What discovery questions should you ask, by buyer role?
Ask each person on the call questions built for their role, and give every question one job: to learn the cost of the problem, who decides and how, what has to be true for them to switch, or what they personally win or lose. Our bank below is a 4x4 grid: four buyer roles (economic buyer, champion, user, gatekeeper) by those four jobs, sixteen questions in all, and you ask three per person present, never the whole grid. That keeps a two-person discovery call near the 11-14 questions that Gong's data from 519,291 sales conversations on its platform tied to the best outcomes.
How we research this: facts come from each vendor’s public pages and named studies, dated where shown; the math is ours and labeled. Vendor prices and limits are as of October 9, 2026.
Updated
Why sort discovery questions by role instead of by topic?
Because the signer, the pusher, the user and the vetoer hear the same question four different ways. Ask "what does this cost you?" and a user gives you hours, a CFO a budget line, and IT a shrug. One list for everyone is why reps sound like a form.
The roles come from b2b-buying-committee-size. The four jobs are our framework, not a standard.
- Cost: what the problem costs, and in what unit.
- Decision: who decides, in what steps, by what date.
- Switch: what has to be true for them to leave the current way, including what they already tried.
- Stake: what this person wins or loses personally.
A question that fits no job is a pitch in disguise. Cut it.
| Role | Best placed to answer |
|---|---|
| Economic buyer | Decision, Cost in money |
| Champion | Stake, Switch |
| User | Cost in hours, Switch |
| Gatekeeper | Decision steps, Switch conditions |
Our guidance, not a measured result.
What is the 4x4 discovery question bank?
Columns are who is in the room; rows are what you still do not know. Written for a generic B2B software sale; swap the nouns.
| Job | Economic buyer | Champion | User | Gatekeeper |
|---|---|---|---|---|
| Cost | If this problem stays exactly as it is for another year, what does that show up as on a number you report? | When this goes wrong, what do you end up doing by hand, and how often in a month? | Think of the last time this broke. How long did the fix take, start to finish? | What has a vendor in this category cost you in review time before, and what made it drag? |
| Decision | When you last approved something at this price, what did you need to see before you said yes? | If you took this upstairs next week, who would be in the room and who would push back first? | Who asked you to be on this call, and what do they want you to tell them afterwards? | What does your review need from us, in what order, and which step usually stalls? |
| Switch | What would make you keep the current way even if the numbers favor a change? | What did the last fix for this look like, and why didn't it stick? | Which part of today's setup would you miss most if it vanished? | Which requirement is a hard no, and which is a strong preference? |
| Stake | Which goal of yours does this problem get in the way of this year? | How does fixing this change the way your own year gets judged? | What part of your week gets better or worse if this shows up? | What happens to you if a vendor you cleared turns out to be a problem? |
Pick cells, never read rows: find the job you cannot yet put a number or a name next to, and ask that cell.
No budget cell, on purpose. You back into budget from Cost plus Decision; asking it cold on call one earns a reflex no. If it comes up anyway, see how-to-handle-no-budget-objection, and bant-vs-meddic for where budget sits in a framework.
Our prompt for one named contact's four cells:
We sell [PRODUCT, one line] to [INDUSTRY]. My discovery call is with [NAME] ([TITLE], [COMPANY], [SIZE]). Treat them as the [ECONOMIC BUYER / CHAMPION / USER / GATEKEEPER].
Write 4 discovery questions for this person, one each for: the cost of the problem, how the decision gets made, what has to be true for them to switch from the current way, and what they personally win or lose.
Rules: open questions only, each starting with What, How, Which or Who, one question each, no product features, no claims about [COMPANY] beyond what I gave you. After each question, add one follow-up that asks for a number, a date or a name.
What's the question budget for each person on the call?
Our rule: three per person present, one per job, never the full grid. Alone with one person? The other roles' cells become "who should I ask about this?"
For the outer limit, lean on Gong. It looked at 519,291 sales conversations (blog first published 2018), and its data puts the sweet spot at around 11-14 questions per discovery call, with diminishing returns past 14. Gong also reports a 46:54 talk-to-listen ratio for top performers in discovery, who ask questions as they come up while average performers fire theirs at the start. That is Gong's data on its platform, not a market average.
Cold call pacing (three questions total) lives on discovery-call-without-interrogating.
| People on the call | Prepared questions | Where the rest of the grid goes |
|---|---|---|
| 1 | 3, plus follow-ups | Ask for introductions to the missing roles |
| 2 | 6, plus 5-6 follow-ups | 11-12 total, inside Gong's band |
| 3 or more | 9 | 9 + 9 = 18 is past 14: cap follow-ups at 5 or split the call |
Our guidance. One more rule, also ours: every prepared question earns at most one follow-up, and it asks for a unit (hours, dollars, a date, a name). No unit, no answer.
What does a good answer sound like for each job?
| Job | Weak answer (keep digging) | Good answer (write it down) |
|---|---|---|
| Cost | "It's a pain." | "About 30 hours a month across the support team." |
| Decision | "I'd need to check." | "Our COO signs after a two-week procurement check." |
| Switch | "We're open to anything." | "The last tool died because nobody updated it." |
| Stake | "It would be nice." | "My bonus rides on first response under two hours." |
The unit follow-up turns the left column into the right one. A good answer closes the cell: log it in the CRM under the job name, not the question.
What does a two-person discovery call look like with the bank?
Invented example, our math, for illustration. Lakeview Supply (invented, 220 people) is a prospect for our invoice-matching software. Thirty minutes with Dana, controller and likely champion, and Raj, CFO and economic buyer.
Question budget. 2 people x 3 = 6 prepared, plus at most 6 follow-ups: a ceiling of 12, inside Gong's band. We plan 6 + 5 = 11.
Time budget. At the 46:54 ratio Gong reports for top performers, we talk about 30 x 0.46 = 13.8 minutes and the buyers about 30 x 0.54 = 16.2. 16.2 / 11 = roughly 1.5 minutes of buyer talk per question. Past three minutes on one answer, move on (our reading).
Cells chosen. Dana: Cost, Switch, Stake. Raj: Decision, Cost, Switch; his Stake is skipped (first meeting, see below).
Answers (invented). - Dana, Cost: "About 120 invoices a week keyed by hand, 6 minutes each." Unit check: 120 x 6 = 720 minutes = 12 hours a week, about 52 hours a month (x 4.33). - Dana, Switch: "We tried an OCR plug-in. Nobody trusted it." - Dana, Stake: "My target is closing the books in 5 days. We take 8." - Raj, Decision: "Over $25k a year needs me plus a three-week IT security review. I sign." - Raj, Cost: "$14,000 in late-payment fees last year. Partly AP." - Raj, Switch: "I'd keep the current way if rollout collides with year-end close."
Ledger. 6 cells filled, 10 not: 5 skipped on purpose (Raj's Stake, four Gatekeeper cells) and 5 open (Dana's Decision, four User cells). 6 + 5 + 5 = 16.
| Job | Economic buyer (Raj) | Champion (Dana) | User | Gatekeeper |
|---|---|---|---|---|
| Cost | filled | filled | empty: ask who | skipped |
| Decision | filled | empty | empty: ask who | skipped |
| Switch | filled | filled | empty: ask who | skipped |
| Stake | skipped | filled | empty: ask who | skipped |
The empty User column becomes the next step: "Dana, can you bring one AP clerk to Thursday's demo?" IT security is a known three-week step with no contact yet.
CRM entry. Cost = about 52 AP hours a month plus part of $14k in late fees. Decision = Raj signs, three-week security review, $25k threshold. Switch = failed OCR plug-in; rollout must miss year-end close. Stake = Dana's close, 8 days to 5. Every filled cell has a unit. Now it's a deal, not a vibe.
Which questions do you skip, and when?
When should you skip each role's questions?
- Gatekeeper questions: skip them on call one unless security or procurement is obviously in play. Add the gatekeeper late.
- Stake questions to an economic buyer you have never met: skip them. Ask the champion to tell you instead.
- Switch questions when there is no current way (greenfield): swap in "What would you do without us?"
- Most of the bank on a renewal with a known buyer: ask only Decision.
Whoever should answer an empty column is your next booking. See how-to-find-the-decision-maker, and why-b2b-deals-end-in-no-decision for why a second person early matters.
More cells per meeting means fewer meetings; see how-many-meetings-to-close-one-deal. And when your champion quits, re-run the bank with the replacement: what-to-do-when-your-champion-leaves.
See also: How do you prep a sales call in three minutes with AI?
Sources
Research and vendor pages accessed October 6, 2026.
The four-job model, the sixteen questions, the three-per-person rule and the worked example are Pipe Playbook guidance; the example is invented.