Whitespace studies

Most empty space is empty for a reason.

Finding a gap in a patent landscape is easy — plot anything and holes appear. The hard part is telling a real opening apart from a thin corpus, a vocabulary mismatch, or an idea that simply does not work. So every gap is attacked before it is reported.

Hypothesis strength

multiplicative
Density82
Rarity74
Semantic novelty68
Evidence quality55
Crowdedness71

Strength is the product of its pillars, not their average. One collapsed pillar collapses the score — a beautifully rare gap with no evidence behind it does not get to rank highly.

Validation record

6 attacks planned
Synonym shifted0 hitsCLEAN
Semantic paraphrase2 hitsWEAKENING
CPC adjacent0 hitsCLEAN
Assignee pivot1 hitWEAKENING
Red team0 hitsCLEAN
LiteratureNOT_RUN

Five attacks ran, one could not — and the reason is recorded. A study that quietly skips its own tests would report a stronger result than it earned.

Pipeline

5 stages, audited end to end

Gap taxonomy

10 reasons a space is empty

Validation

6 attacks, 6 diagnostic gates

Handoff

Straight into novelty search

The pipeline

Five stages, and the last one exists to destroy the first four.

Discovery is the cheap half. The validate stage is adversarial by design: its job is to refute what the earlier stages proposed, and a hypothesis is only interesting once it has been through it.

  1. 1

    Frame the field

    FIELD_MAP

    The scope is written down first — concepts, classifications, explicit exclusions, and the assumptions being made — so the study has a definition you can argue with before any compute is spent. Every edit to it is recorded on an audit trail.

  2. 2

    Cluster the corpus

    CLUSTER

    Documents from 2000 onward are embedded and split into clusters, and each cluster is graded well-defined, usable, or diffuse. A diffuse cluster is reported as diffuse rather than dressed up as a finding.

  3. 3

    Measure density, rarity, and divergence

    SIGNALS

    Filing density over time, term divergence between clusters, claim-element families, and genuinely rare feature pairs. This is where an absence starts to become measurable instead of anecdotal.

  4. 4

    Read the rare region closely

    DEEP_DIVE

    The sparse pairs are examined against the documents that surround them, to establish what is actually claimed nearby and what the gap is bounded by.

  5. 5

    Attack the hypothesis

    VALIDATE

    Each candidate opening is then actively attacked — six adversarial search strategies and six diagnostic gates. A hypothesis that survives has been shot at; one that does not is recorded as refuted, with the query that refuted it.

Six gates

Before you call it an opportunity, rule out the boring explanations.

Each gate is a specific reason a region of a landscape can look empty without being available. They return passed, passed with weakening, failed, advisory, or unassessed — and unassessed is reported honestly rather than treated as a pass.

G1 · Data

Is the corpus even dense enough here to support a conclusion?

G2 · Terminology

Does this exist under another name?

G3 · Adjacent claims

Is a neighbouring claim already reading on it?

G4 · Feasibility

Is it empty because it cannot be built yet?

G5 · Commercial

Is it empty because there is no market?

G6 · Regulatory

Is it empty because regulation forecloses it?

Attack strategies

Six ways to try to prove your own finding wrong.

Each attack is a real search with a recorded query and hit count, and each returns clean, weakening, or refuting. A refuting result means the full combination already exists — better to learn that here than in an examination report.

Synonym shifted

SYNONYM_SHIFTED

The same idea in other words.

Semantic paraphrase

SEMANTIC_PARAPHRASE

The same idea, restructured.

CPC adjacent

CPC_ADJACENT

The neighbouring classifications.

Assignee pivot

ASSIGNEE_PIVOT

What the obvious filers already hold.

Red team

RED_TEAM

A deliberate attempt to refute it.

Literature

LITERATURE

Non-patent publications.

Gap taxonomy

Ten reasons a space is empty. Only one of them is good news.

A study does not return 'whitespace found'. It returns which kind of emptiness it found, which is the difference between a filing opportunity and a warning that your corpus is too thin to say anything.

Genuine

GENUINE

Survived the attacks and the gates. A real opening that appears both unclaimed and worth claiming.

Patent whitespace

PATENT_WHITESPACE

Nothing in the patent record — but the literature or the market may already know about it.

Claim whitespace

CLAIM_WHITESPACE

Present in disclosures but never claimed. Often the most useful kind, and the easiest to miss.

Data whitespace

DATA_WHITESPACE

The gap is in the corpus, not the world. Coverage is too thin here to conclude anything — and the study says so.

Terminology whitespace

TERMINOLOGY_WHITESPACE

The idea exists under different words. An artefact of vocabulary, caught by paraphrase and synonym attacks.

Scientific whitespace

SCIENTIFIC_WHITESPACE

Published in the literature but not carried into patents.

Product whitespace

PRODUCT_WHITESPACE

Shipping in products without patent coverage.

Technical feasibility

TECHNICAL_FEASIBILITY_WHITESPACE

Empty because it does not work yet. The gate that most often explains a suspiciously clean result.

Commercial whitespace

COMMERCIAL_WHITESPACE

Technically open, commercially uninteresting. Nobody filed because nobody would buy it.

Regulatory whitespace

REGULATORY_WHITESPACE

Blocked or shaped by regulation rather than by the state of the art.

What happens next

A surviving hypothesis becomes an invention you can search and draft.

Converting a hypothesis rewrites it into the same feature shape the novelty search and drafting flows already consume — with no new model call and no new claims invented at the boundary. The study's own audit trail travels with it: scope, runs, edits, hypotheses, challenges, and notes.

Point it at a field you already know well.

Run a study on a technology area you have opinions about. The gates and the attack log will tell you quickly whether the result is worth anything — including when the honest answer is that the corpus cannot support one.

Trial requests are reviewed by a person, usually within one business day. No card, no auto-renewal.