Singapore Property Market Expert

Domain knowledge for interpreting Singapore property market data correctly.

Trigger: Market terminology, bedroom classification, district/region decisions, URA data quirks, deal ratings

Market Segmentation

Core Central Region (CCR) - Prime Districts

code

D01: Boat Quay / Raffles Place / Marina
D02: Shenton Way / Tanjong Pagar
D06: City Hall / Fort Canning
D07: Bugis / Rochor
D09: Orchard / Somerset / River Valley
D10: Tanglin / Bukit Timah / Holland
D11: Newton / Novena

Characteristics: Highest PSF, luxury market, more freehold/999-year, lower volume

Rest of Central Region (RCR) - City Fringe

code

D03: Queenstown / Alexandra / Tiong Bahru
D04: Harbourfront / Keppel / Telok Blangah
D05: Buona Vista / Dover / Pasir Panjang
D08: Little India / Farrer Park
D12: Balestier / Toa Payoh
D13: Potong Pasir / MacPherson
D14: Geylang / Paya Lebar / Eunos
D15: East Coast / Marine Parade / Katong
D20: Bishan / Ang Mo Kio

Characteristics: Mid-range, good connectivity, mix of tenure types

Outside Central Region (OCR) - Suburban

code

D16-D19: East (Bedok, Changi, Tampines, Pasir Ris, Serangoon, Hougang, Punggol)
D21-D28: West/North (Bukit Timah, Jurong, Bukit Batok, Woodlands, Yishun, Seletar)

Characteristics: Most affordable, highest volume, predominantly 99-year

Bedroom Classification (Three-Tier System)

Why three tiers? AC ledge harmonization in June 2023 changed how floor area is measured. Post-harmonization units appear smaller for the same bedroom count.

Tier 1: New Sale >= June 2023 (Post-Harmonization)

code

1-Bedroom: < 580 sqft
2-Bedroom: 580 - 780 sqft
3-Bedroom: 780 - 1150 sqft
4-Bedroom: 1150 - 1450 sqft
5-Bedroom+: >= 1450 sqft

Ultra-compact sizing after AC ledge rules changed.

Tier 2: New Sale < June 2023 (Pre-Harmonization)

code

1-Bedroom: < 600 sqft
2-Bedroom: 600 - 850 sqft
3-Bedroom: 850 - 1200 sqft
4-Bedroom: 1200 - 1500 sqft
5-Bedroom+: >= 1500 sqft

Modern compact sizing with AC ledges counted.

Tier 3: Resale (Any Date)

code

1-Bedroom: < 600 sqft
2-Bedroom: 600 - 950 sqft
3-Bedroom: 950 - 1350 sqft
4-Bedroom: 1350 - 1650 sqft
5-Bedroom+: >= 1650 sqft

Legacy larger sizes from older developments.

Code: backend/services/classifier.py - classify_bedroom_three_tier()

URA Data Quirks

Month-Level Transactions

All URA transaction dates are 1st of month. There is no day-level granularity.

python

# WRONG: "Last 90 days"
date_from = today - timedelta(days=90)

# CORRECT: "Last 3 calendar months"
date_from = date(max_date.year, max_date.month - 2, 1)

Exclusive Date Bounds

Date ranges use exclusive upper bound to avoid double-counting at month boundaries.

sql

WHERE transaction_date >= :date_from
  AND transaction_date < :date_to_exclusive  -- NOT <=

Sale Types

Sale Type	Meaning
New Sale	Initial purchase from developer
Resale	Secondary market (owner-to-owner)
Sub Sale	Sale before TOP (Temporary Occupation Permit)

Tenure Types

Tenure	Meaning
Freehold	Perpetual ownership
999-year	Effectively freehold
99-year	Leasehold (most common)

Timeframe Resolution

Canonical timeframe IDs (backend resolves to dates):

ID	Period	Label
M3	3 months	3M
M6	6 months	6M
Y1	12 months	1Y (default)
Y3	36 months	3Y
Y5	60 months	5Y

Rule: Frontend sends ID, backend resolves to date_from and date_to_exclusive.

Deal Rating Framework

Percentile-Based Rating

Compares unit PSF against distribution:

Rating	Percentile	Meaning
Good	< 25th	Below market average
Fair	25th - 75th	At market
High	> 75th	Above market average

Comparison Levels

•Project-level: vs other units in same project
•District-level: vs all transactions in district
•City-wide: vs all Singapore transactions

Exit Risk & Liquidity

Transaction Velocity

Measures market health based on transactions per 100 units:

Velocity	Level	Meaning
< 5 txn/100 units	Low	Illiquid, harder to sell
5 - 15 txn/100 units	Healthy	Normal market
> 15 txn/100 units	Elevated	High turnover (investigate why)

Price Metrics

PSF (Price per Square Foot)

Primary comparison metric. Normalizes for unit size.

code

PSF = Transaction Price / Floor Area (sqft)

Median vs Average

•Median: Resistant to outliers, preferred for market summaries
•Average: Useful for total volume calculations

Outlier Handling

Transactions are flagged as outliers using IQR method at data ingestion. Every query MUST exclude outliers:

sql

WHERE COALESCE(is_outlier, false) = false

Common Data Questions

"Why do bedrooms seem wrong?"

Check sale type and date. A 750 sqft "2-bedroom" is correct for:

•Tier 1 (post-Jun 2023 new sale): 580-780 sqft = 2BR
•Tier 3 (resale): 600-950 sqft = 2BR

"Why are there gaps in monthly data?"

URA updates semi-annually (H1/H2). Recent months may have incomplete data until next release.

"Why does filter show no transactions?"

Check:

•Date range - URA data starts from specific year
•Combination rarity - some district+bedroom combos have few transactions
•Outlier exclusion - transactions may be filtered as outliers

"Why is median different from what I calculated?"

Ensure:

•Same date range (exclusive upper bound)
•Same outlier exclusion
•Same bedroom classification tier
•Same sale type filter