Skip to content

Workflows · Guide

Sales Discovery: Build Evidence Across the Opportunity

Maintain source-backed facts, hypotheses, conflicts, stakeholder authority, proof, risks, and decisions across the full opportunity—not one call.

Updated August 8, 202613 min readSiddharth GangalBy Siddharth Gangal
Workflows

13 min read · Updated August 8, 2026

Sales discovery is the continuous discipline of building, testing, and updating decision-grade evidence about an opportunity across calls, emails, documents, product evaluations, stakeholder changes, and buyer decisions. It is not one meeting, a question count, a talk ratio, or a completed CRM form.

This page owns discovery across the opportunity. Use the sales discovery call plan for a live meeting, the discovery call framework for branches and stages, discovery questions for a question library, sales call preparation for pre-meeting research, and deal qualification criteria for investment gates. Those are narrower jobs.

Sales discovery is an opportunity-wide evidence discipline

The purpose of discovery is not to fill every field. It is to reduce consequential uncertainty enough for the buyer and seller to make the next decision safely. Some opportunities require several stakeholder conversations and artifacts; others should stop after evidence shows no relevant problem, authority, permission, or fit. There is no universal call length or number of questions.

Begin with a decision: should both parties investigate, validate proof, involve another owner, address a risk, agree commercial work, pause, or stop? Then identify the propositions that decision depends on. A proposition is precise enough to verify or disprove—for example, “Finance owns approval for spend above the buyer's stated threshold,” not “Finance involved.”

Discovery is distinct from qualification. Discovery gathers and updates evidence; qualification applies an approved investment rule to that evidence. It is also distinct from a deal review, where managers inspect evidence and choose actions.

Build a proposition-level discovery record

Use one row per proposition, not one narrative blob. A useful record contains:

FieldRequired question
PropositionWhat precise statement could be supported or contradicted?
DomainAccount, problem, impact, outcome, stakeholder, decision, proof, risk, or action?
EvidenceWhat exact statement, artifact, event, or system record supports it?
Source authorityIs this source authorized and competent for this proposition?
ProvenanceWhere did it come from, when, under which version and permission?
StateVerified, contradicted, unknown, stale, proposed, or not applicable?
Owner and expiryWho verifies it, and what change or date triggers review?
Next actionWhat evidence would resolve or materially advance it?

Separate buyer statement, observable fact, seller calculation, inference, and approved claim. Confidence without a rationale is not a state. Preserve the original evidence and later correction rather than silently rewriting history.

The U.S. GAO's 2025 Green Book applies to federal internal control, not sales discovery. Its focus on objectives, reliable reporting, sources of data, and documentation offers a bounded control-design reference: records should support the decision and remain reviewable.

Map eight evidence domains across the opportunity

DomainEvidence soughtDo not assume
Context and triggerBuyer entity/use, current event, owner, source, timingA public signal creates permission or priority
Current state/problemWorkflow, example, affected population, constraints, existing responseOne stakeholder represents every user
ImpactBuyer-defined unit, baseline, consequence, owner, time windowA seller ROI estimate is buyer fact
Desired outcomeSuccess definition, measurement, trade-offs, acceptanceFeature interest equals a committed outcome
Stakeholders/authorityRoles, decision rights, influence, users, risk ownersTitle proves authority
DecisionCriteria, process, gates, alternatives, dates, rejection pathA target date is a buyer process
Proof and riskClaims, substantiation, test, privacy/security/legal/implementation limitsA demo proves production performance
Next actionDecision, owner, artifact, dependency, date, stop condition“Follow up” is a mutual commitment

Not every domain needs equal depth for every decision. Scope evidence to consequence and uncertainty. A low-risk exploratory conversation needs less proof than a claim about regulated use, security, financial return, or employee monitoring.

Match evidence to source authority

Authority is proposition-specific. A user can explain their workflow; a budget owner can explain their authority; security can define its control requirements; legal/procurement can describe its process; an approved product owner can substantiate product behavior. A champion's summary may be useful but should not silently replace the responsible owner's evidence.

Build a source matrix with proposition type, primary source, acceptable secondary source, permission, expiry, and escalation. Record when a source is speaking from experience versus authority. If a buyer says “security will approve,” preserve it as that stakeholder's statement until security evidence arrives.

External research can form a hypothesis, not private account truth. Verify company identity, date, provenance, and relevance before using public events. Never represent scraped or inferred personal information as something the buyer told you.

Preserve conflicts, unknowns, and change

Conflict is discovery output. If finance and the champion name different approval routes, store both statements, sources, dates, and scope. Do not pick the answer that best supports the close date. Assign an authorized resolver and ask a neutral verification question.

Unknown is also valid. Required fields that prohibit unknown encourage invented values. Use state plus reason: not asked, source unavailable, permission denied, stakeholder disagreement, artifact pending, or changed since verification. Stale evidence remains visible but cannot support the current decision without review.

Material changes include reorganization, budget revision, new alternative, security finding, implementation constraint, stakeholder departure, changed metric, revised procurement route, or altered scope. Each should invalidate affected propositions, not every unrelated discovery record.

Govern seller claims and buyer proof

Discovery is reciprocal: the seller asks for buyer evidence and must substantiate its own objective claims. Maintain a claim register with exact wording, intended audience/use, evidence, limitations, version, approver, expiry, and prohibited extrapolations.

The FTC's Advertising Substantiation Policy is U.S. advertising law guidance, not a global sales script. It states that objective advertising claims require an appropriate reasonable basis before dissemination. Apply relevant law with qualified counsel; operationally, do not promise performance, savings, accuracy, compliance, or customer outcomes before approved evidence supports the exact claim.

Map proof to the buyer's criterion. A security document supports specified controls as of its date; it does not prove a buyer's full risk acceptance. A case study describes its named context; it does not guarantee another customer's result. A pilot tests predeclared cases; it should disclose data, environment, exclusions, failures, and acceptance rules.

Control access, privacy, and sensitive inference

Discovery may collect personal information, recordings, employee concerns, financial assumptions, technical architecture, legal risks, and confidential plans. Define purpose, minimum necessary fields, source, access, sharing, retention, correction, deletion, export, and incident handling.

The NIST Privacy Framework is voluntary privacy-risk guidance, not law or certification. Use it as a governance reference; determine applicable recording, employment, privacy, marketing, sector, and jurisdiction rules with qualified owners.

Do not infer sensitive traits or hidden motives from tone, facial behavior, transcripts, or public data. AI-generated summaries must label sources and inference, respect permissions, abstain when evidence is missing, and require human approval before consequential CRM writes or buyer-facing claims.

Carry discovery across interactions and owners

Before each interaction, select unresolved propositions relevant to the next decision. During it, capture source and context rather than optimizing question count. Afterward, reconcile evidence, contradictions, permissions, and next actions. Send a buyer-facing recap only for information appropriate to share and invite correction.

When ownership changes, hand off the latest proposition register, source artifacts, unknowns, contradictions, claim limits, privacy restrictions, decisions, and next actions. Do not hand off only an AI summary. Use stable account, contact, opportunity, meeting, and artifact IDs.

Salesforce documents that Opportunity History can record selected field changes and actor details. That does not guarantee full discovery history: configuration decides tracked fields, and external artifacts and relationships need their own controls.

Reconcile a worked discovery inventory

Fictional example—not a benchmark: a frozen opportunity record contains 20 decision-relevant propositions: nine verified, four contradicted, five unknown, and two stale.

StateCalculationCoverageNext treatment
Verified9 ÷ 2045%Retain source and expiry; use only within scope
Contradicted4 ÷ 2020%Preserve both sources; assign authorized resolver
Unknown5 ÷ 2025%Ask, obtain artifact, abstain, or stop
Stale2 ÷ 2010%Reverify after material change

The states reconcile: 9 + 4 + 5 + 2 = 20 and 45% + 20% + 25% + 10% = 100%. This is evidence coverage, not a discovery quality score, win probability, forecast, or permission to advance. One contradicted security gate may matter more than nine verified contextual facts.

Refresh discovery when material facts change

Reopen only affected propositions when a source, scope, stakeholder, system, assumption, or decision changes. A dedicated rediscovery call may help after a major gap or reset, but updates can also come through documents, tests, stakeholder meetings, or approved system events.

Use a change record: event, affected propositions, before/after state, source, actor, time, decision impact, owner, next action, and buyer correction where appropriate. Feed unresolved high-risk issues into the deal review; do not hide them to preserve forecast optics.

Opportunity discovery checklist

☐ Name the next mutual decision and required propositions

☐ Separate facts, buyer statements, seller calculations, hypotheses, and claims

☐ Store exact evidence, source authority, provenance, permission, and as-of time

☐ Permit verified, contradicted, unknown, stale, proposed, and not-applicable states

☐ Map context, problem, impact, outcome, stakeholders, decision, proof/risk, and action

☐ Preserve disagreements and assign an authorized resolver

☐ Substantiate seller claims before use and disclose scope and limits

☐ Minimize sensitive information and control access, retention, correction, and deletion

☐ Hand off source artifacts, unknowns, conflicts, permissions, and version history

☐ Reverify only affected propositions after material change

Good discovery does not create certainty on demand. It makes the evidence, uncertainty, conflict, and next decision visible enough for both sides to act responsibly.

Sources and evidence

Sources support the specific claims linked from this article. Vendor documentation establishes documented behavior, not independent outcomes.

  1. 01
    2025 Green BookU.S. Government Accountability Office · Effective fiscal year 2026
  2. 02
    Policy Statement Regarding Advertising SubstantiationU.S. Federal Trade Commission · Accessed August 8, 2026
  3. 03
    Opportunity HistorySalesforce Help · Accessed August 8, 2026
  4. 04
    Privacy FrameworkNational Institute of Standards and Technology · Accessed August 8, 2026

Keep reading

Related posts

Ready to evaluate the workflow?

Review the configured system with your team.

Confirm integrations, permissions, write authority, human review, failure handling, and current commercial terms before rollout.