Methodology

What this page covers

Texas Home Intelligence publishes a continuously updated layer of local homeowner data and plain-language interpretation, plus free homeowner tools. This page explains how the data layer works: where each figure comes from, how often it refreshes, and what we do when a source fails.

Where does Texas Home Intelligence get its data?

Every figure comes from a named public source, and every page states what it measures, its geography, its reporting period, and when the feed was last confirmed. We track 18 public sources; 13 are currently connected and live. The table below is generated from our source registry, so it reflects the real state of each feed rather than a description of it.

Public / Observed Data vs. Historical Context vs. THI Analysis

We separate what a source actually reported (Public / Observed Data) from how that compares to prior conditions or seasonal norms (Historical Context) from our own plain-language read of what it might mean for homeowners (THI Analysis). Modeled estimates and anything a homeowner tells us directly are labeled as Estimates & Homeowner Information, never presented as an observed fact.

Stale and failed data

If a source feed fails or returns nothing useful, we do not silently show a zero or blank value as if it were current. We preserve the last known valid figure and mark it as stale, with the date it was last confirmed. If no valid value has ever been obtained, we say so explicitly rather than substituting a fabricated number.

How often does the data update?

Ingestion runs on a schedule and appends to an archive rather than replacing it. Several sources publish on their own lag — NOAA's storm reports run months behind the events they describe, and the EIA revises recent months as utilities report — so each page shows two separate dates: how far the records themselves run, and when we last confirmed the feed. Collapsing those into one "updated" date would overstate how current the data is.

Historical observations

Where practical, we store historical values rather than overwriting the previous number, so a data page can show a trend, not just a snapshot.

Do you ever publish placeholder numbers?

No. Sources we have not connected yet are listed above asNot yet connected and carry no figures at all. Published data pages contain no placeholder rows: a page is only built once its feed has returned real records, and the fabricated rows used to bootstrap a dataset during development are retired from it the first time a real fetch succeeds. That retirement failed once, in August 2026, and we published a figure that had never been measured — see below.

Have you ever published a number that was wrong?

Yes. Twice, and these are the records of them.

28 August – 1 October 2026 — a placeholder electricity price published as a measurement

From 28 August to 1 October 2026, the Texas residential electricity price shown on this site — 13.88¢ per kilowatt-hour, labelled August 2026 and sourced to the U.S. Energy Information Administration — was not a measured figure. It was a placeholder written while the dataset was being set up, which should have been discarded the first time the real feed succeeded and was not. The real series had stopped at 16.44¢ in May 2026, and the placeholder sat on top of it, so the page also showed a LIVE badge over a feed that was four months stale.

It first appeared on 28 August, on the Texas electricity price page and its downloadable CSV, the day that page was published. From there it reached the homepage, both HVAC pages, the AC lifespan tool, and one analysis article. Because the placeholder values ran about 2.5¢ below the real ones, the published series read as a sharp recent fall in prices. It was not one — the drop a reader saw was the 2.5¢ gap between the placeholder and the real series, not a movement in the price, and nothing in this record is a claim about how the price has moved since. Texas residential electricity prices did not fall as the published series suggested. Over the ten months the real data covered at the time of this correction, they rose 6.3%, from 15.46¢ in August 2025 to 16.44¢ in May 2026. The current series is on the Texas electricity prices page.

What we have done about it:

• The placeholder rows no longer publish a figure — seven removed and three relabelled as placeholders, across seven datasets. Every page that carried the figure was rebuilt from the real series, and the electricity feed's badge is resolved from the age of the data rather than from whether the fetch succeeded, which at the time of this correction made it Out of date.
• The article built on the figure has been withdrawn, not corrected. Its central claim could not be restated: the month it reported on does not exist in the data, and the ten-month series it had could not support the year-over-year comparison it led with. Its URL returns 410 Gone.
• The detection is now a test that runs on every build. The old check tried to recognise a placeholder by what it looked like, and these rows did not look like one. The new one regenerates what the placeholder generator would have written and matches it exactly, so a placeholder cannot reach a published page without failing the build. The test also proves itself each run, by re-running the old check and confirming it still misses rows the new one catches.

September – October 2026 — San Antonio permit counts counted records, not permits

Until 1 October 2026, the San Antonio permit figures on this site counted rows in the city's file rather than permits. The city lists one permit on several rows when it covers more than one type or stage, so the counts were too high. The figures were real and sourced; the arithmetic over them was wrong.

The page it mattered on is San Antonio plumbing, which from 4 September to 1 October 2026 said the city issued about 2,040plumbing permits a month and 26,515 over the thirteen months to September 2026. The correct figures are 1,891 a month and 24,587 — about 7% lower. The same page described activity as flat, and still does: it read as a 0.5% rise between the two halves of the window and reads as a 1.6% rise now. No claim on any published page changed. The error could have changed one — in electrical, a trade we publish no page for, counting permits instead of rows turns a 4% fall into no trend at all — but on the pages a reader can open, it did not.

San Antonio's HVAC and re-roof pages moved by about 1% in the other direction. Austin's counting was never wrong — its records are one row per permit — but its figures moved by under 1% too, from a second fix shipped alongside this one: the whole window is now recounted on every run, so permits the city adds to a month we have already counted are no longer missed.

We are recording this rather than quietly correcting it for one reason in particular.The city's file cannot be downloaded by an ordinary script — it refuses the request — so nobody outside could have checked these numbers against the source. A site that asks to be trusted on its sourcing does not get to fix a figure nobody could have verified without saying so.

What has changed: both cities' permit counts now read the permit number and count each permit once; the method note on every permit page says so; a test fails the build if the counting reverts; and the whole thirteen-month window is recomputed on every run, so no month can be left on an older method.

How is the Home Stress Index calculated?

The Home Stress Index combines five signals — Roof & Storm, Foundation & Soil, HVAC, Water & Irrigation, and Trees & Yard — into a single 0–100 reading for an area. Its weights, bands, storm-decay curve, sources, and the inputs we deliberately leave out are all published on its own page:the Home Stress Index methodology.

Not a substitute for professional evaluation

Nothing on Texas Home Intelligence is an inspection, a diagnosis, or a substitute for a licensed professional’s on-site evaluation. Our readings describe conditions recorded across an area; they cannot confirm the state of any individual property.