Most questioning advice is a list. Ask open questions rather than closed ones. Ask about pain. Ask eleven of them. The lists survive because they are easy to teach and hard to falsify. The research underneath them says something narrower and more useful: in a complex B2B sale, what separates a strong conversation from a weak one is not how many questions a seller asks, but what work each question is doing. A call can hold twenty questions and do one job. A call can hold nine and do four.

This article sets out the four jobs a question can do, what the evidence says about each, and how to use the distinction when planning a call. The short version: intake questions, the ones that gather facts, are the cheapest questions in the room and the ones buyers have most often already answered on their own. The three jobs that move deals, which are insight, test, and trust, are the ones most call plans leave out.

6sense's 2025 B2B Buyer Experience Report, built on surveys of more than 4,000 recent B2B buyers, found the buying journey has moved from roughly a 70/30 split between independent research and seller engagement to 60/40. The more pointed finding sits in the same data: 94% of buying groups had ranked their preferred vendors before first contact with any seller, and they bought from that preliminary favorite 77% of the time. Gartner frames the same constraint from the other side, reporting that B2B customers spend only about 17% of the purchase process talking to potential suppliers at all, and that sliver is divided among every vendor under consideration.

Information scarcity is not the buyer's problem. In a Gartner survey of more than 1,000 B2B customers, 89% said the information they encountered during the purchase was high quality. What they struggle with is reconciling it, not finding it. So a seller who spends the first fifteen minutes of a first call collecting facts that sit on the buyer's website, in their annual report, or in the AI summary the buyer generated that morning is spending the scarcest asset in the deal, which is attention, on the most abundant output, which is information.

That is the case against intake as a default, not against intake entirely. Some facts are genuinely private: how the last vendor decision actually went, who signs when the number crosses a threshold, what happened when the team tried to fix this themselves. Intake is a real job. It is simply the smallest one.

Gong's analysis of more than 519,000 recorded B2B discovery calls found the highest success rates when reps asked between 11 and 14 targeted questions. Below that, a call tends to be too thin to establish anything. Above it, the call starts to read as an interrogation. The same analysis found that top performers spread questions across the call while average performers front-load them, which is what you would expect if questions are meant to follow a conversation rather than run it.

Two caveats matter. The data is correlational and drawn from calls recorded on one vendor's platform by customers who chose to record them, so it describes reasonably sophisticated sellers rather than the whole profession. And a count says nothing about content: eleven questions about headcount and tech stack will land inside the range and teach you nothing. Read 11 to 14 as a budget. The rest of this article is about how to spend it.

The largest observational study of sales conversations is still Neil Rackham's Huthwaite program: more than 35,000 sales calls observed across 23 countries over twelve years, with behaviour coded live by trained observers. Its central result was negative. The techniques that worked in small transactions, feature emphasis and hard closing, correlated negatively with success in large ones. What correlated positively was a sequence that moved from the buyer's situation to a specific problem, then to the implications of that problem, and only then to the value of solving it.

The implication question is the interesting one, because it collects nothing. It asks the buyer to do arithmetic out loud: if the handoff between support and engineering takes three days, what does that cost across a quarter, and who feels it. The seller usually suspects the answer already. The point is that the buyer says it, in their own numbers, in front of their colleagues.

Rackham himself has been careful about how the result gets used. Describing the validation work, he notes that the first thousand people trained on the model outperformed control groups by 17% in sales volume, then adds that the gain came from the coaching process built around the model rather than from "the four little questions." That caveat is the honest version of every questioning framework, including the one below. A taxonomy of questions is a planning aid. It does not execute itself, and it does not survive a rep who has never practiced it with a manager who listens to calls.

In The JOLT Effect, Matthew Dixon and Ted McKenna analyzed more than 2.5 million recorded sales calls and found that 40-60% of qualified pipeline ends in no decision. Of those losses, 56% came from indecision, meaning the buyer wanted the change and could not commit to it, usually out of fear of getting it wrong, and 44% came from genuine preference for the status quo.

That split has a direct consequence for questioning, because the two failure modes answer to different questions. Preference for the status quo shows up under intake and insight: the problem is not yet expensive enough to act on. Indecision does not show up there at all. A buyer who is afraid of being blamed for a bad purchase will answer every question about pain enthusiastically, agree with the business case, and still not sign.

Test questions exist to surface that. What would have to be true for this to look like the wrong call six months from now. If this stalls, what is the most likely reason. Who on your side would be quietly relieved if it did not happen. These are uncomfortable to ask because a candid answer can cost you the forecast, and that is exactly their value. An earlier article in this series set the rule that every qualification claim should carry a claim, a source, a date, and a falsifier. The falsifier comes from a test question. A deal claim without one is not a forecast, it is a hope.

The reason most sellers skip test and trust questions is a prediction error, and it has been measured. Kardas, Kumar, and Epley, writing in the Journal of Personality and Social Psychology in 2022, paired strangers and assigned them either shallow questions or deep ones. Participants predicted the deep conversations would be awkward and uncomfortable. Afterward they rated those same conversations as more enjoyable, more connecting, and less awkward than they had expected. The barrier is miscalibration: people systematically underestimate how interested the other party is in a real answer.

The companion finding concerns follow-ups. Huang, Yeomans, Brooks, Minson, and Gino, across three studies of live conversations, found that people who ask more questions, and follow-up questions in particular, are better liked, with speed daters who asked more questions more likely to be granted a second date. The mechanism is perceived responsiveness, which is the sense that the other person is listening, understanding, and validating. As with the depth finding, people did not anticipate the effect. It is worth noting that the Journal of Personality and Social Psychology published a correction to that paper in March 2025: an independent audit found minor errors in how some results were reported, and the substantive result of every hypothesis test in the paper was unchanged. A sales article that cites a psychology finding should also report what later happened to it.

Practically, the follow-up is the cheapest instrument available to a seller. It requires no preparation at all. It is simply the second question about the same answer: say more about the three days, or, when you say leadership pushed back, who specifically.

Questioning and listening are usually trained as separate skills, and they do not behave that way. Itani, Goad, and Jaramillo, in a meta-analysis of roughly two decades of research on salesperson listening published in the Journal of Business Research in 2019, found that listening drives both relational outcomes and sales performance, operating directly and through adaptive selling, and that it is about as consequential for performance as adaptive selling itself.

The practical form of that finding is blunt. A question you do not follow is worse than a question you never asked, because it tells the buyer the conversation is a form being filled in. The test is whether your next question came from their last answer.

Gartner surveyed 645 B2B buyers between August and September 2025 and published the results in May 2026. Buyers used an average of seven information sources during a purchase, 45% used generative AI, and 69% said they prefer to validate AI-generated insights with a sales rep. Robert Blaisdell, the VP Analyst who presented the work, framed it as a change in the seller's role rather than a reduction of it: sellers need to "show up differently, engaging where they can help buyers validate information, reduce risk and move forward with greater confidence." 6sense's data points the same way, with 89% of 2025 purchases including AI features and 58% of buyers engaging sellers earlier to clarify AI details they could not resolve on their own.

This quietly changes what a good question looks like. The buyer arrives holding an answer assembled from seven sources, one of which is a model that knows nothing about their contract terms, their integration backlog, or the vendor consolidation their CFO announced in July. The useful question is no longer what are your goals for this year. It is closer to this: the comparison you ran, what did it assume about implementation time, and who on your side has checked that assumption.

Every question in a B2B conversation does one of four jobs. Naming the job is what makes a call plan testable.

Intake gathers facts you cannot get anywhere else. Its failure mode is asking for facts that are already public, which spends buyer attention on information and signals that you did not prepare.

Insight makes the buyer see their own situation differently, usually by asking them to quantify a consequence out loud. This is the implication question from the Huthwaite research, and it is the only job that leaves the buyer better off whether or not they buy. Its failure mode is the seller supplying the arithmetic instead of the buyer.

Test tries to falsify your own read of the deal. It produces the falsifier a qualification claim needs, and it is the job most often skipped because a truthful answer can damage the forecast. Its failure mode is asking only questions whose answers confirm what you already believe.

Trust makes it safe to say the uncomfortable thing: the internal politics, the last failed rollout, the fear of owning this decision. It is how indecision becomes visible, and indecision accounts for the majority of no-decision losses. Its failure mode is the seller assuming the question is too intrusive to ask, which the evidence on deep conversations says is usually wrong.

The planning rule follows in one step. Write the call plan, then label each question with one of the four jobs, then run three checks. If more than half the plan is intake, you are interrogating the buyer about facts they published. If nothing on the plan is a test, you will leave the call with a forecast you cannot falsify. If nothing is trust, you will not hear about the indecision that kills more deals than your competitors do. Keep the total inside the 11 to 14 range, and spread it across the call rather than stacking it at the front.

Do your intake before the call, not during it, and bring the two or three private facts you genuinely cannot look up. Plan one insight question per problem you believe is real, phrased so the buyer, not you, supplies the number. Plan at least one test question you are slightly afraid to ask, and treat its answer as the falsifier attached to your forecast. Plan one trust question, and ask it later in the call once you have earned the right. Then follow every answer with a second question about that same answer, because the follow-up is free and it is the part the buyer actually feels. A call run this way will usually be shorter than the one you planned, and the notes will contain something your CRM has never held before: the reason this deal might not happen, in the buyer's own words.