Skip to content
Evalúa/Trust & accuracy/Methodology
Methodology overview

How we actually grade a listing.

Evalúa is a research tool, not a black box. This page lays out what data goes into a report, how the score is weighted, and where the model is least confident — so you can judge our reasoning for yourself. The exact calibrations behind the score are proprietary; everything else is here.

The model · in production
0–100score

A weighted composite — three published category weights, independent indicators, five verdict bands.

Updated regularlyHuman-reviewed
Data points / listing70+
Score range0–100
Category weights3
Core marketSamaná
01  ·  PIPELINE

Six stages, about a minute.

From a pasted URL to a finished report. Each stage is checkpointed and logs its own confidence.
01
IngestScrape + parse
02
NormalizeDisambiguate
03
Comp matchFind comparables
04
Yield modelSimulate returns
05
Tax & legalApply DR rules
06
VerdictRoll up

Scrape + parse.Pull the public listing page and extract the underlying facts — price, size, type, location, and features. We work only from facts, never from the listing’s photos or marketing copy.

Disambiguate.Resolve the zone, unit type, and build year, and flag listings whose fields don’t add up. A meaningful share of portal data is mislabeled.

Find comparables.Match against active listings that are nearby, similar in size and type, and recently listed. Outliers are trimmed; the exact matching rules are internal.

Simulate returns.Estimate twelve months of rental performance — season-adjusted daily rate × occupancy, net of operating expenses and tax.

Apply DR rules.Apply Dominican jurisdiction rules — CONFOTUR eligibility, transfer tax, IPI, HOA, and rental-licensing checks.

Roll up.Roll the indicators into the weighted 0–100 score, then write the strengths, watch-items, and plain-language verdict.

02  ·  DATA SOURCES

What we pull, how often, how much we trust it.

Sources are weighted by reliability before anything enters the comp set. The exact weighting is internal.
SourceCadenceUsed for
Public DR listing portalsaggregated market listings
frequentComps
Broker & agency websitesfederated · many agencies
dailyComps
CONFOTUR registryMinisterio de Turismo · public
weeklyTax status
Public short-term-rental datadaily rate · occupancy signals
dailyRevenue
Banco Central de la Rep. Dom.FX · CPI · macro · public
weeklyContext
Public title & property recordsRegistro Inmobiliario · checks
as neededFlags
Evalúa field teamLas Terrenas · on-the-ground
ongoingVerification
Note on source weighting

Why we don’t treat every portal as equal.

The largest DR portals still carry listings that are stale, mislabeled, or duplicated. We run a cross-portal dedup on every ingest, verify against public records where they exist, and treat sharp price drops as signals to investigate rather than as clean comps.

  • Comp-set entries need corroboration across independent sources
  • Qualitative signals only shape the written context, never the number
  • A field-team visit overrides any automated classification
  • We keep an internal audit trail of every run
03  ·  CORE FORMULAS

The numbers a report is built on.

No proprietary magic in the formulas themselves — the same calculations any careful analyst would run. The work is in the parameters we calibrate.
Formula 01 · Fair price

Adjusted $/m² vs. comps.

A weighted median price per m² across the comparable set, then adjusted for unit-level features — floor, view, finish grade, and age.

Pfair =Σ wi · Pi/m²iΣ wi×subj × αfeat
wicomp weight · recency × proximity × size similarityαfeatfeature adjustment · floor, view, finish, ageNcomp count · thin sets flagged low-confidence
Formula 02 · Net yield

After everything annual rate.

Annual rent net of operating expenses, divided by the all-in acquisition cost. The number that actually matters if you’re buying to rent.

Ynet =(ADR · OCC · 365) − OpexPpurchase + Closing + Furnishing
ADRaverage daily rate · season-blendedOCCoccupancy · zone-specific, 12-month rollingOpexHOA, management, utilities, tax, maintenance reserve
Formula 03 · Liquidity proxy

How fast a zone moves.

The DR has no public sold-price registry, so we can’t measure closed sales. Instead we proxy liquidity from listing dynamics — how quickly listings are removed, and whether days-on-market are rising or falling.

L =Removed90dActivetoday·DOMtrend
L > 1.2tightening · listings clearing quickly0.8 – 1.2balanced · healthyL < 0.8cooling · favor the buyer
Formula 04 · CONFOTUR savings

The incentive math.

Applies only if the project is registered and the closing falls inside the 15-year eligibility window. We verify against the Ministerio de Turismo registry on every run.

S =(Ttransfer + Tipi · 15y)· β
Ttransfer3% of purchase · waived under CONFOTURTipi1% annual property tax · waived 15 yearsβeligibility flag · 0 or 1
04  ·  COMPOSITE SCORE

Three weights, one number.

Independent indicators roll up into three weighted categories. The category weights are published; indicator-level calibration is proprietary.

How a score gets built.

The composite is the weighted sum of three category scores, each built from independent indicators. We publish the breakdown on every report — a single headline number is useful, but you should always read the breakdown before acting.

No subjective inputs. Every factor traces back to a measurable number — a price comparison, an occupancy rate, a regulatory flag. If a subscore can’t be computed, the whole report is flagged low-confidence rather than fudged.

78/100Good — revenue-led, CONFOTUR-confirmed, costs clean.
01
Revenue potential40%Gross yield, occupancy, average daily rate, and net operating income vs. zone benchmarks.
02
Cost of ownership35%Acquisition cost, monthly operating burden, break-even occupancy, and upkeep.
03
Value assessment25%Asking price vs. comparable listings, plus a premium-features score.

Category weights are published on every report. How each indicator is computed and calibrated is proprietary.

05  ·  KNOWN LIMITATIONS

Where we’re less confident.

We publish our weaknesses. Read the confidence flag on a report before acting on it.
Data gap

Pre-construction pricing is unreliable.

Developer price sheets are marketing, not market, and comp sets for planned buildings are thin by definition. We flag these as low confidence and hold the score back until enough comparable units are verified.

Flagged low-confidence
Coverage

Rural & secondary markets are thin.

Outside our core tracked zones — centred on Samaná, with wider DR tourism markets — comp density drops below our floor, and the report is gated behind a manual review.

Outside core zones
Model assumption

Rental occupancy is a point estimate.

We use trailing zone occupancy from public short-term-rental data. New supply, regulation, and macro shocks can move it, so our stress test applies a sensitivity band to every yield.

Sensitivity published
Scope

We don’t score intangibles.

Neighborhood feel, view quality, neighbor relations, HOA politics — these matter, and we deliberately don’t invent a number for them. They surface as qualitative flags in the “watch” section of a report.

Design decision
Legal

CONFOTUR status can change.

Projects can be added to or removed from the Mitur registry between our checks. We verify at report time, but any closing more than a few weeks later should be re-verified directly with a DR attorney.

Verify at closing
Human judgment

This is not a substitute for a lawyer.

Evalúa flags what it can measure. Title disputes, inheritance structures, and foreign-ownership edge cases need a licensed DR real-estate attorney and an in-person site visit. Always.

Always
06  ·  CHANGELOG

We version the model like software.

Older reports remain viewable at the version they were generated under, for audit.
Now2026

Continuously improved. The scoring model is versioned and updated periodically. Material changes that affect scores are summarized here on release; detailed calibration notes stay internal. Older reports remain viewable at their original version.

Current
07  ·  FAQ

Questions we keep getting.

If yours isn’t here, write contacto@evalua.do and it’ll shape the next revision.
№ 01Is Evalúa an appraisal?+
No. An appraisal (avalúo) is a legal document signed by a licensed appraiser, typically required for a mortgage closing. Evalúa is a research tool — a data-driven second opinion you use before or during due diligence, to decide whether to make an offer and to pressure-test the one you’re considering. For a closing you’ll still need a DR-licensed tasador.
№ 02How do you handle listings with bad data?+
A meaningful share of portal listings have at least one misclassified field — wrong square meters, wrong bedroom count, wrong zone. Our normalize step cross-checks the numbers against public records and the rest of the listing, and if it can’t reconcile them, a person reviews the run. We don’t silently correct — the report tells you which field we overrode and why.
№ 03Why don’t you score Punta Cana like Samaná?+
Punta Cana’s market is dominated by large resort operators with bundled rental programs, which changes the yield math fundamentally. The model accounts for that, so a “good” Punta Cana score typically caps lower than a “good” Samaná score. It isn’t a bias — it’s a structural difference in who ends up owning the unit’s income stream.
№ 04Can I audit a specific report?+
Every report shows the comparable listings used and the category-by-category score breakdown, and each report has a shareable link. The model internals — the indicator weights and adjustment formulas — are proprietary and aren’t exported.
№ 05How do you make money?+
The tools are free. Evalúa is built by a Las Terrenas real-estate team, and when you’re ready to act on a report we can introduce you to a vetted local partner; if that introduction leads to a transaction, we may earn a referral commission. That relationship is disclosed, and it never changes a listing’s score — the model runs the same math whether or not you ever contact us. We don’t sell placement in reports, and we don’t let developers or agencies pay to move a number.
№ 06How often is the model updated?+
The underlying data refreshes on a regular cadence. The model itself — the weights, formulas, and eligibility logic — is versioned like software and updated periodically after a backtest cycle. Material changes are noted in the changelog above, and older reports stay viewable at the version they were generated under.

Read the reasoning. Then run a listing.

Our confidence and our reasoning are published. Drop in a URL and you’ll see the same methodology applied to a property you actually care about.