marmos

What is marmos

marmos is launched by Tencent Beacon team,AI-native data analytics platform. Through natural language conversation, users can complete data integration, analysis, chart generation, and interactive dashboard building in one sentence, supporting multi-source data such as Excel, MySQL, ClickHouse, and cross-source correlation. marmos integrates multiple models including Hunyuan, DeepSeek, Zhipu, Moonshot, adopts a zero-leak security architecture, and has served 100+ business teams within Tencent.

marmos's main features

• Natural language analysis: Enter one sentence to automatically generate data analysis reports and visualization charts.
• Multi-source data integration: Supports multiple data sources such as Excel, CSV, Tencent Docs, MySQL, ClickHouse, etc.
• Online information retrieval: Can automatically retrieve public information on the internet and generate structured decision dashboards.
• Interactive dashboards: Generated data dashboards support modification, addition, and deletion of chart cards at any time, with the entire process editable.
• Intelligent attribution analysis: Built-in attribution engine automatically breaks down the core driving factors behind data fluctuations.
• Proactive monitoring and alerts: Supports setting data monitoring rules, automatically pushing alerts for anomalies, enabling "proactive monitoring".
• Cross-source correlation analysis: Enables correlation analysis across multiple data sources through a unified query layer, breaking down data silos.

How to use marmos

• Register an account: Visit the marmos official website https://marmos.qq.com/ to complete account registration and login.
• Connect data: Upload local Excel/CSV through "Connect New Source", or authorize connection to data sources such as Tencent Docs, MySQL, ClickHouse, etc.
• Ask in natural language: Directly enter analysis needs in the dialog box, such as "generate a dashboard for this week's sales data".
• AI generates dashboard: The system automatically analyzes the question, retrieves data, generates charts, and builds an interactive dashboard.
• Edit and adjust dashboard: Click on a chart card to modify, insert, or delete charts, and adjust analysis dimensions and styles as needed.
• Set monitoring alerts: Configure data monitoring rules, and automatically push reminders for abnormal indicators, enabling proactive monitoring.
• Share with the team: Share the generated dashboard link or permissions with specified colleagues, supporting collaborative viewing.

marmos's core advantages

• Zero-threshold conversational analysis: No need to write SQL or understand BI; a single natural language sentence can drive the entire data analysis process.
• Full-chain closed-loop coverage: From data integration, dashboard generation, editing and adjustment to monitoring alerts and team sharing, all completed in one stop.
• Multi-model intelligent fusion: Integrates multiple models including Hunyuan, DeepSeek, Zhipu, Moonshot, providing more comprehensive analysis capabilities.
• Enterprise-grade zero-leak security: LLM never touches real data, supports private deployment and sensitive data probes.
• Cross-source correlation analysis: A unified query layer connects multi-source data such as Excel, databases, and online documents, breaking down information silos.
• 1 person matches 1 team: Verified by real cases, it can compress the work of an original 2-3 person data team to be completed by 1 person.
• Large-scale internal real-world validation: Originating from Tencent Beacon team, has served 100+ business lines including WeChat and Tencent Video.
• Fully editable dashboard end-to-end: AI-generated charts and dashboards support modification, addition, and deletion at any time, enabling continuous iteration without starting over.

Comparison of marmos with similar competitors

Application scenarios of marmos

• Automation of weekly content operations reports: Automatically integrates multi-terminal (OTT/mobile) data, automatically generates and pushes operational dashboards such as MAU, DAU, and IP playback rankings at 8 AM every Monday, turning content support decisions from "by feel" to "by rankings".
• Product Governance and Deduplication Analysis: Facing millions of SKUs, multi-layer deduplication is performed through "product content fingerprints," automatically locating duplicate sources and prioritizing them by contribution level, turning governance actions from "guessing" into "looking at data."
• Financial Operations Analysis (P&L Breakdown): When a finance BP asks ad hoc follow-up questions during a business analysis meeting, a waterfall chart is generated in one sentence, breaking revenue down into multiple factors, achieving "one question, one answer, with data chasing the questions."
• Marketing Campaign Attribution: During major promotions, campaign performance is broken down across three dimensions—channel/creative/audience—automatically flagging ineffective channels in red, directly reusing the winning combination, reducing ineffective metrics by 40%, and increasing overall ROI by 35%.
• Consumer Decision Dashboard Generation: Public information is retrieved online to generate structured decision dashboards, such as rankings of popular outbound travel destinations, visa policy matrices, price trend forecasts, etc., assisting personal or corporate decision-making.