01
Ingest
Paste a broker message. It's stored verbatim, then extracted into structured price observations. Low-confidence extractions are flagged.
Housing price intelligence
Know what a home is actually worth.
Dwelling Fee collects the fragmented price signals you already have — broker chat, listings, screenshots — and turns them into structured, queryable market intelligence. Every fact links back to its raw text and carries a confidence score.
No single confident numbers — only honest distributions.
conf 92%
n = 14 observations
Vinhomes Grand Park
apartment · Quận 9 · living page
3.2M
price / m²median · IQR p25–p75
“Bán 2PN, 60m², 3.2 tỷ, sổ hồng. TL.” → click any point for the raw signal.
Built for the messiness of real price signals — wherever they live
Broker chatZaloMessengerSMSWeb listingsScreenshots
Why it's different
A median that silently mixes rent, asking and transacted prices is confidently wrong. Dwelling Fee refuses to do that.The product's defining trait: intellectual honesty over a single confident number.
How it works
From a pasted message to queryable intelligence
Four surfaces, one pipeline. Messy text goes in; structured, provenance-backed market facts come out.
02
Review
A human-in-the-loop queue. Link an ambiguous observation to a candidate property, create a new one, or dismiss it — your call.
03
Properties
Resolved entities become living pages that aggregate every observation over time — a price scatter and an honest IQR distribution per property.
04
Analytics
Price/m² distributions, segmented by listing type and deal status — never mixed. Segments with n < 5 are flagged as underpowered.
The real work
The hard problems aren't CRUD
Anyone can store a row. Dwelling Fee is built around the three things that actually make housing data trustworthy.
Extraction
Turn messy, abbreviated, multilingual chatter into correct structured facts. The shorthand is the hard part:
2PN = 2 bedstỷ = billion ₫sổ hồng = titleTL = negotiableEntity resolution
Decide whether two messages describe the same property — then merge their observations into one living page instead of scattering duplicates.
needs reviewauto-link 92%
Statistical honesty
Never mix rent, asking and transacted prices. Show distributions — median and IQR — with sample-size guards, never a lone confident figure.
median · IQR p25–p75n < 5 → underpoweredTrust, made visible
Every number traces back to where it came from
Because the product is about honesty, trust is encoded in the interface itself — not buried in a methodology page.
- Provenance on every factClick any data point to jump straight to the raw signal it was extracted from.
- Confidence, shown plainlyA small mono percentage and a sage→amber→terracotta dot. No hidden assumptions.
- Caveats said out loud“Only 3 sale observations — too few for a reliable estimate.” Thin data is labelled, never smoothed over.
“…3.2 tỷ, 60m², sổ hồng”
3.2M ₫/m² · sale · asking
92%3.2M ₫/m² · sale · asking
“cho thuê 12tr/tháng”
rent · transacted
61%rent · transacted
“giá tốt, LH”
no price extracted
38%no price extracted
Low-confidence and unpriced signals are excluded from analytics until resolved.
Questions
Honest answers
Where does the price data come from?
From the signals you already have — broker messages on Zalo, Messenger and SMS, web listings, and screenshots. Each is stored verbatim, then extracted into structured price observations that link back to the original text.
How do you handle asking versus transacted prices?
They're never mixed. Distributions are segmented by listing type and deal status, because a median that silently blends rent, asking and transacted prices is confidently wrong.
Can I trust a single number from the tool?
Dwelling Fee never shows a single confident number. It shows distributions — a median with an interquartile range — and flags segments with too few observations as underpowered, so you always see how much data is behind an estimate.
Does it understand Vietnamese broker shorthand?
Yes. Extraction reads real shorthand such as 2PN (two bedrooms), tỷ (billion VND), sổ hồng (land title) and TL (negotiable), turning abbreviated chatter into correct structured facts.
Start with a single message
Paste one broker message and watch it become a structured, provenance-backed price observation. No spreadsheets, no guesswork — just honest housing-price intelligence.