Methodology

Data sources

Rentu blends two market-rate sources and surfaces a third as a reference benchmark.

• Rentcast (active listings)

  IQR median of current rental listings in the ZIP, filtered by bedroom count and property type. De-duplicated by building so a single large complex cannot skew the result. Weight: 60–85%, depending on comp count.

• Zillow ZORI (Observed Rent Index)

  Monthly asking-rent index for ~8,200 US ZIP codes. Represents current market-rate listings across all unit types. Updated monthly. Weight: 15–40%, depending on comp count.

HUD SAFMR — reference only

HUD Small Area Fair Market Rent is shown alongside the estimate as a benchmark but is not blended into it. HUD targets approximately the 40th percentile of gross rents paid by recent movers — it is an affordability floor, not a market-rate estimate. It is used internally to identify likely income-restricted listings in the Rentcast pool.

Zillow ZHVI and US Census ACS data were evaluated and removed. ZHVI adds no independent signal when current listing data is available. ACS lags 2–3 years and adds noise rather than accuracy.

Estimate calculation

Weighted average

The most likely estimate is a weighted average of Rentcast and ZORI. Rentcast's weight increases with the number of matching comps — thinner samples get a lower weight as a precaution against a small pool dominating the result.

These weights are a cautious policy informed by a 40-ZIP backtest, not a data-fitted formula — the backtest's ground truth was itself Rentcast-derived, which limits how far those numbers can be trusted. Weights will be revisited as real usage data accumulates. If only one source has data for the ZIP, that source is used at full weight.

Consensus-adaptive weighting

When sources disagree by more than 25% of the mean, the outlying source has its weight halved before averaging. This prevents a single extreme value from dominating when markets are mixed. With two sources that strongly disagree, this naturally produces a near 50/50 split.

Time-decay weighting

Fresher Rentcast listings carry more weight than stale ones. Each building's weight follows an exponential decay: a listing with a median age of 30 days gets weight 0.5, 60 days gets 0.25, and so on, with a floor of 0.10 so even old comps contribute something. Listings with no age data are treated as fresh. Decay is applied after all filtering and building deduplication, before the weighted IQR median is computed.

Subsidized-listing filter

Before computing the median, Rentcast comps are screened against the HUD SAFMR for that ZIP and bedroom count. Two tiers apply: listings priced more than 25% below the SAFMR are removed unconditionally; listings priced 10–25% below the SAFMR are also removed if they have been on the market for more than 180 days (borderline price combined with extended time on market is a strong signal of income-restricted housing). If fewer than 3 market-rate comps remain after filtering, the full pool is used and a note is shown on the results page.

Two-cluster markets

Some ZIPs have two genuinely distinct rent bands — often older or affordable stock sitting alongside newer market-rate units. When the comp pool clearly separates into two clusters with a meaningful price gap between them, the report shows both cluster medians near the comps table so you can see what the estimate is averaging across. The estimate itself is not changed; the note exists so you can interpret the range with that context. This flag fires only when the gap is unambiguous — thin or smoothly-spread distributions are not flagged.

Range

The range (market low to market high) is anchored to the min and max of the active source values, floored at 92% and ceilinged at 108% of the most likely estimate. When sources diverge by more than 25% — typically in mixed markets where older and newer stock coexist — the high end is extended by the full spread. A note is shown when this applies.

Out-of-range estimates

When a condition adjustment pushes the estimate above the market high or below the market low, the page explicitly states this. The range marker is pinned to the edge of the bar and the caveat line reads "above/below the typical market range for this ZIP" so the number is never silently misrepresented.

Confidence

Reflects both how many sources returned data and how closely they agree. "Sources agree" means both sources are within 15% of each other. "Sources mostly agree" means 15–25% apart. "Sources vary" means more than 25% apart, or only one source is available. The 25% boundary is the same one used for the wide-spread range caveat — a report showing that caveat will always also show "Sources vary."

Adjustments

Adjustments are industry conventions drawn from appraisal literature. They are fixed multipliers — not calibrated to local market data — because none of our sources provide condition or property-type breakdowns at the ZIP level.

Condition

Needs work −12% · Average ±0% · Renovated +10%

Property type

House +8% · Condo 0% (baseline) · Townhouse +3% · Apartment −3%

Square footage

Square footage does not change the estimate. It is used only to display an implied $/sqft rate. None of our sources provide sqft-level data at the ZIP scale, so no calibrated adjustment is possible.

Known limitations

• • ZORI is an all-bedroom average. It underrepresents larger units since studio and 1BR inventory dominates most ZIPs. For 3BR+ units, ZORI may run below the actual market median.
• • Rural ZIP codes may have no active Rentcast listings. If only ZORI data is available, confidence shows as "Sources vary" and the range is wider.
• • Condition and property-type adjustments are fixed multipliers. A renovated apartment in San Francisco and one in rural Ohio both get +10%, which may not reflect actual market premiums in each location.
• • HUD FMR targets the 40th percentile of recently moved renters by design. In high-demand markets, actual median rents run above HUD — this is expected, not a data error.
• • Flat condition adjustments can push estimates outside the market range. When this happens, the results page labels it explicitly.
• • Results are estimates, not appraisals. Use for informational purposes only.

Data freshness