in a specific format.

Once the AI generates a draft, another script formats the content into JSON and pushes it directly into Secoda through the API. The result is an automated workflow that documents new or updated assets in line with Miinto’s internal standards.

The impact was transformative: the team eliminated over a month of documentation work (based on 20 minutes per table) while significantly improving documentation quality. The automated approach delivered consistent formatting, comprehensive coverage, and eliminated human error.

As Martin put it, "We will essentially never need to document a table or column ever again."

By combining Secoda’s Automations with their own API-driven processes, the team created a system that keeps documentation accurate and up to date without relying on manual edits or check-ins. This freed up time for the team to focus on more strategic projects and helped ensure documentation stays reliable as the data environment grows.

What’s next

Miinto’s BI team is continuing to explore how AI can support analysis across the organization. One current initiative focuses on helping users investigate profitability drivers across product categories. Dashboards often show what changed, but not why. By combining BigQuery data with existing metrics and using Secoda AI to ask questions, the team is working to surface explanations tied to performance factors like return rates, shipping costs, and partner agreements.

On deck is also Secoda’s AI charting features to make insights easier to explore. These tools can help users visualize trends and receive plain-language summaries,  especially in cases where dashboards don’t exist or the analysis pulls data from multiple systems.

Another focus is expanding their set of domain-specific AI personas. Each persona will be trained on a specific area of the data stack and designed to support business users who need quick answers without navigating technical tools. This builds on their documentation efforts and makes it easier for teams to access the information they need to make informed decisions.

Martin summed it up: “We believe that making data available for self-service, and making information about what you're looking at available, will lead to overall better results.”

By continuing to integrate AI into both search and analysis, Miinto is improving how data is used across every team.

If you're interested in exploring how Secoda can support your data initiatives, book a demo today to learn more.