We set out to write about the eight things you should extract from every discovery call. The number came from us, not from the evidence, and once we went looking for support it did not survive.
What replaced it is more useful. The research supports an editorial structure rather than a count, and it draws a line most discovery checklists ignore: the information that predicts whether you win is not the same information that predicts whether the engagement succeeds.
There is no magic number
There is no source demonstrating that eight discovery items, or seven, or any fixed count, maximises win rate or delivery success in professional services. Every framework asserting one is asserting an editorial preference dressed as a finding.
The frameworks disagree on purpose
The major methodologies are not competing versions of one checklist. They optimise different moments in a sale, which is why merging them produces an incoherent list.
SPIN wants Situation, Problem, Implication and Need-payoff: context, difficulty, consequences, and then the value of solving it. Huthwaite describes these as a persuasive framework rather than a rigid sequence. Its centre of gravity is making a problem consequential enough to act on.
MEDDPICC wants Metrics, Economic Buyer, Decision Criteria, Decision Process, Paper Process, Identified Pain, Champion and Competition. That is a qualification instrument. It is concerned with whether a deal is real and how it will get signed, not with whether the work will succeed.
Challenger emphasises reframing the buyer's understanding, and JOLT concentrates on indecision rather than preference, which is a materially different diagnosis of why deals stall.
Read together, they are answering different questions. A checklist that merges all four produces thirty fields, most of which nobody fills in.
The won-deal record
Five categories cover what the qualification evidence and the commercial datasets repeatedly point at. These are not five independently proven causal variables. They are the smallest practical synthesis that covers the constructs supported across independent research.
- Problem and consequence. What is wrong now, and what happens if it stays wrong. SPIN separates problem from implication for a reason: naming a problem is much weaker than understanding its operational or economic consequence.
- Desired outcome and value metric. What the buyer wants to be true afterwards, expressed in something measurable.
- Stakeholders, authority and alignment. Who decides, who can veto, and whether they currently agree.
- Decision criteria and buying process. What they will compare you against, and the mechanics of how a signature happens.
- Urgency and decision risk. Why now, and what makes doing nothing attractive. This is where the risk of inaction case gets its raw material.
The delivery record
A seller can qualify a deal perfectly and still sell a bad engagement. Three further categories matter far more once the outcome you care about is delivery rather than signature.
Requirements, constraints and dependencies. What has to be delivered, and what it depends on: systems, data, people, access, decisions. What is explicitly out of scope. Which assumptions would change the effort if they turned out false. The nearest rigorous evidence here comes from software and project work rather than boutique GTM consulting, and that limitation is worth stating plainly. Kamata and Tamai examined requirements quality against project cost and found the relationship real but conditional.
Success criteria and expectations. How the client will judge the work, in their words, captured before anyone writes a scope.
Working assumptions about the environment. Prior attempts, why they failed, and what the organisation has already rejected. A programme that failed because managers never changed their behaviour tells you more about the next engagement than any stated requirement.
The delivery record is the part that gets skipped, because none of it is needed to get the deal signed.
What sellers actually fail to capture
The available evidence here is less flattering to the sales methodology industry than its marketing suggests.
Ebsta's dataset reports that only 36% of deals which had passed discovery contained both a qualification score and supporting notes. In the same proprietary dataset, stronger qualification was associated with higher conversion and less slippage. Treat that carefully: it is evidence that sellers frequently leave the CRM without a usable record. It is not evidence that filling the boxes mechanically would produce the uplift, and the vendor publishing it sells qualification software.
The more consequential omission is usually the delivery record, precisely because nothing in the sales process forces it. Nobody is blocked from sending a proposal because the dependencies were never captured. The cost lands months later, on a different team.
When more discovery is the wrong move
The strongest counter-argument to any extraction checklist is not that discovery is unimportant. It is that more discovery can be the wrong behaviour.
The independent sales evidence supports accuracy and adaptation rather than information volume. Franke and Park found adaptive selling more strongly linked to performance than customer orientation alone, including on objective measures. Hall, Ahearne and Sujan found that accurate judgements help, but inaccurate deliberative judgements can actively compromise performance.
A seller who asks thirty questions and records every answer can end up worse informed than one who asks five discriminating questions, because volume creates the feeling of understanding without the substance. The purpose of a structure is to make sure you did not miss a category that matters, not to maximise fields completed.
Common questions
Should we just adopt MEDDPICC and be done with it?
MEDDPICC is a strong qualification instrument and a weak delivery instrument. If you adopt it, adopt it knowing it will tell you whether a deal is real and almost nothing about whether the engagement will go well.
Who should own the delivery record?
Whoever will be accountable for the work. The practical failure is that discovery is run by someone who will not deliver, and the categories that only matter later never get asked about while the client is still in the room.
Can this be extracted from a transcript automatically?
The factual categories can, and that is what we build. What cannot be automated is the judgement about which unresolved question actually gates the deal, which is why extraction should produce a reviewed record rather than a finished document.
What if the client will not answer the harder questions?
Record that as the finding. A buyer who cannot describe their decision process, or will not name a consequence, has told you something specific about the deal. Absence is data, provided you write it down rather than leaving the field blank.