Methodology

What Gets Published

Public company records must meet a minimum publishable threshold: name, slug, website or canonical domain, description, location or country, classification, provenance or manual review context, and a freshness timestamp. Candidate records that do not meet that bar can remain in review without appearing in the public directory.

Why Some Companies Are Not Shown

“Publishable” is a test each record has to pass on its own, not a judgement about the company. A record is held back when a required field is missing — most often a location or country, sometimes a working website, a description, or the source URL that supports the entry. That means a real, well-known company can be absent simply because the directory has not yet verified enough about it to present a record honestly. Absence here is a statement about our data, never about the company.

The same rule runs in the other direction: nothing is published to fill a gap. If a field cannot be traced to a source it is left blank and labelled as not yet verified, rather than estimated, inferred, or copied from a competitor's listing. Counts shown across the site are counts of records that pass this test, which is why they are described as publishable rather than as the size of the market.

A small number of companies in the current corpus are held back for exactly this reason: they publish no address anywhere, and no business register we can reach carries an entry we can match to their domain with confidence. We would rather show nothing than invent a country for them.

Logos and Trademarks

Company names, logos and other marks shown on deftech.directory are the property of their respective owners. We reproduce them to identify the company each entry describes — nominative use — and for no other purpose.

Nothing here implies that a company endorses, sponsors or is affiliated with this directory, or that we are affiliated with it. We are not a reseller, a partner or an agent of any organisation listed.

Each logo is collected from the company's own website where possible, and every one records how it was obtained: parsed from the company's homepage, retrieved from a logo service keyed on its domain, or taken from a favicon service. Where the mark came from a favicon service we say so, because that image is a third party's best guess rather than something the company published. Logos are resized and re-encoded for consistent display; we do not alter them otherwise.

Listing is editorial. A company does not pay to appear and cannot pay to be removed for commercial reasons — but if you are a rights holder and want your mark taken down, or you would rather we used a different one, write to contact@deftech.directory and we will act on it. The entry itself will remain, without the logo.

News and Job Listings

News items are collected from published RSS and Atom feeds. Only the headline, the publisher's own summary, and a link back to the original article are stored — every story is attributed to its publisher and every link points at the publisher's site, not at a copy hosted here. Categories are assigned by a fixed keyword rule set, so the same story always lands in the same category. Job listings are read directly from each company's applicant tracking board and link to that employer's own application page.

Source Strategy

The pipeline prioritizes high-signal public sources such as defense innovation programs, accelerator cohorts, conference exhibitor lists, venture portfolios, curated landscape maps, and company websites. Controlled open-ended discovery is only appropriate when it remains source-backed, rate-limited, robots-aware, deduplicated, observable, and review-gated.

Classification and Evidence

Defense relevance is treated as a reviewable classification, not a claim of endorsement or comprehensive market coverage. Public profiles should show what the company does, why it appears relevant to defense or dual-use technology, and which source URLs support the record.

Known Limits

The directory is not exhaustive, and freshness depends on source recency, extraction health, and reviewer activity. Missing fields should be interpreted as not yet verified, not as proof that a fact does not exist.