Methodology and limitations

What each number is computed from, and the specific way each one can be wrong.

QAPs.net scores sites against a state’s published Qualified Allocation Plan. Every point, deduction and threshold check traces to a section of the rulebook, and the citation links in the app open the agency’s own PDF at that page. That part is exact.

The inputs are not. Several checks run on documented proxies rather than the authoritative source, because the authoritative source is not published as data, is not published at all, or exists only as a determination an agency makes on a specific application. Where a proxy is used, the app labels it on the screen where it appears. This page is the full list.

A proxy result is a screening signal. It is not a finding of vacancy, buildability, ownership, zoning, environmental condition, utility availability, access, or site control.

The proxies in use

Whether a competing LIHTC project sits within two miles

What we actually use: HUD's national LIHTC database, a point-in-time inventory that lags placements in service and does not track an agency's current pipeline.
Where it can be wrong: A project awarded or placed in service since HUD's last publication is invisible, so a site can pass the two-mile check and still fail it at the agency.

Whether a tract is a Qualified Census Tract or Difficult Development Area

What we actually use: The published HUD QCT and DDA designations for the current year, joined by tract.
Where it can be wrong: Designations change annually and a deal is scored under the designation in force at application, not at the time you searched.

Whether a site is rural, which sets the service radius at 5 miles instead of 3

What we actually use: Census urban-area geography, standing in for USDA's own rural determination.
Where it can be wrong: USDA's determination is the one that governs and is made on different boundaries. A site near an urban-area edge can be classified differently by each.

Whether a parcel is vacant and developable

What we actually use: County assessor records showing acreage with zero improvement value or zero buildings. Which field carries the signal varies by county.
Where it can be wrong: This is a tax-record signal, not a site inspection. It says nothing about buildability, wetlands, utilities, access, easements, topography, or whether anything is standing on the land today.

Whether a grocery store, pharmacy, bank, convenience store or medical office is in range

What we actually use: Overture Places, an open point-of-interest dataset built from OpenStreetMap and partner contributions, as a point-in-time extract.
Where it can be wrong: Absence of a record does not establish absence on the ground, and presence does not establish that a business is still open. A closed anchor store scores exactly like an open one.

How far away those services are

What we actually use: Driving distance by road, computed with a self-hosted OSRM router on OpenStreetMap data, from the parcel centroid.
Where it can be wrong: Each agency names its own measure and neither is this one: AHFA measures by vehicle odometer between specific physical points, and ADFA measures the shortest driving distance per Google Maps. Against Google Routes on 54 Arkansas parcel-to-amenity pairs drawn deliberately at scoring boundaries, the median disagreement was 0.02 miles and the 90th percentile 0.44, with an observed worst case of 1.38. Any result within about half a mile of a scoring limit is therefore ambiguous and needs to be re-measured the agency's way, not computed. The centroid is also an assessor-record centroid, not a surveyed corner or a proposed building location.

A census tract's Area of Opportunity Index, worth up to 10 of Arkansas's 99 points

What we actually use: ADFA's own currently published AOI feature layer, read whole and cached, joined by 2020 census tract. The number is the agency's; we look it up rather than deriving it.
Where it can be wrong: ADFA republishes the index between cycles on a new service and leaves the previous one answering, so the value in force is the one published at application, not the one cached here. The values ADFA stores carry one decimal place, where the QAP text describes two and ADFA's own map pads the number out for display; we show the stored value exactly, and neither pad it nor derive a second decimal to match the wording. One Arkansas tract carries no value at all, which is reported as unknown and never as zero.

Whether a site is outside the 100-year floodplain

What we actually use: FEMA's published flood hazard layers.
Where it can be wrong: Maps are amended, and a Letter of Map Amendment or a pending revision will not appear here. Unmapped areas are unmapped, which is not the same as being outside the floodplain.

Whether an incompatible use sits next to the site

What we actually use: A materialized extract of categorised OpenStreetMap features around each parcel.
Where it can be wrong: The deduction list is the QAP's, but detection depends on the feature being mapped and tagged. On the batch-scored candidate set some parcels carry an explicit deductions-not-checked flag, which the card shows rather than hides.

County award history and whether a county may be closed this cycle

What we actually use: AHFA's own published award lists for 2021, 2024, 2025 and 2026, extracted from agency PDFs and its approved-applications feature layer.
Where it can be wrong: It is a recency signal, not an allocation model. It does not know the current cycle's applications and cannot tell you a county is closed.

What the Service never checks

Zoning and site control are never verified. Whether a parcel is zoned for multifamily use, whether it can be rezoned, how long that takes, and whether anyone has it under contract are questions this product does not answer and does not attempt to answer. Rezoning routinely takes months and is the single most common reason a well-scoring site is not a real deal.

What a ranking is, and what it is not

The Deal Sourcer ranking orders sites by the QAP points a site’s own characteristics earn. It is not a prediction of which projects receive an allocation. The agency runs an ordered set of selection lanes, then a tiebreaker cascade that ends in a public drawing. None of that is modelled here.

AHFA states it directly: “Regardless of strict numerical ranking, the Point Scoring System does not operate to vest in an applicant or project any right to a reservation or allocation of Housing Credits in any amount.”

Searched counties are the only counties searched. A county absent from the candidate list has not been examined and found empty; it has not been examined.

Packet tools

Generated packets, manifests and previews are working documents. They are not filing-ready applications and are not a certification of completeness, including when every tracked checklist item is populated: the tracked checklist may not cover every agency requirement. Agency filings also impose requirements this product cannot produce, such as printed index pages on specified paper, original signatures in a specified ink, sealed envelopes, and large format survey sheets.

This page is referenced by the Terms of Service. Independently verify every material fact before acquiring a site, committing capital, or filing.