Glossary

Terms, defined

Plain definitions of the terms used across this site and in the research and advisory market: verified primary research, research provenance, signed recommendation, synthetic respondent, expert network and more. Each definition is one paragraph and is written to be quoted.

#

Verified primary research

Research in which every human source has been identity and role checked, every document has been opened and read by a person, and every claim in the deliverable can be traced to one or the other. The term distinguishes this work from primary research that relies on self-reported panels, unverified expert profiles or AI-generated summaries. More.

#

Research provenance

The documented chain that connects every claim in a research deliverable to its source: who said it or where it was published, how the source was verified, when it was gathered and where the record is held. More.

#

Provenance appendix

The section of a research deliverable that lists every interview (role, organisation type, verification method, date, duration, record location) and every document (citation, passage relied on, who checked it and when), so the client can inspect the evidence chain. More.

#

Signed recommendation

A written recommendation that takes a position, states its confidence and the conditions that would change it, and carries the name of the individual researcher responsible for the work. Contrasted with an unsigned summary or a firm-branded report with no accountable author. More.

#

The Theory Standard

The published set of four rules Theory Intelligence applies to every deliverable: every source is real, every claim is traceable, AI use is disclosed, and a named person signs the recommendation. Written so it can be pasted into any research brief as contract terms. More.

#

AI use disclosure

A written statement in a research or advisory deliverable specifying where AI tools were used (for example scheduling, transcription, coding, drafting) and confirming that no factual claim, figure, quotation or citation was AI generated. Seventy percent of companies say they would not trust a report prepared with AI without such assurance (Source Global Research, 2026). More.

#

Synthetic respondent

A language model prompted to answer research questions as if it were a person with specified characteristics. Useful for drafting and pre-testing; unreliable for decisions because synthetic answers compress toward the mean, miss the tails and cannot know anything unpublished or after the model's training cutoff. More.

#

Expert network

A firm that recruits practitioners, screens them for compliance and sells paid conversations with them, typically by the hour. GLG, AlphaSights, Guidepoint, Third Bridge and Tegus are the largest; the market was roughly USD 3 billion in 2025. Networks sell access; they do not verify beyond compliance, interview, synthesise or sign. More.

#

Signal Brief

Theory Intelligence's fixed-price service that answers one business question in ten business days with typically six to eight verified sources drawn from people, records and data, and a signed two-page answer with provenance appendix, for USD 7,500 to 12,000. More.

#

Decision Study

Theory Intelligence's fixed-price engagement supporting one consequential decision with a full verified evidence base (typically fifteen to thirty sources across people, records and data), structured analysis, a signed recommendation and a working session, in four to six weeks, for USD 25,000 to 75,000. More.

#

Intelligence Retainer

Theory Intelligence's monthly service keeping a market, a set of counterparties or a risk under standing verified coverage: six to ten verified sources a month plus records and data, a signed monthly briefing with provenance appendix, indicators watched, and a named researcher who answers urgent questions in one to five business days, for USD 6,000 to 15,000 a month with a six-month minimum. More.

#

Commercial due diligence

Research conducted before an investment to test the commercial case: market size, customer behaviour, retention drivers, competitive position and pricing power. Theory delivers it as a Decision Study built on verified customer and operator interviews. More.

#

Win-loss analysis

Interviews with recent buyers who chose your product and buyers who chose a competitor, run by an independent researcher, to establish why deals are won and lost. Buyers say things to an independent interviewer that they do not say to a vendor's sales team. More.

#

Survey panel fraud

Compromised responses in online survey panels from bots, click farms, duplicate respondents and inattentive participants. Kantar estimates 30 to 40 percent of online survey responses are compromised; academic tests found bots pass quality checks 99.8 percent of the time. More.

#

Citation hallucination

A reference generated by an AI model that does not exist, does not resolve, or does not say what is claimed. The failure behind the 2025 Deloitte Australia refund and the 2026 PwC report corrections. Prevented by a person opening every source before use. More.

#

Independent vendor evaluation

An assessment of a technology product or supplier conducted by a firm with no commercial relationship to any vendor in the category, based on verified interviews with current users. Theory accepts no vendor fees, referrals or partnerships. More.

A term missing?

Write to info@theoryintel.com and we will add it, with a source.

Start a brief