Market Scan

Strategy & Knowledge • Work sample

Market Scan |full

What it is

Enter a market and it produces what the first day of a commercial due diligence produces by hand — but in under a minute, and with its sources attached. Four layers stack up: a qualitative landscape (snapshot, a competitor table of positioning, pricing and differentiators), a voice-of-customer read (real G2 / Capterra user ratings per player), a quantitative layer of public comparables — live financials, valuation multiples (EV/Revenue) and forward analyst estimates (consensus revenue growth and price targets) — and a recent-developments news flow (funding, launches, M&A, partnerships). It’s the shape of a real CDD, automated end to end.

Sample markets are pre-run and cached; “scan your own” runs live and is rate-limited.

How I built it

Three data layers, each from the right source. Firecrawl searches and scrapes the comparison articles and vendor pages, and an LLM synthesises them into a grounded, cited landscape — no ungrounded claims. For any player that is publicly traded, Financial Modeling Prep supplies real market cap, revenue, growth, gross margin, the EV/Revenue multiple, consensus forward revenue growth, analyst price targets and its latest headlines; private vendors are correctly left out of the comps. Per-player user ratings are pulled from G2 / Capterra with a stealth scrape (their ratings are JS-rendered behind anti-bot, so a plain fetch fails — the stealth path gets clean structured data). A second, recency-focused Firecrawl pass feeds the LLM the industry news flow. Everything is tied back to a source URL.

Highlights

Qualitative and quantitative — a landscape table plus real public-company financials
Sources cited throughout; the model refuses to invent numbers or events
The pipeline mirrors how a CDD analyst actually works — search, read, quantify, track news

Stack

Python · Streamlit · Firecrawl · OpenRouter (LLM synthesis) · Financial Modeling Prep