Build distinct shortlists.
Start with Cheras condos for lower budgets, Mont Kiara condos for the higher-price tier, and Country Heights Damansara bungalows for a smaller premium segment.
ABC
Company
The price
of a place.
00Find the right market.
The brief
Find the right market.
Then question the price.
ABC lists properties in Malaysia. With a recession concern shaping its 2023 planning, the company wants to focus its resources and grow revenue by 10% over one year.
The approach: understand Kuala Lumpur's asking prices, find promising neighborhood segments, and test what property features can tell us about price.
These are listings, not completed sales. Supply can guide a shortlist; it cannot prove buyer demand or revenue growth.
- Cleaned listings
- 4,801
- Neighborhoods
- 65
- Median asking price
- RM1.30m
- Regression sample
- 177
01The average is not
A market with a long tail
The average is not
the typical listing.
The mean asking price is RM2.17 million. The median is just RM1.30 million. A small number of very expensive listings pull the average upward.
Half of the listings fall between RM715,000 and RM2.50 million. That middle range is a more useful starting point than a single market-wide average.
FIG. 01
How much of the market fits a budget?
50.3%of listings at or below this price
All 4,801 listings included. Mean and median marked below.
02Start with a neighborhood,
Two price tiers, different opportunities
Start with a neighborhood,
not a city-wide average.
The lower-price tier contains 1,202 listings at or below RM715,000. The other 3,599 listings sit above that boundary. This is a split within this dataset, not a universal definition of affordability or luxury.
Lower-price tier
Cheras
171 listings / RM450,800 median
Condominiums make up 80 of these listings. A useful starting point for a budget-led shortlist.
Higher-price tier
Mont Kiara
651 listings / RM1.90m median
521 are condominiums. A concentrated pool of listings for buyers with a larger budget.
FIG. 02
The market, neighborhood by neighborhood
Top 10 / Listing count
A smaller premium niche. Country Heights Damansara has 27 higher-price listings, with an RM8.18m median. Eighteen are bungalows.
Watch the sample size. The RM180,000 median for Bandar Tasik Selatan comes from just one listing. Bandar Damai Perdana's RM725,000 higher-tier median comes from two.
03Bigger properties cost more.
A closer look / Desa ParkCity
Bigger properties cost more.
How much can that explain?
A separate sample of 177 Desa ParkCity listings puts price alongside rooms, bathrooms, parking spaces and size. Size has the strongest individual correlation with price: 0.80.
Together, the four features explain 74.0% of price variation in this sample. That is a measure of fit, not 74% prediction accuracy.
FIG. 03
The relationship, and what it leaves unexplained
177 listings. Each dot is one observation.
THE CATCH
Rooms and bathrooms overlap strongly (correlation 0.82). Once the other features are accounted for, the rooms coefficient is uncertain. Dropping rooms barely changes the model's fit.
Desa ParkCity / Scenario study
Put a price
on the features.
This is a fitted asking-price estimate, not a valuation. Changing a feature shows an association in the model, not the value caused by a renovation.
Fitted price / RM
Fitted asking price / RM
Adjusted R-squared73.4%
Residual standard errorRM435,863
Individual inputs stay within observed ranges; their combination may not occur in the sample. Model assumptions need further review. No held-out validation is available.
What does each feature add to the model?
Coefficient estimates and conventional 95% OLS confidence intervals. Size is shown per 100 sq ft; other features per one unit. Intervals depend on model assumptions and are not robust to unequal error variance.
04Use the market to focus.
The decision
Use the market to focus.
Use the model with care.
Test the furnishing offer.
Compare fully furnished options in Cheras and Mont Kiara. Measure qualified inquiries and completed deals before shifting more resources. Listing counts alone cannot establish demand.
Validate before pricing.
Review unusual records and regression assumptions with the analytics lead. Test on unseen listings and actual transactions before using model estimates in pricing decisions.
The path to 10% revenue growth is a testable commercial plan, not a result this dataset has already demonstrated.
Sources & study notes
The details behind the story.
Data preparation and definitions
The source starts with 5,000 listing records. Cleaning removes duplicates and unusable price or size records, standardizes room counts, simplifies property-type labels, and converts area to square feet. The cleaned table contains 4,801 listings, with missing room, bathroom, parking and furnishing values retained where present. Summary statistics exclude missing values for the relevant feature.
The tier boundary is the first price quartile, RM715,000, as implemented by the workbook formula. The median is RM1.3m. All neighborhood prices called "median" here are recalculated medians, not means. The 370 price outliers use Q3 + 1.5 × IQR and are retained, not silently removed.
Built-up area and land area are not interchangeable. Their measurements are preserved; comparisons by size should consider that distinction. The minimum asking price (RM1,150) and extreme sizes are data-quality flags, not verified market benchmarks.
Regression scope and uncertainty
The model uses the supplied 177-row Desa ParkCity sample, not all 348 Desa ParkCity entries in the wider listing table. The upstream selection rule is not established. Both ordinary least-squares models include an intercept and are reproduced directly from those 177 observations.
The four-feature model has R-squared 0.7399, adjusted R-squared 0.7338 and residual standard error RM435,863. Its rooms coefficient has p = 0.136. The reduced model removes rooms; that does not establish that all collinearity or other assumption issues have been resolved.
The RM2,129,275 and RM2,226,460 example estimates use 4 bathrooms, 3 parking spaces and 2,200 sq ft; the full model also uses 3 rooms. Residual error is reported separately, not subtracted from the prediction. Plus or minus one residual standard error is not a 95% prediction interval. These are in-sample fits, with no measured out-of-sample accuracy.
Business context and source material
The case describes a 20% joint-profit-sharing business model and a one-year revenue-growth target of 10%. Profit per deal, completed transactions, conversion rates and operating costs are not provided, so revenue impact cannot be calculated from listing prices.
Analysis workbook · Historical case study, not a live property feed. The article uses a saved workbook export from September 2026.
Interactive-story architecture informed by Al Harkan's scrollytelling documentation. Original implementation with locally hosted D3. The architectural cover is AI-generated illustration and does not depict the listed properties.