How AI‑Powered “Deep Research” Supercharges Data Retrieval for Modeling

Doing mathematical modeling and data analysis faces a major challenge: finding reliable data.

Unclear data search direction

Unable to locate trustworthy data sources

Insufficient data quantity

Difficulty spotting meaningful patterns in inconsistent data

Large‑language‑model AI tools with a “web search” mode (e.g., DeepSeek) make data acquisition more convenient.

However, some retrieved data are unreliable or even fabricated by the AI, leading to untrustworthy conclusions.

AI advances have further improved data‑retrieval efficiency.

To advance further, three directions are suggested:

Enhance personal domain knowledge and data insight

Design better prompts for more targeted output

Choose more capable platforms

This article focuses on the third point and introduces a tool I recently tried that greatly boosts efficiency: Metaso AI “Deep Research”.

First, open the website: https://metaso.cn/

Enter a clear question and select the “Deep Research” function.

Example topic: “Research Shanghai food‑delivery market”.

The output differs from other LLMs:

It provides a “data retrieval process” with a “visualization + data summary + reference sources” format for each node:

Using growth‑rate data as an example, it reports that the 2025 market size is about 12 billion RMB, but notes the conclusion is incomplete because the data are national, not Shanghai‑specific.

Metaso then gives a staged data summary, highlighting three shortcomings:

1) Missing Shanghai 2025 overall market size data (only sector or national data available); 2) Rider income data are from 2023‑2024 surveys, lacking 2025 empirical data; 3) Platform market‑share data are outdated (latest 2023), not reflecting the 2025 competitive landscape.

It then asks for official data from the Shanghai Statistics Bureau on the 2025 food‑delivery market size and growth rate:

Shanghai Statistics Bureau 2025 food‑delivery industry overall market size and growth‑rate official data.

The process continued, taking 4.8 minutes, and produced:

A complete data‑collection flowchart

A report

180 reference materials (Word, PDF, web pages, etc.)

It can also generate a more intuitive “interactive report”:

Share link: https://metaso.cn/s/Ux2wQP0

In summary, Metaso AI “Deep Research” offers these key features:

Systematic data‑retrieval workflow

Stage‑wise data summarization

Automatic supplementation and real‑time updates

Interactive reports with traceable data sources

Easy sharing and team collaboration

Really great! I sincerely thank the AI developers and product managers (not limited to Metaso) for creating tools that dramatically improve research efficiency.

I hope this article helps you; feel free to try the tool and share your experiences in the comments.