The Shortlist Before the Conversation

There is a well-known figure, attributed to Gartner, holding that buyers spend 17% of their time meeting potential suppliers during a B2B purchase. It appears in conference slides, vendor blogs and strategy documents, usually without a report title or a year. Following it to a source is unrewarding: the pages that host it do not open to outside readers, the underlying research appears to predate 2019, and no version of it carries a methodology anyone can inspect.

The claim it makes is probably directionally right, which is why nobody checks. That is the more interesting observation. An assertion about how buyers behave has circulated for the better part of a decade in an industry that describes itself as data-led, without a single retelling stopping to establish where it came from. A statistic that is repeated because it is familiar has stopped functioning as evidence and started functioning as a shared belief.

The underlying question is worth asking properly, and it can be, because there is openable research on it.

What can be established

6sense’s 2025 B2B Buyer Experience Report, based on roughly 4,000 responses, reports that buying groups first contact a seller 61% of the way through their buying journey, earlier than the 69% recorded in the previous edition. The same study finds that 94% of buying groups had already assembled a ranked shortlist before speaking to any seller, and that 95% of completed purchases went to a vendor that was on that Day One shortlist, up from 85%.

Read together, these describe a specific structure rather than a general trend. The majority of the decision happens before contact; contact happens slightly earlier than it used to; and the set of vendors under consideration at the moment of contact is almost never expanded afterwards. The conversation is where a preference is confirmed, not where the field is opened.

It is worth being precise about what the study is. It is a survey of buyers’ own accounts of their process, conducted by a company that sells software premised on the existence of an unobservable pre-contact stage. Self-reported process reconstructions are imperfect, and the vendor’s interest is real. The figures are used here because the sample is large, the report states its basis, and the finding is consistent with independently gathered evidence about how information is gathered before a vendor is approached. They are not used as a precise measurement of anything.

The composition of the invisible period

If the decisive work occurs before contact, the question becomes what material it draws on, and here the picture has changed quickly.

The same 6sense study reports that 94% of B2B buyers say they used large language models during their buying process. This is a statement about tool use rather than about influence, and it should not be inflated into one about outcomes. But it establishes that the pre-contact period is now substantially mediated by systems that assemble accounts of vendors from public material, which is a different information environment from the one the older folklore described.

That matters given what has been measured about those systems. The Tow Center for Digital Journalism found in March 2025 that eight AI search tools answered source-attribution queries incorrectly more than 60% of the time. A buying group forming a ranked shortlist through such a system is not receiving a neutral rendering of the available evidence; it is receiving a synthesis with a measured failure rate, and it has no more visibility into the failures than the vendor does.

The traditional intermediary that once occupied this space is in retreat. TrustRadius and Pavilion’s 2024 study of 2,164 technology buyers found analyst-report usage at a seven-year low of 16%, with 49% of buyers seeking conversations with existing users to check what a vendor had told them. Buyers have not stopped wanting independent verification. They have stopped getting it from the parties that formerly supplied it, and are assembling it themselves from a mixture of peer conversation and machine synthesis.

The consequence for how vendors describe themselves

Most vendor material is written for the meeting. It assumes a reader who has raised a hand, arrives with context, and can ask a question when something is unclear. Discovery calls, tailored decks, references and demonstrations are all designed for that setting.

The evidence above says that setting is downstream of the decision. During the period when the shortlist is formed, the vendor is represented by whatever is publicly readable about it, assembled by parties it cannot see, with no opportunity to clarify. A vendor whose public record is thin, out of date, or comprehensible only to someone who already knows the category is not making a weak case at that stage. It is making no case, and the shortlist is formed without it.

Two implications follow, both unwelcome.

The first is that the pipeline is a lagging indicator of a process that concluded earlier. A vendor that is not contacted has not lost a deal; it was never in the set, and nothing in its own systems records the omission.

The second is that the material which matters most at the deciding moment is the material least suited to measurement: registry entries, category descriptions, third-party analysis, documentation that explains what the product is in terms an outsider can follow. None of it produces an attributable outcome. All of it is what a buying group and its tools have to work with during the phase that decides the result.

The folklore statistic and the survey evidence agree on the shape of the problem. The difference is that one of them can be opened, and the one that cannot has been doing the industry’s thinking for it for ten years.