Methodology

Reading methodology v2 · last revised 22 August 2026

A reading is one word — HOLD, WATCH or ACT — about a ZIP code's housing market. This page says where the numbers come from, what they measure, how they are scored, and how often they change. Every danger line we use is published below; none of them is proprietary.

Readings are live. Since 20 August 2026 readings are published for 5,002 of the 22,874 U.S. ZIP codes that have a page here. Coverage grows ZIP by ZIP; a page whose reading is not live yet says so instead of showing a rating. This page describes the method behind every live reading — what came before the current basis, and when it changed, is disclosed in section 7.

1. Where the data comes from

Readings are computed from RentCast, a licensed market-statistics API. We pay for it, and we use its ZIP-level statistics about homes currently listed for sale. This is the only source that feeds a reading.

Three further sources support the product without feeding any reading:

SourceWhat it is used forFeeds a reading?
Federal Housing Finance Agency house price index (U.S. Government, public domain)Testing our danger lines against what actually happened the following year, and as an independent price benchmarkNo
U.S. Census Bureau ZCTA and CBSA files (public domain)Mapping ZIP codes to metropolitan areas, and place namesNo
Freddie Mac Primary Mortgage Market SurveyThe mortgage-rate context shown on paid reportsNo

No source sponsors, endorses, or partners with this site, and no source has any influence over a reading.

2. What the numbers measure — and what changed

Our readings are computed from active-listing statistics: the prices sellers are asking, and how the pool of homes for sale is moving. These differ from sold-price statistics. Asking prices run higher than sale prices, and time-on-market measured across unsold listings runs longer than time-to-contract measured across homes that sold.

We use listing data because it leads the market rather than trailing it: price cuts and lengthening listing times show up here weeks before they show up in closed sales. A reading is a direction, not an appraisal.

This is a change. Until August 2026 our readings were computed largely from closed sales. A number on a page today is a different kind of number than one from before that date, even when the reading word is the same. We hold the reading comparable over time; we cannot hold the underlying measurement identical, because it is not.

3. The formula, and what it lost

The reading weighs three signals. It used to weigh five. Two have no equivalent in active-listing data and were not replaced:

SignalStatus
Months of supplyGone. It needs the number of homes that sold, and active-listing statistics cannot see closings. This was the highest-weighted check in the old formula.
Share of listings cutting priceGone. No equivalent field.
Year-over-year price trendSurvives, now measured on asking prices
How long listed homes have been on the marketSurvives, now measured across active listings
How the pool of homes for sale is changingSurvives

Which three survived is not luck: every survivor was already a year-over-year ratio, and both casualties were levels. A level does not port across a change of basis — an active-listing median sits higher than a sold median — but the direction and size of a year-over-year change stays comparable.

We would rather say this plainly than not: a three-signal reading is less informed than a five-signal one. With two of five danger signals genuinely gone, the engine has less evidence, and it calls fewer ACTs than the old one did. We left that drop in place. Tuning the thresholds until the old distribution reappeared would manufacture confidence the data no longer supports.

4. The danger lines

Each signal has one published line. Crossing a line scores points; the points decide the word.

SignalDanger linePoints
Asking prices vs. a year agobelow −2%2
Asking prices vs. a year ago (steeper)below −5%3
Time on market vs. a year agoup more than +10%1
Homes for sale vs. a year agoup more than +30%1

ACT at 3 points or more · WATCH at 1 or more · HOLD at zero. Because the steepest price line alone scores 3 points — the ACT threshold — a single signal can reach ACT when prices are falling fast. ACT does not mean several lines are crossed.

A market with no lines crossed and all three strength conditions — asking prices up +5% or more, time on market down −20% or more, homes for sale down −15% or more — reads as a strong seller's market and is shown as STRONG, qualified on the page as a seller's market rather than a warning: it is a reason to consider selling, not a danger sign. So the published vocabulary is four words — HOLD, WATCH, ACT and STRONG. Internally the engine files STRONG under the same action class as ACT, because both mean “this is a moment to do something”; we do not show it that way, because an ACT badge on a seller's market would read as a warning and mean the opposite of what it is.

If fewer than 2 of the three signals can be computed for a ZIP, we publish no reading at all: the page says the data is insufficient rather than showing a rating we do not have the evidence for.

The formula is versioned. This is v2. The lines above were calibrated against real active-listing data in August 2026, not carried over: two of them fired almost never when the old numbers were applied to the new basis, so they were re-fitted to flag a similar share of markets. Changes to a line are versioned and restated, not applied silently.

5. How often a reading refreshes

Not every ZIP refreshes at the same rate, and we would rather say so than imply a uniform freshness we do not have. Market data is licensed per ZIP code and costs money to refresh, so the busiest ZIP codes will refresh most often and the rest less often.

The exact cadence is not settled yet. We have not yet run a second cycle on the new source, so we are not going to publish a schedule we cannot yet keep. The plan is monthly for roughly the busiest thousand ZIP codes and less often — quarterly or twice a year — for the rest. When the cadence is set, this section will state it exactly.

Every page carrying a reading shows the month that reading’s own data runs through — today that is August 2026 on all of them. A page with no reading shows no date, because there is no reading date to show.

6. What a reading is not

A reading is general information about a housing market. It is not financial, investment, legal, or real-estate advice, it is not an appraisal or valuation of any property, and it should not be the basis of a decision to sell. It describes a market, not your home, and it knows nothing about your property's condition, your finances, or your circumstances. Consult a licensed real-estate professional before making sale decisions. The full Terms of Service govern.

7. What changed, and when

Readings published before August 2026 were computed from a prior data vendor’s sold-home statistics. That source was discontinued on 14 August 2026 and every reading since has been rebuilt on licensed active-listing statistics, which measure the market differently — see section 2. Figures derived from the earlier basis are no longer published.

8. Corrections

If a figure here disagrees with what the site shows, the site is wrong and we want to know: press@shouldisellyet.com. Our danger lines are checked against the engine automatically on every build, so a discrepancy is a bug rather than a difference of opinion.