Scoring Methodology
How Avena Terminal rates and ranks every new build property in coastal Spain
Overview
Avena Terminal assigns every tracked property a composite investment score between 0 and 100. The score is not an opinion. It is the output of a quantitative model that decomposes each property into five measurable dimensions, weights them according to their empirical contribution to long-term investment returns, and normalises the result against the full universe of tracked listings. A score of 80 means the property sits in the top quintile across all five dimensions relative to every other new build we monitor in coastal Spain.
The model is re-run daily. When a price changes, a new comparable sale appears, or occupancy data updates, the affected scores adjust automatically. There is no manual override and no pay-to-rank mechanism. Every property is evaluated by the same formula, using the same data pipeline.
The theoretical foundation is hedonic regression, a widely accepted econometric technique in real estate valuation research. Originally formalised by Sherwin Rosen (1974) and later extended by property economists such as Sirmans, Macpherson, and Zietz (2005), the hedonic approach treats a property's market price as the sum of implicit prices paid for its individual characteristics: size, location, build quality, amenities, and yield potential. Avena Terminal adapts this framework to the Spanish new build market by incorporating region-specific benchmarks and short-term rental performance data that do not exist in traditional hedonic datasets.
The Five Scoring Dimensions
Each dimension captures a distinct aspect of investment quality. The weights reflect the relative importance that backtested data assigns to each factor in predicting total five-year returns (capital appreciation plus rental income) across the Spanish coastal property market.
1. Value
40%The value dimension measures how a property's asking price per square metre compares to the prevailing resale benchmark for its specific municipality. We source resale transaction data from the Registradores de Espana (Spanish Property Registrars) and supplement it with listing-price indices from Idealista and Fotocasa to construct a rolling 12-month median price per square metre for each town.
A new build priced 15% below the local resale median, for instance, signals embedded value: the buyer is acquiring a brand-new asset at less than existing stock trades for in the same area. The model converts this percentage discount (or premium) into a 0-100 sub-score using a sigmoid mapping function that rewards deeper discounts with diminishing marginal returns, so a 30% discount does not score twice as high as 15%.
Value carries the largest weight (40%) because empirical research consistently shows that entry price relative to local benchmarks is the single strongest predictor of medium-term investment outcomes in residential property. Overpaying at the point of purchase is the hardest mistake to recover from.
2. Yield
25%The yield dimension estimates gross annual rental income as a percentage of the purchase price. We derive nightly rate estimates from short-term rental platforms (Airbnb and Booking.com comparable listings within the same postcode area) and multiply by an occupancy-adjusted annual revenue figure.
Occupancy rates are seasonally weighted: peak summer months (June through September) carry higher utilisation assumptions than winter months, calibrated by region. A property on the Costa del Sol with year-round tourism may carry an 70% average occupancy assumption, while a Costa Blanca North listing with a shorter season may use 55%.
The resulting gross yield is normalised against the full tracked universe to produce a 0-100 sub-score. Properties generating yields above 6% gross typically score in the 75-100 range, while yields below 3% fall into the bottom quartile. This dimension receives 25% weight because rental income represents the primary ongoing cash-flow return for buy-to-let investors and substantially influences total return calculations.
3. Location
20%Location quality is evaluated through a composite of measurable proxies: straight-line distance to the nearest beach, driving time to the nearest international airport, density of amenities (restaurants, supermarkets, healthcare) within a 2-kilometre radius, and the historical five-year compound annual growth rate (CAGR) of property prices in the municipality.
Price appreciation data comes from the INE (Instituto Nacional de Estadistica) housing price index at the provincial level, supplemented by Registradores transaction-level data where municipal granularity is available. Municipalities with CAGRs above 8% over five years score higher than those with flat or declining price histories.
At 20% weight, location captures the appreciation potential and lifestyle desirability that drive both resale value and rental demand. Two otherwise identical properties can score very differently if one sits in a high-growth beachfront town and the other in a stagnating inland municipality.
4. Quality
10%The quality dimension evaluates the physical and specification attributes of the property itself. Inputs include energy efficiency rating (A-rated builds score highest), the presence of private parking, communal or private pools, terrace and garden area relative to total built area, and the number of bedrooms and bathrooms relative to the development average.
Developer track record is also factored in where data is available. Repeat developers with completed delivery histories receive a small bonus versus first-time developers. This dimension carries 10% weight because, while quality matters for long-term maintenance costs and tenant satisfaction, it is secondary to price and yield in determining investment returns.
5. Risk
5%Risk measures the delivery and execution uncertainty associated with a given listing. Key-ready properties with completed building licences score highest. Off-plan developments that have not yet broken ground carry the highest risk penalty. The model also considers estimated completion timelines: a project 18 months from delivery scores lower than one delivering in 3 months, reflecting the opportunity cost of capital and the probability of construction delays.
At 5% weight, risk is the smallest component. This is deliberate: the Spanish new build market benefits from strong consumer protection (bank guarantees on deposits under Ley 20/2015), which limits downside in most scenarios. However, the dimension still penalises early-stage projects appropriately and ensures that key-ready stock receives a small but meaningful advantage in the composite score.
Hedonic Regression Model
The term "hedonic regression" refers to a class of statistical models that estimate the implicit price of individual product characteristics by regressing observed transaction prices on a vector of attribute variables. In the context of real estate, this means decomposing the price of a property into the value contributions of its location, size, condition, amenities, and other features.
Avena Terminal uses a semi-log hedonic specification where the natural logarithm of price per square metre is regressed against dummy variables for municipality, property type (apartment, townhouse, villa), bedroom count, and a continuous variable for distance to coast. The residual from this regression represents the "unexplained" portion of price, which the model interprets as the degree to which a property is over- or under-priced relative to its peers.
This residual feeds directly into the Value dimension. A large negative residual (price significantly below what the model predicts for a property with those characteristics) produces a high Value sub-score. The regression is re-estimated monthly using rolling 12-month transaction data to capture evolving market conditions.
The approach draws on established academic work including Rosen (1974), Sirmans et al. (2005), and Malpezzi (2002), as well as the European Central Bank's residential property price methodology, which also employs hedonic techniques for cross-country housing market comparisons.
Data Sources
Primary listing data: prices, specifications, locations, images, and availability for new build developments across coastal Spain. Updated daily.
Official Spanish housing price index at provincial level. Used for location CAGR calculations and macro-market benchmarking.
Transaction-level resale price data at municipal level. Powers the Value dimension benchmarks.
Nightly rate and occupancy estimates for short-term rental yield calculations, sourced from comparable listings within the same postcode.
Supplementary listing-price indices for local market benchmarking where Registradores data lacks municipal granularity.
Mortgage rate data and housing market indicators used in macro-economic context layers.
Confidence Intervals
Not all scores carry equal certainty. The confidence of any individual property score depends on the richness and recency of the underlying data. Avena Terminal assigns each score a confidence tier:
Municipal resale data available within the last 6 months, at least 5 rental comparables within the postcode, and key-ready status confirmed.
Provincial-level benchmarks used (municipal data unavailable), 2-4 rental comparables, or off-plan status with confirmed building licence.
Sparse transaction history in the area, fewer than 2 rental comparables, or very early-stage development with limited specification data.
These confidence margins are derived from bootstrap resampling of the hedonic regression residuals. By re-estimating the model 1,000 times on randomly drawn sub-samples of the transaction dataset, we obtain an empirical distribution of predicted scores for each property and report the 90% confidence interval width. Properties in data-rich municipalities (such as Marbella or Torrevieja) tend to have narrow intervals, while listings in smaller towns with few transactions carry wider uncertainty bands.
Academic References
- Rosen, S. (1974). "Hedonic Prices and Implicit Markets: Product Differentiation in Pure Competition." Journal of Political Economy, 82(1), 34-55.
- Sirmans, G.S., Macpherson, D.A., & Zietz, E.N. (2005). "The Composition of Hedonic Pricing Models." Journal of Real Estate Literature, 13(1), 1-44.
- Malpezzi, S. (2002). "Hedonic Pricing Models: A Selective and Applied Review." Housing Economics and Public Policy, Blackwell.
- European Central Bank (2023)."Residential Property Prices: Methodological Framework." ECB Statistics Paper Series.
- Banco de Espana (2024)."Spanish Housing Market Monitor." Financial Stability Report.