Adjudication Atlas
32,192 clearance decisions to search, compare and map.
Published DOHA security-clearance decisions from 1996–2026. Look one up by case number, narrow by guideline, outcome and year, compare up to four side by side, and export citations that link to the official decisions. Or see them all on one map. It runs in your browser, with no account, no server and nothing to install.
Clustered by guidelineOpen the map →
Read these as history, not odds. They cover published DOHA decisions only, not every clearance adjudication, and describe past decisions. They do not predict the outcome of any individual case.
Four ways in
One set of filters (guideline, outcome, year, decision type and more) drives every view, and each view keeps the others' search and filters.
Search decisions
Find a case by number, or decisions by guideline or cited condition. Read each result's date, guidelines and outcome, compare up to four, and export CSV or citations.
Search decisions →Map
A WebGL constellation: one dot per decision, one cluster per guideline. Color by guideline or outcome, split each cluster into granted vs. unfavorable, and trace hearing → appeal reversals.
Explore the map →Table
Every visible decision in one sortable list, by date, outcome, decision type or case link, with a link to the decision in DOHA's public repository.
Open the table →Insights
Did SEAD 4 change outcomes? What happens on appeal? Which guideline pairs and mitigating conditions go with grants? Every chart has 95% intervals and a data table.
View insights →Search decisions
Search case numbers, guidelines, and available metadata. This is not a full-text search: the Atlas holds no decision text. Every result links to the official decision on DOHA's site. To search the words of the decisions themselves, use Search decision text.
Browse by guideline
What the data shows
Descriptive findings from the current release. These are patterns in published decisions, not predictions. Rates count initial decisions only; appeals have their own line.
Financial Considerations (Guideline F) initial decisions were granted 36% of the time before SEAD 4 and 27% after. The 95% intervals (35%–37% vs. 26%–28%) do not overlap.
Appeals: of the 4,125 Appeal Board decisions linked to an initial decision in this data, 85% kept its outcome, 10% reversed it and 6% sent the case back (remanded).
Guideline spread: among guidelines with at least 100 initial decisions, grant rates run from 19% (Personal Conduct) to 58% (Foreign Influence).
Read it carefully
It covers published DOHA decisions only, not every adjudication. Appeals are self-selected. Guideline, date and condition fields are extracted from decision text by documented rules and can be wrong. Nothing controls for the facts of individual cases.
Rates are granted ÷ (granted + denied + revoked) among initial decisions; remanded cases have no result and are left out. Rates based on fewer than 20 known outcomes are never shown. Not legal advice.
Coverage and limits
What is in the data, what is missing, and how old it is.
| What it holds | 32,192 published DOHA decisions from 1996–2026, held in a governed DCSA evidence library: 27,282 initial decisions by Administrative Judges (ISCR and ADP) and 4,910 Appeal Board decisions. Not every DOHA decision, and not every clearance decision: most clearance determinations never reach a DOHA Judge. |
|---|---|
| Uneven years | Fewer than 50 initial decisions for 1996 and 1999–2000; Appeal Board decisions come from across 1996–2026. A thin year says more about which decisions DOHA has published and the library holds than about DOHA's caseload. |
| Missing fields | 99.7% of decisions have an exact date; the rest show "Not available", and their year is the case-number year. 100.0% link to the exact DOHA decision, using the official address the library records for it; a decision without one would link to DOHA's list for its year, or show no link. The program (ISCR or ADP) is not recorded for Appeal Board decisions. Values the data lacks are shown as "Not available", never guessed. |
| Outcomes | Initial decisions are granted, denied or revoked. Appeal Board decisions affirm, reverse or remand (send back) a Judge's decision. Counts are of decisions, not people: one case can have an initial decision, an appeal and a decision on remand, so initial and appeal figures are always kept apart. |
| Case history | Decisions with the same case number are listed together. Which initial decision an appeal reviewed is inferred from the case number and dates, not read from the appeal, so the Atlas calls those links "linked", not verified. |
| Dates | Data last converted from the library on 2026-11-04; the latest decision in it is dated 2026-10-28. Each link is the official DOHA address the library records for the decision. The site build date at the bottom of each page says when the pages were assembled, not when the data changed. |
Connect your AI assistant
The Atlas is also a Model Context Protocol (MCP) server, so Claude and other MCP clients can search and count these decisions for you, with the same metadata and the same official links. Six tools over the same 32,192 decisions: five answer from metadata, and one searches the decisions' words and returns short passages, as the text search does. Every FSOtools connector, with its status, is listed on the connectors page.
| Server address | https://fsotools.ai/atlas/mcp (Streamable HTTP, no sign-in, no key). Opened in a browser, the address shows a guide to the tools. |
|---|---|
| Claude.ai | Settings → Connectors → Add custom connector. Name it Adjudication Atlas, paste the server address and save. Then ask, for example, "How did Guideline F decisions come out after SEAD 4?" |
| Claude Code | claude mcp add --transport http doha-atlas https://fsotools.ai/atlas/mcp |
| What it is | A read-only endpoint over the Atlas dataset, served on this domain by the doha-atlas-mcp Worker. Five tools answer from metadata; the sixth returns short passages from published decisions, the same ones the text search shows. It adds nothing to those decisions: no names, no notes, no predictions. Every result links to the official decision on DOHA's site. |
How it's built
Transparent sources and rules, so every number can be traced back.
| Source | Published DOHA decisions held in a governed DCSA evidence library, converted read-only into structured fields. No decision text, summaries or applicant details are copied. |
|---|---|
| Guidelines | The guidelines found by scanning each decision. The primary guideline comes from the decision's Formal Findings (the first guideline found against the applicant; for grants, the first listed). |
| SEAD 4 era | By decision date: on or after 8 June 2017 is post-SEAD 4. |
| Conditions | SEAD 4 disqualifying and mitigating conditions the decision cites (e.g. AG ¶ 20(b)). Judges also cite conditions they find do not apply. |
| Links | Only addresses recorded when the decisions were collected from DOHA's public repository: the exact decision, or DOHA's listing of that year's decisions. Links are never guessed. |
| Search and exports | Search reads case numbers, guidelines and cited conditions. CSV and citation exports carry the same metadata, the export date and the dataset version; they hold no decision text, and nothing is sent anywhere: files are made in your browser. |
| Engineering | A single self-contained page (WebGL via regl-scatterplot), no server and no runtime network calls, tested in CI on every change. |
Corrections
Spotted a wrong date, outcome, guideline or link? Email a correction and include the decision id (for example 14-05302.a1). The same address is behind every page's Report a correction link. The official decision on DOHA's site, not this dataset, is the record.