Research
AI market research agent that delivers a sourced, decision-ready report
Market Analyst breaks your research question into 3–5 sub-questions, scans primary and recent sources on the web, combines the findings with your own Google Analytics and social data, and writes a report that ties every claim to a numbered source.
- As a workflow node
agent.marketAnalyst- Default model
- Claude Opus 5
- Economy model
- Claude Sonnet 5
- Tools
- 4
- Report
- Market research
What the agent does
AI market research runs at three depths: Quick scan produces a report of about 600 words from 4–6 searches and 2–3 pages, Standard about 1,200 words from 8–12 searches and 5–8 pages, and Deep dive about 2,500 words with comparison tables from 15–25 searches and 10–15 pages. Queries are written in the market's language and in English: industry reports, statistics offices, trade associations, marketplaces, review sites, competitor pricing pages and news.
The agent never invents numbers: every claim rests on a source it actually opened, dates are noted where data may be stale, and anything it can't verify is marked “not found”. With Google Analytics connected it reads the last 90 days of channel and page data, and with social accounts connected it sees what resonates, so recommendations tie back to your own demand.
The report always has the same sections: executive summary, market size and trends, segments and pains, channels, pricing benchmarks, opportunities, risks and sources. Numbers go into tables and citations use the [n] format. Segments and opportunities also come back as structured data, so a workflow can, for example, hand the top opportunity to Content Creator as a topic.
Inputs / Outputs
Inputs
- What do you want to find out?required
- State the research topic in one sentence: product or service, customer type, region and time frame. Example: “Vegan café market and price ranges in Istanbul, 2025–2026.” A generic question gets a generic report.
- Region
- Country, city or region to cover:
Turkey,DACH,US Midwest. Leave it empty to infer it from your project's market; fill it in to get local sources. - Research depth
- Quick scan, Standard (default) or Deep dive. It sets the number of searches and pages read, the report length and the cost; for Deep dive, raise max tool steps and the run budget.
Outputs
- report
- The full report in Markdown: fixed sections, tables for numbers and numbered source citations.
- segments
- Customer segments: name, estimated size with its source (or “unknown”), pains, how to talk to the segment and the channels to reach it.
- opportunities
- Opportunities: title, rationale, impact and effort (high, medium, low) and a suggested first step.
- sources
- Sources used: title, URL and a note on what each one supports, with a date where relevant.
Saves a report to the Reports page on every run: Market research.
Tools
- Web search
web_search - Read a web page
fetch_url - Analytics summary (GA4)
get_analytics_summary - Post performance
get_post_performance
Example tasks
- “Research the vegan café market in Istanbul: how demand is changing, typical price ranges and the 3 best districts for a new opening.”
- “Research demand for online dietitian services in Turkey. Find at least 4 customer groups, their main problems and the best channel to reach each.”
- “Look into the skincare habits of women aged 25–35: which products they buy and where, how much they spend a month and which brands they trust.”
- “We're considering expanding to Germany. Summarize price ranges and localization expectations in the DACH market for our category; use only sources from 2025 onwards.”
Tips for good results
- Give topic, customer type, region and time frame together: “D2C skincare brands in Turkey, 2024–2026” instead of “cosmetics market”.
- Say what the report is for: setting prices, opening a branch, launching a service. Opportunities are written for that decision.
- Set a source rule: “Only official statistics, company reports and industry associations; don't count blog posts as sources.”
- Fill in your project's offerings and competitors; the agent sees that context on every run and judges opportunities against it.
- It is built for occasional, in-depth work: before a launch, when assessing a new market, at the start of a quarter. For weekly tracking, Analytics Strategist fits better.
Model advice
Default model
Claude Opus 5
Per 1M tokens: $5 input · $25 output
Economy model
Claude Sonnet 5
Per 1M tokens: $2 input · $10 output
Synthesizing long context and connecting sources consistently is the heart of this agent's job, so its default model is flagship tier; keep it for deep dives and for decisions where mistakes are expensive, such as an investment or a launch. For quick scans and most standard-depth research, the balanced model listed as its economy option is enough. Economy-tier models are weak at source synthesis and are not recommended here.
Frequently asked questions
Can I trust the numbers in the report?
Every number is tied to a numbered source, and anything the agent can't verify is marked “not found”. Before an important decision, open the source and check its date.
Which accounts do I need?
Serper or Tavily is recommended for web search; without one the source list stays short. Google Analytics and social accounts are optional and ground the recommendations in your own data.
How often should I run it?
Once a quarter or before a major decision. For weekly performance tracking, Analytics Strategist is a better fit.
Does it research local, non-English markets?
Yes. Queries are written in the market's language as well as English, and the report is written in your project's language.
Get started with OrqLabs
Build your first workflow today
Start on the free plan: connect your accounts, pick a template and stay in control with approval steps. No credit card required.