Inspiration
A property with high rental revenue is not necessarily a good investment—it may also be expensive to buy. We wanted to turn public data into a clear answer to DDD’s question: where should a fictional fund allocate $10 million across Vancouver short-term rentals?
We focused on the trade-off between buying a few expensive properties and buying more modest ones.
What our project does
Our analysis compares neighbourhoods, property types, and bedroom counts using estimated annual revenue divided by estimated purchase price.
Our final gross-revenue allocation proposes 15 condos across three neighbourhoods:
- Downtown Eastside: seven one-bedroom condos.
- Downtown: three one-bedroom and two two-bedroom condos.
- Kitsilano: three one-bedroom condos.
The estimated purchase cost is $9.87 million CAD, supporting approximately $765,000 in estimated annual gross revenue under a constant-revenue scenario for 2027.
The resulting 7.75% is a gross revenue-to-price ratio, not profit or net investment return. Operating expenses, acquisition costs, furnishing, and working capital require a separate all-in feasibility assessment.
How we built it
We combined two public data sources:
- Inside Airbnb: listing characteristics and estimated revenue and occupancy.
- City of Vancouver Open Data: property assessments and parcel information used to construct neighbourhood-level price proxies.
Using Python, we cleaned price fields, filtered minimum-stay requirements, and grouped active entire-home listings by neighbourhood, property type, and bedroom count. We adjusted condo price estimates using explicit bedroom multipliers.
We then applied transparent allocation rules:
- At least 30 comparable listings per segment.
- A maximum of 40% of the purchase budget in one neighbourhood.
- A purchase limit tied to each segment’s observed listing supply.
- Near-tied bedroom sizes could share an allocation.
We visualized the comparisons and documented the reasoning behind our decisions.
Challenges we faced
The data contains estimates rather than verified booking receipts. Assessed values are also not confirmed purchase prices, and listings within the same segment can differ substantially.
Small samples were another challenge. We tested how sample size affects estimated median revenue through repeated subsampling. This helped inform our screening threshold, while showing that larger samples still do not guarantee accurate predictions.
Finally, balancing revenue efficiency with concentration and bedroom mix required explicit trade-offs rather than simply choosing the highest-ranked option.
What we learned
We learned to distinguish gross revenue, operating profit, and investment return. We also learned that a transparent method is only useful when its assumptions and limitations are clear.
Our comparisons support condos over houses within the eligible segments and price assumptions we tested. However, that finding does not establish that every individual condo is a better investment.
AI assisted with coding, analysis, and presentation drafting. We checked important calculations against source tables and reviewed allocation totals, constraints, and alternative rules.
What’s next
Before turning the allocation into an executable investment plan, we would validate available purchase prices, confirm building permissions, obtain operating-cost estimates, and reselect properties within an acquisition-and-launch budget.
We would also test revenue assumptions against another time period.
Data attribution
Data: Inside Airbnb, Vancouver, 15 June 2026, CC BY 4.0.
Additional data: City of Vancouver Open Data, property assessments and parcel information.
The project follows the business case’s simulation rule waiving the principal-residence requirement.
