What drives house prices?

A mortgage lender's collateral valuation team checks every appraisal it receives against a statistical benchmark. Appraisals far outside the benchmark range go to a senior valuer, which costs time and delays the loan. The team asks you to build that benchmark for Wadi Noor, a fictional city with five districts, from 800 villa and townhouse sales registered in 2023 (dataset housing, synthetic).

The team needs answers to three practical questions:

1. Which property features are associated with price, and by how much — in terms a credit officer understands?
2. Why do some colleagues say “bedrooms add SAR 290,000 each” while others find bedrooms add almost nothing?
3. For a specific property, what range of prices is normal, so that an appraisal outside it is flagged?

The decision that depends on your work: where to set the flagging range. Too narrow and valuers drown in false alarms; too wide and inflated appraisals slip through, putting the lender's collateral at risk.

All sales are synthetic and generated for teaching; they describe no real city or market.