ad-buyer-group-mapper

Buyer Group Mapper

Given a deal size, industry, and offer, identify the roles likely to sit on the buying committee, then pair each role with a messaging angle and the targeting proxy that actually reaches it on an ad platform. The output is a segment-level role map for campaign planning - role patterns for a market segment, not a named-contact map of one live deal, which is a different artifact with a stricter evidence standard.

Buying committees are B2B-native. The underlying model partially transfers to multi-decider consumer purchases (households) - see that section for what carries over and what does not.

This skill produces the role map. Turning it into sized, budgeted, sequenced audiences is mbfinotti/advertising-skills@ad-audience-targeting. Writing the ads themselves is mbfinotti/advertising-skills@ad-copy-variants.

Clarifying questions

Ask once, briefly, before mapping; skip anything already answered.

1. What is the offer - and is this B2B, or a multi-decider consumer purchase?
2. What is the deal size / ACV band? (Drives committee size and whether per-role creative is affordable.)
3. What industry, and what buyer company size - SMB, mid-market, enterprise?
4. How large is the addressable market - hundreds of named accounts or tens of thousands?
5. What evidence about past buyers exists: closed-won/lost data, buyer interviews or call notes, review mining, account-level intent data?
6. Which platforms will run this, and at what monthly budget?
7. Is the purchase net-new for the category, a replacement, or an expansion? (Changes who is present.)
8. By what date must results land? A hard deadline promotes evidence you already own and reach you can buy this week; it demotes buyer interviews and anything needing a standing job to configure.
9. One-off campaign win, or a compounding asset the team re-uses each quarter? Compounding promotes closed-won analysis, buyer interviews and an owned account list; one-off promotes published bands and function-level reach.
10. What is the effort ceiling - research hours per account, creative variants you can actually produce, and how many roles you can coordinate across? A low ceiling promotes a merged map; a high one is what makes per-role creative affordable.

Every ranking in this skill is a default for a median team, not a law: it shifts with context and with who executes it. Re-rank each one against these answers and against what you already know about the user, then say in the output which answer moved which option. Signals worth weighing:

• An account list already owned.
• An in-house research function.
• An intent feed already licensed.
• A CRM too dirty to diff.

Evidence gate

• Every role is a hypothesis until evidence supports it. Never assert a committee; propose one and mark each role's status.
• Pick evidence sources by efficiency - roles confirmed per hour spent - not by strength alone. The axes disagree:
  • efficiency: closed-won/lost diff > call notes > review mining > buyer interviews > published research > intent data > title inference
  • value: closed-won/lost diff > buyer interviews > call notes > review mining > published research > intent data > title inference
  • effort: buyer interviews > intent data > closed-won/lost diff > review mining > call notes == published research > title inference
  • compliance cost: intent data > buyer interviews > all others == none
• Effort in orders of magnitude:
  • Buyer interviews: a week of scheduling across roles.
  • Intent data: a standing job to configure and re-tune topics.
  • Closed-won/lost diff: a day per deal-size segment.
  • Review mining: a few hours.
  • Call notes or published research: an hour to read what already exists.
  • Title inference: near-zero.
• Call notes and published research tie on effort because both are an hour of reading a document someone else already produced; neither buys any new collection. They do not tie on value - notes are about your buyers, the bands are about somebody else's.
• Act on the two places the axes disagree. Call notes carry nearly the value of buyer interviews at a fraction of the effort, because the recordings already exist. Intent data sits high on effort and low on value here, because it resolves to accounts and so never confirms a role.
• Compliance cost, where it applies. A third-party intent feed needs a licensing and privacy review before first use, and its terms constrain re-use once the data is inside your targeting stack. Recorded interviews need per-participant recording consent, and a call recorded without it cannot be cited as evidence later.
• Default rung: diff which roles appear in won deals against lost deals, filling gaps from call notes that already exist - the strongest derivation available to most teams. Move up to buyer interviews when the diff leaves a role's presence ambiguous across segments; add intent data only when the unknown is which accounts to cover, never which roles exist.
• What the efficiency order starves: buyer interviews, second on value and first on effort, lose every round to sources that already exist. Promote them the moment a role's presence stays ambiguous after the diff and the notes - that ambiguity is the one thing no existing document can resolve.
• Re-rank this order against the user: clean CRM stage history makes the diff near-free, no CRM hygiene drops it below review mining, an in-house research function makes interviews cheap, and an already-licensed intent feed makes intent near-zero effort.
• Delete, don't demote: a team that cannot obtain per-participant recording consent deletes buyer interviews from the menu and says so in the output; a team with no licensing route for third-party intent deletes intent data. A ruled-out source parked at the bottom of the ladder reappears later as scope nobody budgeted.
• Title inference is the only free source and the weakest one - it is the floor of the ladder, never a rung to stop on.
• Never invent an org chart from a title. Classify a role from seniority plus how close its function sits to the product category - nothing else. When title and department contradict each other, output "unknown", not a guess.
• A role with no evidence survives only as a labelled hypothesis with a stated disproof test (e.g. "cut security review if it appears in none of the next ten closed-won notes"). A role with no evidence and no disproof test gets cut.
• Apply the working test from persona research: if a role changes no targeting, creative, or offer decision, delete it. Roles that survive because they sound plausible are liabilities, not audiences.

Committee sizing

Derive committee size from the user's own closed-won data first. Published figures are a sanity band only - primary studies disagree by a factor of two because each defines "involved" differently and samples different deal sizes.

Source	Size	Qualifier to keep attached
Gartner	6-10 decision makers	complex solutions only - the qualifier is almost always dropped in citation
TrustRadius 2024 (2,164 buyers)	96% of groups have ≤5; SMB peaks at 2-3, enterprise at 4-5	tech purchases, self-report
6sense	~10.6 (North America)	behaviorally captured active participants
Forrester 2024 (>16,000 buyers)	13 internal stakeholders	broadest definition of "involved"

• Do not rank these four sources against each other: they sample different populations under different definitions of "involved", so any ordering would read as a quality ranking the disagreement does not support. Use them as a band and let the user's own closed-won data pick the number inside it.
• Seniority rises with deal size: 79% of purchases require CFO approval; 52% of groups include VP+ and 53% a C-suite executive (TrustRadius 2024).
• No primary study segments committee composition by industry or by purchase type (net-new vs. replacement vs. expansion). Say "no published data" instead of extrapolating.
• Sanity check both directions: a $6K ACV offer whose closed-won notes show two people does not get an eight-role matrix; a $500K enterprise deal mapped as champion-only is equally wrong.

Role model

Use this six-slot vocabulary and define it inline in the output - no canonical role list exists, and the frameworks in circulation genuinely disagree.

Role	Definition	Typically arrives at
Initiator	Surfaces the problem and starts the search	problem identification
Champion	Feels the pain, drives the evaluation, spends political capital publicly	problem identification → throughout
End user	Hands-on daily; judges impact on their own job	problem identification
Economic buyer	Owns the budget line; can approve or veto the spend	business-case stage
Technical evaluator	Screens against specs and architecture; veto power, no approval power	requirements building
Gatekeeper-blocker	Procurement, security, legal, compliance - substantive veto holders	validation

Vocabulary flags to carry into the output:

• Champion, mobilizer, and coach are not synonyms:
  • Coach: feeds intelligence, may stay hidden.
  • Champion: advocates publicly.
  • Mobilizer: coined because traditional champions are often friendly contacts who cannot drive consensus.
  • When evidence shows your champion can't move the group, the map needs a consensus-driver, not a louder champion.
• "Gatekeeper" has two meanings: the classic administrative information filter, and the modern procurement/security/compliance functions holding substantive vetoes. This skill means the second; say which you mean.
• Never use "influencer" as a role label. It covers anyone whose opinion is sought, so it yields no targeting attribute and no messaging angle.
• Merge slots freely downward: SMB committees of 2-3 typically collapse to champion-who-is-the-user plus economic buyer. Do not fill empty slots for completeness.

Per-role output

For each surviving role, fill all six fields - the angle is the deliverable; the first three are its inputs, and the proxy is what makes it operational:

Field	Answers
Measured on	What number or outcome is this role's job judged by?
Personal risk	What happens to them if this purchase goes wrong?
Likely objection	The first reason they'd say no, in their own words
Messaging angle	The one framing that connects the offer to their metric and defuses their risk
Proof / offer type	Evidence format and CTA weight this role will accept
Targeting proxy	The platform attribute combination that actually reaches this person

• Fine-grained per-role objection maps are weakly evidenced in published research - seniority-level content behavior is well evidenced, per-role objections mostly are not. Treat the starter patterns in references/worked-example.md as defaults to verify against the user's own call notes and reviews, not as facts.

Offer types are a menu, and its axes disagree sharply - the cheapest asset to produce is not the one to build first:

• efficiency: compliance/security pack > ROI or savings calculator > ungated guide or checklist > demo or trial request > sandbox access > benchmark report

• value: demo or trial request > sandbox access > ROI or savings calculator > compliance/security pack > benchmark report > ungated guide or checklist

• effort: benchmark report > sandbox access > ROI or savings calculator > demo or trial request > compliance/security pack > ungated guide or checklist

• compliance cost: compliance/security pack > sandbox access > all others == none

• Effort in orders of magnitude:

  • Benchmark report: a quarter to collect data nobody else holds.
  • Sandbox access: a quarter of engineering.
  • ROI or savings calculator: a week to build.
  • Demo or trial request: a standing job, since each one spends sales capacity again.
  • Compliance/security pack: an hour to assemble from certifications already held, a quarter if none exist.
  • Ungated guide or checklist: near-zero.

• The pack leads because it is nearly free when the certifications exist and it is the only asset that removes a validation-stage veto - the highest-value outcome any single offer buys in this map.

• The demo does not lead despite the best per-person value: match offer friction to role warmth, because low-friction offers (guides, calculators, benchmarks) convert around 10-15% and high-friction offers (cold demo or trial requests) around 1.5-4%. Cold gatekeepers do not book demos at all.

• Compliance cost, where it applies: publishing a pen-test report or audit letter ungated needs security and legal sign-off, and a document published ungated cannot be recalled; a sandbox touching real data pulls in a data-processing review.

• Default rung per role: the ungated low-friction offer. Move up one rung once a role's proxy audience engages twice, never on the first touch.

• Re-rank against what the user owns: an existing SOC 2 pack promotes the compliance rung, and a proprietary data set already in hand promotes the benchmark report from last to first.

• Delete, don't demote: with no sales capacity to staff them, demos and trials leave the menu entirely - say "deleted: no sales capacity" rather than ranking them last, where they read as a stretch goal and quietly become one. Same for the sandbox with no engineering quarter to spend, and for the benchmark report when no proprietary data exists.

• Write angles in verbatim customer language wherever review mining or call notes supply it - exact phrases beat polished descriptions because they are how the buyer actually thinks.

• Every angle must be distinct. Run th

...