DWELLVENTA SIGNAL V2.3 · OPEN METHODOLOGY

A score you can
look inside.

Follow the evidence from a published asking price to a useful research shortlist. Every factor has a source, a weight and an evidence strength. Missing facts stay visible.

Explore scored properties ↗
THE EVIDENCE-ADJUSTED FORMULAStart at 50.

Add each factor's weighted deviation from 50, multiplied by its evidence strength. Check eligibility first: explicit severe restrictions withhold a number pending source review. Cap and round to 0–100.

50 + Σ(w × r × (factor − 50)) − flags

What does my score mean?

Every assigned score runs from 0 to 100 and is rounded to a whole number. Higher means a stronger signal for that score’s research purpose. 50 is the neutral starting point, not an average property or a 50% chance of success. A 70/100 is not a 70% return, a percentile, a valuation or a recommendation to buy.

Asking-price Signal

Higher means stronger supported asking-price and housing-market signals. The standard and custom Signal share this scale, but custom weights can change the result.

  1. 0–49Weaker signal
  2. 50–69Mixed signal
  3. 70–100Stronger signal

A lower asking price relative to compatible peers, positive housing trends, more stable indices, published price reductions and shorter metro time to pending can raise a home’s score. Opposing supported evidence can lower it. Explicit active environmental or demolition restrictions and investor exclusions withhold a numeric score pending source review. A low ask may still reflect repairs or restrictions. Land uses only compatible land asks and published price movement.

Flippability

Higher means stronger supported resale-research signals, not a predicted resale value or flip profit. Read the number relative to its neutral starting point:

  1. 0–49Below neutral
  2. 50Neutral baseline
  3. 51–100Above neutral

There are no weak/mixed/strong numeric bands for this model. The separate “Strong flip research lead” badge requires all of: a score of at least 65, an asking price at least 20% below the compatible asking-price benchmark, and at least 40% evidence coverage, after eligibility and peer checks. A score of 65 alone does not earn the badge.

Comparable asking-price position, housing momentum and metro time to pending move the score. Reported research inputs affect evidence coverage only and add no opportunity points. None establish after-repair value, contractor costs or legal access.

Rental Potential

Higher means stronger supported long-term rental-research signals. Its stronger-signal threshold is 65, not the Asking-price Signal’s 70.

  1. 0–49Weaker signal
  2. 50–64Mixed signal
  3. 65–100Stronger signal

Compatible acquisition asks, local rent growth and stability, longer-term rent trends and bounded economic context can raise or lower it. Reported research inputs affect evidence coverage only and add no opportunity points. When an acquisition comparison is missing, no property Rental Potential score is assigned. Available area rent history remains separate descriptive context, with only its supplied scope, segment, observation date and direction displayed. A missing date or scope is left unreported. Area rent indices are not property-specific rent, cash flow, vacancy or permission to rent.

Research priority

Higher means review the source record sooner. It is a research queue, not a claim of greater distress, confirmed availability or investment value.

  1. 0–49Lower current priority
  2. 50–69Standard review
  3. 70–100Review sooner

More specific published record signals, recent source observations, clear parcel/address identity, actionable next steps and reported details can raise priority. Historical or expired notices reduce it. A record can have a priority score without an asking-price Signal.

Compare like with like. A Rental Potential score of 70 and a Research priority score of 70 answer different questions. Compare only within the same model, snapshot context and custom strategy. A high score with limited evidence deserves more investigation, not more certainty.

Score strength is not evidence strength.

Read two separate measures. The 0-100 number describes the selected research signal. Weighted input coverage describes how much of that model has usable support after source, geography, sample-size and age adjustments. A listed input may have only partial support. Neither number is a probability, a percentile or a measured accuracy rate.

Evidence coverage is the weighted share of usable model inputs, adjusted for support such as sample size, geography and observation age. It is displayed separately from the 0–100 score:

For example, 72/100 with 38% weighted input coverage means a stronger Asking-price Signal with moderate supported evidence. It does not mean a 72% chance of profit or 38% accuracy. Missing factors contribute no change from 50; weaker evidence shrinks each factor’s influence toward zero rather than transferring its weight to known factors.

Dates answer another question. “Calculated” is the model’s as-of date. A listing snapshot records a source retrieval; the authority’s observation may be older. Refreshing a download does not make an old record new. Read original observation dates, status and deadlines before acting.

A dash is not a zero. A score may be loading, temporarily unavailable, not applicable, awaiting a fresh snapshot, or missing required price, peer, local rent or identity evidence. The property or record explains the reason. Unscored does not mean worthless or low priority.

Pro reports show signed factor contributions: positive points raise the result, negative points lower it, and unsupported factors stay neutral. The unknown-evidence sensitivity range shows how unsupported modeled inputs could affect the result; it is not a statistical confidence interval and does not cover every due-diligence risk.

Different scores. Different decisions.

Asking-price Signal

For compatible homes or land with a verified asking price and independent peers. Describes asking-price and market context, with evidence coverage. Asking prices are not completed sales.

Flippability

Resale research for recent, priced residential offerings. Requires at least five compatible retail REO asking-price peers, an effective sample of three, price reliability of 0.35 or more and no extreme subject-price outlier. Closed-sale after-repair value and renovation costs remain unknown.

Rental Potential

Long-term rental research using local rent trends and supported acquisition evidence. Area rent indices are not property rent, net income, cash flow or rental permission. No national rent fallback is used.

Research priority

For imported inventory, tax, foreclosure, vacancy, safety and holdings records. Helps order the work queue even when no price is published. It does not assert that a property is available or distressed today.

Evidence coverage

The weighted proportion of model inputs supported by usable evidence. It is separate from the score and is not a probability of accuracy, profit or success. Title, condition and other due diligence remain separate.

Nationwide coverage, measured.

Loading the published scoring audit…

These counts describe dated imported records, not unique available homes. Record-only leads can receive research priority; an opportunity Signal still requires defensible asking-price comparisons. Coverage is not manufactured by substituting a tax balance, auction bid or assessed value for a purchase price.

Fixed weights for the property type.

FactorHomesLandRule and evidence
Comparable asking-price position40%75%50 + 100 × relative gap to the weighted peer median, with the gap capped at ±45%. Similarity, sample size, dispersion, observation age and geography reduce influence.
12-month housing trend15%0%50 + 250 × annual change. Requires 13 positive consecutive monthly observations.
Longer-term housing trend10%0%50 + 400 × annualized growth over 36 months, or 24–35 months with reduced evidence strength. No gap interpolation.
Index downside stability10%0%75 − 150 × maximum drawdown − 200 × annualized log-return variability over up to 36 months. These are smoothed housing-index characteristics, not individual-property risk estimates.
Published asking-price movement15%25%50 + 100 × asking-price reduction, capped at ±30%. At least two same-source observations seven days apart; latest observation must match the current asking price.
Metro median time to pending10%0%80 − median days from listing to pending, capped at 0–100. Metro context has limited influence and excludes the closing period.

Each factor is capped at 0–100. Missing evidence has strength 0 and contributes no change from 50. We never redistribute missing weights. Explicit reported environmental and demolition flags each subtract 15 points in the standard model. Unreported risks are unknown, rather than cleared.

Pro customers can choose six strategy weights and a flag deduction. Legacy four-factor strategies retain their first four weights. Land uses the strategy's price and price-movement allocation only, proportioned before evaluating missing evidence; a strategy with neither receives no custom land score.

Economic context with a bounded influence.

For eligible homes, Signal v2.3 adds a context adjustment to the six-factor property score. The adjustment is capped at ±5 points and multiplied by property evidence coverage. The reserved 10% housing-stock weight is not used and contributes zero points; it is not redistributed to other factors. It never creates a price, a comparison cohort, a land valuation or an opportunity score for an unscored record.

final = clamp(property score + property coverage × Σ(context weight × evidence × (points − 50) / 10))

Economic conditions · 30%

BLS unemployment (6%), its three-month change (9%), BEA PCE inflation (7.5%) and real GDP growth (7.5%). These national aggregates describe broad conditions.

Mortgage conditions · 35%

Freddie Mac's latest 30-year fixed benchmark compared with the preceding 10-year arithmetic weekly average (22.75%), plus its 13-week change (12.25%). The 1-year, 5-year and since-April-1971 averages are additional descriptive comparisons. Each average excludes the newest observation, needs at least 95% of expected weeks and rejects gaps longer than 14 days. Missing history stays neutral.

Housing supply · 25%

Zillow metro or explicitly labeled national for-sale inventory change (15%) and Census/HUD new-home months of supply (10%). Starts and permits are descriptive pipeline indicators with no extra points.

Housing stocks · 10% reserved

Not used in scoring. No automated stock-price or total-return history is ingested, displayed as model evidence or scored. The reserved weight stays neutral; it is not reassigned. Pro may display delayed provider-hosted quotes for D.R. Horton (DHI), Lennar (LEN), PulteGroup (PHM), Toll Brothers (TOL) and KB Home (KBH). TradingView retains its attribution and controls the delayed Cboe One display. Quotes are not extracted, exported or used as automated score inputs.

Fixed factor points: unemployment = 50 + 10 × (5 − rate); three-month unemployment change = 50 − 60 × change in percentage points; PCE = 50 − 15 × max(0, |inflation percent − 2| − 0.5); quarterly annualized real GDP growth = 50 + 7 × growth percent; mortgage historical gap = 50 − 12 × (current rate − preceding 10-year average), measured in percentage points; mortgage change = 50 − 30 × change in percentage points; inventory change = 50 − 100 × year-over-year fractional change; new-home supply = 50 + 6 × (6 − months). All points are capped at 0–100. These thresholds are transparent heuristics, not fitted or calibrated predictions.

Rising inventory gives buyers more choice while increasing resale competition; the score uses the resale-risk interpretation. Counts refer to listings active at any time during the month, not currently buyable Dwellventa homes. National supply is a labeled fallback, not city inventory. Metro mapping uses exact published geography or an audited existing alias; unsupported county locations cannot borrow a city's metro.

Rates older than 35 days, monthly macro/supply observations older than 120 days, GDP quarter ends older than 150 days, or snapshots retrieved more than 35 days ago contribute nothing. Observation ages are measured from the completed period. Future observations and snapshots are rejected. Exact prior monthly or quarterly periods are required for changes; no interpolation fills gaps. Missing inputs keep their fixed weights and stay neutral. The unknown-evidence sensitivity range also includes their bounded possible economic influence. Source snapshots refresh on demand with a six-hour cache; failed fetches retain their original date.

Pro reports and CSV exports show all values, dates, source links, statuses and contributions. Public browsing receives headline scores and basic evidence/eligibility explanations; peer records, detailed factors and contributions remain in protected reports. Revised historical series cannot establish what was available at a past decision date; no release-vintage backtest is claimed.

Freddie Mac PMMS ↗ · FRED / original BLS, BEA, Census and HUD sources ↗ · Zillow inventory definitions ↗

Building-material trends are cost context, not hidden score points.

Explore all nine material trends ↗ for free, with dated charts and source definitions. No account, property selection or estimate request is needed.

National BLS producer-price indices, published through FRED, track lumber, steel, concrete, gypsum, roofing, paint, plumbing fixtures, electrical wiring devices and hand tools separately. Monthly and yearly changes compare the exact prior month and the same month one year earlier. Missing periods are not interpolated. Higher growth means more producer-price pressure; it does not imply a better property or a measured local renovation cost.

No automatic score points. A material index cannot establish a property’s scope of work, quantities, local quotes or contractor costs. Material trends are independent research context. The optional Deal lab scenario uses a materials-only budget, its baseline month and category percentages supplied by the user. No dollar estimate or material mix is assumed.

The calculation applies the weighted ratio between each selected category’s latest common observed index and its exact baseline-month index to the entered materials budget. Allocations must total 100%; the baseline must be within 36 months. A change outside ±30% is refused rather than clipped. Every selected category needs valid, current observations. Retrievals more than 35 days old or completed observation periods more than 120 days old are not used for scenarios.

Review the difference before choosing to apply it to the rehab assumption. Only the materials-only difference is applied, and a revised scenario replaces the previous adjustment rather than stacking it. It does not change any property score. Labor, permits, freight, taxes, contractor margins and unlisted categories are not estimated; never apply a materials index to an entire rehab budget. Recalculate the deal scenario after changing its assumptions.

Indices are not dollar prices, retail shelf prices, installation costs or forecasts. The separate categories have different baskets and base periods; raw index levels are not comparable prices. Series are monthly, not seasonally adjusted, and may be revised. The category label and source definition state what each series covers.

BLS Producer Price Index ↗ · FRED source series ↗

Peers that belong together.

  1. Match city and state, residential subtype, and price regime. Administrative land-bank prices remain within the source program; HUD and HomePath retail REO can share a cohort. Homes and land remain separate.
  2. Prefer neighborhood or ZIP cohorts with at least three independent peers. Citywide comparison needs five peers with area, or ten homes with identical reported bedroom and bathroom counts when area is missing.
  3. Area must be between half and twice the subject's reported area. Reported home bedroom counts differ by at most one. Acres are converted using 43,560 square feet per acre; no parcel dimensions or usable area are invented.
  4. Exclude the subject and duplicate parcel or address identities. Keep the latest independent observation. Filter log-price extremes using a median absolute-deviation rule, then compute a similarity- and freshness-weighted median.
  5. Publish the scope, peer count, effective sample size, excluded outliers and comparison basis. Weight similarity by area ratio. Sample evidence is reduced below ten effective peers; broader and more dispersed cohorts have less influence.

A low asking price can reflect substantial repairs, restrictions or an unusual transaction. The Signal first gates explicit investor exclusions and active severe restrictions, then measures reported asking-price context; it does not estimate a discount to fair market value.

Local first. Broad context disclosed.

The housing-context hierarchy is neighborhood → ZIP → city → county → metro → state. Geographic strengths are 100%, 85%, 70%, 60%, 50% and 30%, respectively, before history-length and freshness adjustments. ZIP and county coverage can cross municipal boundaries. State fallback is labeled as state context.

Market observations older than 180 days are excluded. Future observations are excluded from the scoring date. Shorter continuous histories can support annual trends without pretending to support a three-year trend. Raw nationwide price-level CSVs are not republished; the portal stores derived indicators, geography metadata, observation dates, retrieval dates and original source links.

Price evidence uses source snapshots: up to 45 days has full recency strength, 46–90 days 90%, 91–180 days 70%, and 181–365 days 45%. Missing observation dates have limited strength. Closed, pending, withdrawn, coming-soon and expired offerings receive no asking-price Signal until a new eligible observation is available.

Residential housing indices never contribute land opportunity points. They may be displayed separately as surrounding housing context.

Research priority for sparse records.

Research priority starts at 50 and uses published record specificity (30%), source observation recency (25%), parcel/address identity (20%), next-step actionability (15%) and reported details (10%). Each contribution is adjusted by available evidence. Historical and expired notices are down-weighted. Download time does not replace an older reported record date.

Priority uses typed source records and no demographic, protected-trait, personal-contact or owner-behavior inputs. It does not determine eligibility for housing, credit or insurance. Free records show headline priority, historical status and authority links. Pro property reports add contribution breakdowns, missing evidence, comparable-price research and market analytics.

Unknowns stay visible.

Flip and Rental Potential also expose a missing-evidence sensitivity range when a numeric score is supported. The bounds vary unsupported portions of the fixed model factors, round outward and are not a probability or a statistical confidence interval. A withheld score has no invented numeric range. Legal, condition and operating-cost risks outside the model are not covered by these bounds.

The missing-input stress-test range varies unsupported portions of fixed model factors between 0 and 100 while keeping supported contributions fixed after eligibility checks. It shows sensitivity to missing inputs, not a statistically estimated interval or a prediction of the property outcome. Higher evidence means at least 65% weighted coverage; moderate evidence means 35–64%; limited evidence means below 35%.

Condition, title and liens, occupancy, access, taxes, zoning, utilities, flood or environmental reports, rehab costs and buyer eligibility need separate verification. We do not invent rents, resale prices, rehabilitation budgets, holding costs, return projections or clear-title findings to increase score coverage.

Validation before prediction.

This version is a documented heuristic, not a model trained or calibrated against investment outcomes. Its checks cover type and price semantics, independent peers, duplicate exclusion, outliers, bounds, order invariance, missing-data neutrality, monotonic price response, time cutoffs, stale/gapped histories, geographic fallbacks, source flags, CSV safety and protected custom analytics.

Future outcome calibration requires frozen data vintages, time-separated holdouts, geographic and property-type slices, coverage-band analysis and reviewed acquisition outcomes. Today's revised index history cannot stand in for the data available at a past decision date.

Flip research: qualify the evidence.

The public shortlist includes recent, priced residential offerings and excludes reported pending, closed, coming-soon, demolition and explicitly investor-restricted records. Homes lacking compatible price evidence remain research candidates. The Pro flip score uses 50% compatible asking-price position, 20% housing momentum, 15% liquidity and 15% research-input coverage. The research-input allocation stays at neutral points; it affects coverage, not opportunity. Unsupported evidence stays neutral. A score requires at least five compatible retail REO peers, an effective sample of three, price reliability of at least 35% and no subject outlier. A strong lead additionally requires a 20% asking-price gap, score of 65 or more and evidence coverage of 40% or more.

Retail REO listings and land-bank program prices are compared separately. Asking prices are not completed-sale comparables or an after-repair value. The Deal lab uses your repair, resale, financing and transaction assumptions, including downside tests. Verify title, lawful access, condition and investor eligibility before treating a lead as actionable. Freddie Mac First Look rules ↗

Primary methodology references.

Data Provided by Zillow Group. Dwellventa derives its own screening indicators and is not affiliated with Zillow Group or the listed authorities. The comparable-selection reference informs research design; Dwellventa does not provide an appraisal.

Strategy scores & evidence rules

Flippability · v1.3

Fixed weights: compatible asking-price position 50%, housing momentum 20%, market liquidity 15%, research-input coverage 15% (neutral points only). Start at 50 and add each supported factor’s weighted deviation from 50, scaled by evidence strength. Missing factors stay neutral. A score requires a recent priced home and the compatible-peer thresholds above. Investor exclusions, expired offerings, reported demolition and environmental flags prevent a score.

Rental Potential · v1.2

Fixed weights: compatible acquisition asking prices 30%, 12-month area rent growth 25%, rent-index downside stability 20%, longer-term rent trend 10%, research-input coverage 5% (neutral points only), broad economic and financing pressure 10%. Start at 50 with evidence-scaled contributions. Missing factors remain neutral, without redistributing their weight. The economic contribution is capped at ±1.5 points.

A score requires a priced retail residential offering observed within 45 days, reported size or bedrooms, at least 13 consecutive local rent observations dated within 180 days and 25% weighted evidence coverage. A numeric Rental Potential score additionally requires the same five-peer, three-effective-peer, 35% price-reliability and outlier rules as Flippability. Without qualified acquisition evidence, dated area rent context remains available with no property score. Program-price homes without verified investor terms remain unscored.

Rent evidence & scope

Data Provided by Zillow Group. Derived Zillow Observed Rent Index indicators cover 5,996 ZIP areas, 573 metros and 445 single-family metros in this snapshot. A matching single-family metro is preferred for explicitly reported single-family homes; otherwise a state-matched ZIP or audited metro is used at lower evidence strength. Mixed rental stock does not establish unit-level rent.

Source observations through August 31, 2026; retrieved October 6, 2026. The Pro chart rebases each area index to 100. Rent-index methodology ↗

Underwriting is still required

Neither score estimates rehab costs, closed-sale resale value, monthly property rent, vacancy, taxes, insurance, cap rate or returns. Verify condition, title, occupancy, lawful rental use, leases, actual costs and First Look or program restrictions. Vacant land and nonresidential property show “Not applicable.” Missing or changed evidence is shown explicitly; no score is invented.