How a GroundWise Soil & Foundation Activity Report is built — what data goes into it, at what geographic scale that data is measured, and what the data does and does not support.
Section 1
A GroundWise Soil & Foundation Activity Report is a data report describing soil and foundation-repair-activity conditions in the mapped area containing an address. It is not a home inspection, a geotechnical investigation, or an engineering opinion.
The report is built for home buyers, owners, and real estate professionals in the expansive-clay geologies of Central Texas, beginning with the Austin and San Antonio metropolitan areas. It assembles publicly available federal soil-survey data and a set of third-party certified foundation repair project records, and presents them at the geographic scale each dataset actually supports.
The appropriate description of what it delivers is a neighborhood-scale reading of soil conditions and local repair activity that identifies locations sitting in more challenging geologies and provides a factual starting point for due diligence. It is not a lot-level engineering assessment, and it is not designed or validated to be one.
Section 2
The report presents two independent readings. They are displayed next to each other and are never combined, weighted, or averaged into a single number.
| Signal | What it describes | Geographic scale |
|---|---|---|
| Soil Condition Tier | Two properties of the mapped soil, combined: its shrink-swell potential — the property most consistently associated with slab movement in expansive clay, and the dominant term — together with its restrictive-layer depth, carried at a smaller weight. Section 5 sets out both, and why depth points the way it does | USDA soil map unit |
| Neighborhood Foundation Repair Activity | How much foundation-repair activity is recorded across the surrounding area, ranked against other census tracts in the metro | Census tract |
Three reasons, each independently sufficient:
Section 3
An address is geocoded to a coordinate; that coordinate is used to identify which soil map-unit polygon and which census tract it falls within; the mapped properties of those areas are what the report describes. The report does not measure anything at the property, and no soil sample, boring, elevation survey, or structural observation is involved at any point.
This matters practically. Two neighbouring homes on the same soil map unit will receive the same soil reading, because the underlying survey does not distinguish between them. A report that scored one materially differently from the other would be asserting something the data cannot support.
Section 4
| Input | Source | Scale |
|---|---|---|
| Shrink-swell potential (Linear Extensibility Percent) | USDA gSSURGO — NRCS county soil survey | Soil map unit, dominant component |
| Restrictive layer depth | USDA gSSURGO corestrictions table | Soil map unit, dominant component |
| Soil percentile ranking | Distribution of mapped soils across the metro area | Metro-wide reference distribution |
| Neighborhood Foundation Repair Activity | Third-party certified foundation repair project records | Census tract |
| Foundation-repair permit activity | Municipal building-permit records, where the jurisdiction publishes them | Census tract |
| Foundation type (slab / pier & beam) | County appraisal district records where a foundation field exists; otherwise estimated from building age and local construction patterns | Property record |
| Topography and tree proximity | USGS 3DEP elevation and LiDAR | Property-level — shown as context only, not scored |
All soil inputs are drawn from published federal survey data. Repair-activity records are certified by a licensed professional engineer. Where a report shows a foundation type that was estimated rather than read from a county record, the report labels it as estimated.
Section 5
Two soil measurements are used, both taken from the mapped soil unit at the location:
Linear Extensibility Percent measures how much a column of soil changes in length between its dry and moist states. It directly quantifies the shrink-swell behaviour of the clay fraction, and it is the dominant term in every version of the model that has been tested. It is the single soil property most consistently associated with slab-on-grade movement in this region.
This is the depth at which a physical restriction to moisture movement begins — typically caliche, bedrock, or a compacted hardpan. Its direction matters and is easy to get backwards: a shallower restriction contributes a lower reading, not a higher one. A shallow restriction leaves only a thin layer of expansive clay above it, so there is less material available to move seasonally. A deeper, unrestricted profile holds a thicker column of active clay and moves more.
The survey does not record a restriction depth for every map unit — across the Austin and San Antonio metros, roughly half the mapped land area has none recorded. The report distinguishes three cases rather than treating them alike, and says on the page which one applies:
| Case | What the report shows | Label |
|---|---|---|
| A depth is recorded for the map unit | That measured depth | no label needed |
| No depth recorded, and the soil is a series confirmed to be deep and unrestricted | Treated as a deep active profile — for these Blackland Prairie clays the absence of a restriction is a real characteristic of the soil, not missing data | Deep Profile |
| No depth recorded, any other series | An average across the surrounding census tract, which is less specific to the location than the shrink-swell value | Tract Estimate |
Shrink-swell is never substituted from another scale. Where a map unit has no recorded shrink-swell value — most often "Urban land" units in dense areas, where the soil profile was cut or filled during construction — the report is not issued for that address rather than borrowing a figure from a wider area.
Both measurements are converted to percentile ranks within the metro reference distribution before being combined, so that the resulting reading reflects relative standing rather than raw values. The two are then combined into a single soil percentile, with shrink-swell carrying the substantially larger share. The weighting was fitted against recorded repair activity using cross-validation rather than being chosen by judgement, and it has been re-tested at map-unit scale.
Only one percentile is shown for soil. The individual soil inputs are displayed as measured values with their USDA classification, without separate percentile bars, so that the tier is not misread as an average of two visible numbers.
The comparison population is the mapped soils of the metro, not its census tracts and not its homes. Each map unit is weighted by the amount of housing it underlies, so the comparison reflects soils in the areas where homes are actually built rather than total land area — without that weighting, large tracts of sparsely-built ranchland would dominate the distribution and pull every developed location upward.
The four tiers are quartiles of that distribution, so each covers a quarter of it by construction:
Tiers are metro-relative: they describe where a location's soil sits against other mapped soils in the same metro, not an absolute hazard classification. Because the distribution of soils differs between metros, bands are calibrated separately for each market. The report also shows the USDA absolute shrink-swell classification alongside the tier, so both readings are available — the USDA class answers "how expansive is this soil in absolute terms," the tier answers "how does it compare locally."
Section 5b
Because the soil reading comes from the mapped unit containing the property, a property near the edge of that unit sits close to where the adjoining unit begins. Where that applies, the report says so and names the tier the adjoining unit would indicate.
The threshold is not a round number chosen for convenience. It is the National Map Accuracy Standards horizontal tolerance at each survey's own published scale, read from the USDA survey metadata. The surveys covering these metros were published at three different scales, so the band differs by county:
Williamson, Hays, Comal, Caldwell and Guadalupe fall in the first band; Travis, Bexar, Bastrop, Medina, Wilson, Atascosa and Bandera in the second; Kendall in the third.
Only when the adjoining mapped unit would indicate a different tier. Adjoining units frequently share the same soil series and the same shrink-swell value, in which case there is no difference to disclose and a warning would be noise. In practice the note appears on roughly 7–8% of reports.
Section 6
This is a census-tract-level reading derived from third-party certified foundation repair project records. It is reported as the tract’s rank among census tracts in the same metro — a percentile and an activity tier.
It describes the surrounding census tract as a whole. It does not indicate anything about the condition of any individual property, and it does not mean that a specific home has or has not had foundation work. A tract is a neighbourhood-sized area containing thousands of homes.
Where a tract has fewer than five underlying records, no figure is shown at all. The report says "not enough data" instead. A rate computed from one to four records is not a reading in either direction, and presenting it as one would be misleading. This suppression affects a meaningful share of tracts — roughly a third in the Austin market and around fifteen percent in San Antonio.
From v2.6 this floor applies to every metric built on a count of records, including the pier-and-beam-specific reading described in Section 7. That reading previously showed a tier from as little as a single observed repair. Of the tracts affected by raising it, most carried an elevated rather than a reassuring reading — so the change removes alarming figures drawn from one or two observations, not comforting ones.
Counts themselves are never published. The number of records behind a reading does not appear on a report in any form. It governs suppression only. Counts of municipal building permits, which are public record, are stated where they are used.
Where a census tract sits only partly inside the city that publishes the permit records being used, the report carries a coverage caveat. In that situation the records cover only part of the tract while the comparison population covers all of it, so the activity reading understates the true level and should not be read as genuinely low activity.
The activity reading is a comparison between census tracts, not a count of repairs. It is not a number of jobs, and it is not a repair rate. Earlier versions of the report also showed a multiplier against the metro median; that figure compared scores on a compressed scale rather than repair rates, so it was removed in favour of the percentile, which needs no such qualification.
Section 7
Foundation type changes which signals are meaningful, so the report routes to a different format depending on it.
Both signals apply. The report shows a soil condition tier and a neighborhood repair activity reading side by side.
Pier-and-beam reports do not carry a soil condition tier. When the soil signals were tested specifically against pier-and-beam repair activity, they were roughly six to eight times weaker than for slab-on-grade construction — at or below the threshold of statistical significance. This is consistent with a crawlspace structurally separating the home from direct contact with the soil. Reporting a soil tier on a pier-and-beam property would present a reading the data does not support, so the report omits it and explains why.
For pier-and-beam properties the primary signal is instead foundation-repair activity built from municipal building-permit records for the surrounding tract, which reflects market-wide activity rather than any single source of records. Where a jurisdiction does not publish suitable permit data, an adjusted pier-and-beam-specific reading is shown in its place and labelled as adjusted.
Where foundation type cannot be determined with confidence from available records, the report does not guess. It routes to a review format that states the type could not be resolved.
Reports prepared for a contractor to attach to a bid include a figure the consumer report does not: the median depth at which installed pilings reached refusal in documented slab repairs across the surrounding census tract, reported separately for each pressed system — steel, hybrid and concrete.
The three systems are not comparable to one another. Each reaches refusal at a different depth by design, so the figures describe the systems rather than ranking them, and the report says so above the numbers. The ordering is not even consistent between tracts: in some areas concrete records a greater median depth than hybrid, in others the reverse. Any claim that one system consistently goes deeper than another is unsupported.
Each system needs five recorded repairs before a depth is shown. Where a system is thin in the subject tract but has history nearby, a pooled median may be drawn from the nearest tracts carrying recorded work of that system; such a figure is labelled on the report as describing the surrounding area, and states how many tracts were pooled. Where neither the tract nor the surrounding area supplies enough history, the row says so and no depth is shown.
This is observed history for repairs that were carried out. It is not a projection of what any particular property will require, and it is not a quotation.
Section 8
These boundaries define how a GroundWise report may legitimately be read.
Section 9
A GroundWise report contains no observation of any structure. In particular it does not assess:
These are matters an on-site licensed professional can evaluate and this report cannot. Any structural question should go to a licensed home inspector or professional engineer.
Section 10
A signal is only included if it measurably improves the reading under cross-validation against recorded repair activity. Several plausible candidates were tested and did not clear that bar, and are therefore not part of the scored model:
One further candidate, stream-network density, does clear the threshold in testing but has not been incorporated into the scored model. It is a candidate for a future methodology version. Nothing enters the report on the strength of plausibility alone.
Section 11
Source datasets are periodically revised, and site conditions change over time. A report reflects the data available on its generation date and is not updated afterwards. A generation date and a report identifier appear on every report; quote the identifier in any correspondence about a report.
Reports state the methodology version they were produced under. This page describes v2.6. Where a methodology change would materially affect how a reading should be interpreted, the change is recorded in the version history rather than applied silently.
If you believe a report contains an error, contact support@groundwisedata.com with the report identifier.