Full transparency on our data pipeline, scoring algorithms, and flood risk modeling. From raw data ingestion to AI-powered ranking - here's the entire stack.
We collect apartment and house listings from public real estate portals in Paraguay. Fully automated, updated periodically.
The data includes prices, areas, coordinates, amenities, construction state, seller type - everything that's publicly available on each listing.
Prices come in both USD and Paraguayan Guaranies. We convert everything to USD at the current exchange rate, because the USD fields on some portals are unreliable for Gs. listings.
Real estate listings are messy. Here's what we catch:
Some listings show a down payment (anticipo/cuota) as the sale price. A "75,000 USD" apartment that's actually $75k down on a $150k property.
When m² includes common areas, hallways, or terrain. We detect these cases and correct when the discrepancy is significant.
Project listings that advertise the cheapest unit. The actual apartment you'd get at that "price" may not exist. Flagged and penalized.
Listings marked "A estrenar" (brand new) that are actually under construction. We reclassify them because "under construction" and "move-in ready" are very different value propositions.
Errors in areas or prices that don't make physical sense. Automatically detected through cross-validation of the data.
Flagged listings aren't removed. They get a quality penalty that pushes them down the rankings. You still see them. You just know they're suspect.
Is this underpriced?
We answer one question: is this property cheap for what it is? Each property is compared to similar ones - same neighborhood, same number of bedrooms. We rank by price per m², so the cheapest rises to the top. Then we layer in condition (new > renovated > used), amenities (parking, pool, security), and how long it's been listed. A score of 80+ means it's in the cheapest ~20% of its peer group while being in good condition.
| Component | Weight | What It Measures |
|---|---|---|
| Price vs Peers | 45% | Percentile rank of price per m² within neighborhood and bedroom count. Cheapest = highest score. |
| Construction Quality | 20% | Brand new and excellent condition score highest. Needs-renovation scores lowest. |
| Amenity Value | 15% | Parking, pool, gym, security - weighted by what buyers value most. |
| Market Freshness | 20% | How long has it been listed? Fresher listings score higher - the market hasn't passed on them yet. |
Will this generate returns?
The investment score is for buy-to-rent investors. We estimate monthly rent by matching similar rental listings, then calculate yield - but with a twist for pre-construction.
An "en pozo" apartment that won't deliver for 2 years can't generate rent during that time. We add the lost rental income to the purchase price before calculating yield. A $100K pre-construction apartment with $500/month estimated rent has an effective cost of $112K (= $100K + 24 months × $500), reducing its yield from 6.0% to 5.4%.
We also factor in whether you can actually exit the investment (neighborhood liquidity), whether the area is appreciating (trend), and physical risk (flooding).
| Component | Weight | What It Measures |
|---|---|---|
| Adjusted Yield | 30% (Apts) / 20% (Houses) | Annual rent ÷ effective cost. Pre-construction adds lost rent to the price. |
| Price vs Neighborhood | 25% | Buying below market is the first win. |
| Neighborhood Liquidity | 20% | More listings + faster sales = easier exit. |
| Neighborhood Trend | 15% | Year-over-year price change. Rising neighborhoods appreciate. |
| Flood Risk | 10% (Apts) / 20% (Houses) |
Listings with data red flags (price looks like a down payment, inflated area, "starting from" pricing) get their scores reduced. No listing drops below 10% of its raw score - even suspicious listings deserve some visibility.
| Issue Detected | Penalty | Effect |
|---|---|---|
| Down payment as sale price | 0.20x | Severe. Price is probably wrong. |
| Area includes terrain/common | 0.50x | $/m² is artificially low |
| "Starting from" price | 0.55x | Real price likely higher |
| Pre-construction mislabeled | 0.90x | Not move-in ready as claimed |
| Stale listing (>90 days) | 0.45x | May be sold or overpriced |
Live breakdown of data quality flags detected across all listings. Red flags indicate deal-breaking issues; amber flags are moderate concerns.
How each scoring component is distributed across all listings - understand where scores cluster and spread.
How individual amenities affect listing prices across the market
All data comes from real estate listings and may contain inaccuracies. Our data cleaning pipeline automatically detects and corrects the most common errors - but always verify details directly with the seller before making decisions.
The scoring weights reflect our best judgment of what matters in the Greater Asuncion market. They are not calibrated against historical sale outcomes - we don't have access to closed-sale prices. Pre-construction yield adjustment assumes 24 months for 'en pozo' and 12 months for 'en construcción' - actual delivery times vary. Use as a screening tool to find interesting listings faster, then do your own due diligence.
| Elevation-based flood model. Higher weight for houses (ground level). |
Houses weight flood risk more (20% vs 10%) because they're at ground level. Apartments in flood zones are typically elevated and less affected. Houses also weight yield less (20% vs 30%) since they're more often owner-occupied.