Primary vs secondary research
for business decisions

The distinction

Secondary research is reading what exists: analyst reports, filings, trade press, academic papers, databases, and now the summaries of all of these produced by AI research tools. Primary research is generating evidence that does not yet exist: interviewing the customers, observing the process, surveying the population, reading the contracts. The first tells you what everyone can know. The second tells you what you need to know.

What secondary research is good for

Framing. Before spending on interviews, secondary research establishes the size of the market as others measure it, the competitors, the regulatory position, the history. It is fast and cheap, and it should be done first. Its limits are that every competitor has the same access, the sources lag the market by months, and the sources are of uneven quality. Since 2024 a growing share of published market analysis has been drafted with generative AI, and fabricated citations have been found in reports from leading firms. Secondary research now has to be verified, which erodes its cost advantage.

What primary research is good for

Answering. When the decision turns on what specific buyers will do, what a channel thinks, whether a target's customers will renew, or how a regulator will apply a rule, no published source knows. The people who know have to be found and asked. Primary research is slower, three to six weeks for a serious study, and costlier, USD 7,500 to 75,000 in Theory's published formats. In return the evidence is specific to the question, current, and yours alone.

What AI has changed

AI research tools have made secondary research nearly free and nearly instant. They have not made it more reliable; they inherit every error in the sources and add some of their own. Synthetic respondent tools claim to make primary research instant too, by simulating interviewees. Their vendors' own research shows they cluster around the average answer and miss the tails where decisions are made. The effect is that the value of verified primary research has risen: it is the one thing a competitor with the same tools cannot produce by pressing a button.

How to combine them

Use secondary research to write the claims to test. Use primary research to test them. Verify the participants in the primary research and record how. Disclose where AI was used. Have a named person sign the conclusion. This is the sequence in every Theory Decision Study, and the reason we describe our work as verified primary research rather than either term alone.