Skip to content

mbfinotti/advertising-skills / skills/ad-buyer-group-mapper

ad-buyer-group-mapper

Verified

Map the likely buying committee for an offer, deal size, and industry, and give each role a messaging angle plus the ad-targeting proxy - job function and seniority, account list, account-level intent - that actually reaches them, sequenced by buying stage. Use whenever the user mentions a buying committee or buying group, decision makers, influencers or blockers, who signs off, targeting the economic buyer or a champion, per-role ad messaging, or whether one broad message beats per-role targeting - even if they never say 'buying committee'. Covers B2B committees and multi-decider household purchases. Do NOT use to size or build the audiences themselves - use mbfinotti/advertising-skills@ad-audience-targeting instead.

FileKindSize
SKILL.mdskill md29.0 KB
evals/evals.jsonconfig27.7 KB
references/worked-example.mdmarkdown14.4 KB

evals/evals.json

View on GitHub
{
  "skill_name": "ad-buyer-group-mapper",
  "evals": [
    {
      "id": 1,
      "prompt": "I run demand gen at Kindra Systems - workforce-analytics platform, $90K ACV, selling to mid-market healthcare (500-2,000 employees). We're building an ABM program and I need to figure out who is actually on the buying committee so we can run ads per role. My plan: kick off 15 buyer interviews next quarter to map the committee properly, and we just licensed a third-party account-intent feed - I'm hoping it can show us who at each account is involved in the purchase. What we have today: 42 closed-won and 30 closed-lost deals in the CRM with clean stage history, and call recordings on most of them. The campaign has to launch in 6 weeks. How should we build the committee map?",
      "expected_output": "A committee-mapping plan that starts from the won/lost diff plus existing call notes, defers the interviews, restricts the intent feed to account ranking, and delivers roles as evidence-graded hypotheses with disproof tests.",
      "files": [],
      "expectations": [
        "Recommends diffing which roles appear in closed-won deals against closed-lost deals as the first evidence step, before any new data collection",
        "Uses the existing call recordings to fill gaps left by the won/lost diff rather than commissioning new research",
        "Does not endorse running the 15 buyer interviews as the first step; interviews are deferred until a specific role's presence remains ambiguous after the diff and the call notes",
        "States that the 6-week deadline demotes buyer interviews (about a week of scheduling) in favor of evidence the team already owns",
        "States that the intent feed resolves to accounts, not people, and therefore cannot confirm which roles sit on the committee",
        "Restricts the intent feed to ranking which accounts to cover, never to confirming roles or identifying committee members",
        "Quantifies intent-data limits: roughly 81% account-level accuracy, or equivalently that about one in five surging accounts has no genuine near-term intent",
        "Labels every proposed committee role as evidenced or hypothesis instead of asserting the committee as fact",
        "Attaches a concrete disproof test to at least one hypothesis role, such as cutting it if absent from a stated number of upcoming won-deal notes",
        "Identifies title inference as the weakest evidence source, a floor never to stop on",
        "Gives effort in orders of magnitude: the won/lost diff about a few hours per deal-size segment, buyer interviews about a week of scheduling",
        "Flags a compliance cost on at least one source: per-participant recording consent for interviews, or a licensing and privacy review before first use of the intent feed"
      ]
    },
    {
      "id": 2,
      "prompt": "We sell Denticor, a $7K/year scheduling and recall tool for independent dental clinics (1-3 locations). Our new VP of marketing read that B2B purchases involve 6 to 10 decision makers, so she wants ad campaigns built for eight personas: practice owner, office manager, lead dentist, IT consultant, procurement, compliance, finance, and front-desk staff. Looking at our CRM notes from the last 40 wins, it's almost always just the practice owner and the office manager on the calls. Can you map the buying committee and the campaigns per role?",
      "expected_output": "A two-role map derived from the user's own win notes that rejects the eight-persona plan, explains why the 6-10 figure does not apply, and recommends one consolidated campaign.",
      "files": [],
      "expectations": [
        "Rejects the 6-to-10 decision makers figure as universal, stating it describes complex purchases only",
        "States that primary committee-size studies disagree by roughly a factor of two because each defines involvement differently and samples different deal sizes",
        "Derives committee size from the user's own CRM win notes (about two people) rather than from published figures",
        "Cites the SMB band of 2-3 committee members as consistent with the user's own data",
        "Recommends a two-role map - a champion-who-is-the-user (practice owner or office manager) plus an economic buyer - instead of eight personas",
        "Cuts procurement, compliance, and finance as separate roles, noting those stages do not exist in a $7K purchase, and refuses to fill empty role slots for completeness",
        "States that no published data segments committee composition by industry or purchase type, rather than extrapolating a dental-specific composition",
        "Treats the published committee-size figures as a sanity band without ranking the studies against each other",
        "Deletes any proposed role that changes no targeting, creative, or offer decision",
        "The committee summary names the evidence source behind the size estimate (the user's own win notes)",
        "Recommends one consolidated campaign or merged map rather than per-role campaigns, since the case sits below both the roughly 500 named-account and roughly $100K ACV thresholds"
      ]
    },
    {
      "id": 3,
      "prompt": "Planning our LinkedIn ads for Loopwell (revenue-intelligence platform, $45K ACV, mid-market B2B). We mapped five committee roles and I want one campaign per role, targeting exact job titles - 'Head of Revenue Operations', 'Director of Sales Enablement', 'VP Revenue Operations' - stacked with seniority filters to keep it tight. Each audience comes out around 2,500-3,500 members, which feels nicely focused. Budget is $9K/month. We also uploaded our 900-account ABM list, so those people already know who we are - I figure we can go straight to demo-request ads for them. Can you sanity-check the targeting plan and give me the messaging per role?",
      "expected_output": "A corrected targeting plan that replaces exact titles with function plus seniority on the account list, merges sub-floor audiences, treats the list as cold, and holds demo asks until warmth is earned.",
      "files": [],
      "expectations": [
        "Rejects exact job-title targeting as the primary proxy and recommends job function plus seniority instead",
        "States platforms understand only roughly 30-50% of job titles, so every title variant not listed is simply missed",
        "States function plus seniority roughly triples the addressable audience at similar engagement",
        "States that on professional networks title targeting and seniority targeting are mutually exclusive and cannot be stacked",
        "Flags the 2,500-3,500-member audiences as below the roughly 5,000 significance floor and far below the 20,000-50,000 practical floor",
        "States the audience floor before splitting roles, and merges roles whose proxy audiences collide or undershoot rather than shipping five sub-floor audiences",
        "Notes each member gets exactly one job function on professional networks, so a cross-functional role like Revenue Operations straddles cells and requires targeting two functions to cover one role",
        "Warns that stacking AND conditions below roughly 50,000 members degrades delivery",
        "Recommends account list plus function plus seniority as the default proxy given the enumerable 900-account market",
        "Notes function and seniority derive from self-reported profiles with no inference, inheriting stale or inflated titles",
        "Corrects the claim that the ABM list means people know the brand: list membership is an account signal and every individual on it is still cold",
        "Does not endorse demo-request ads as the first touch to the cold list audience",
        "Warns that a narrow audience all seeing one generic ad is worse than a broader audience seeing role variants - creative variants first, filters second"
      ]
    },
    {
      "id": 4,
      "prompt": "I'm the only marketer at Corvine Data (data-quality monitoring, $30K ACV). Our market is huge - something like 40,000 companies fit the ICP. I just read a piece on buying committees and I'm sold: I want six separate campaigns, one per committee role, each with its own creative. Monthly budget is $15K, and realistically I can produce about one ad concept per month by myself. Draw up the six-role campaign plan?",
      "expected_output": "A refusal of the six-campaign plan on threshold and capacity grounds, replaced by a merged map (one primary angle plus role-aware proof points) to a broad audience, with the precision-vs-reach dispute stated honestly.",
      "files": [],
      "expectations": [
        "Tests both precision thresholds explicitly - roughly 500 named accounts and roughly $100K ACV - and states this case clears neither",
        "Recommends the merged map, one primary angle plus role-aware proof points in a single campaign, shipped to a 20,000-50,000-sized audience, instead of six per-role campaigns",
        "Cites the out-market argument: roughly 95% of B2B buyers are out-market at any time, so hyper-narrow targeting wastes the reach that builds future demand",
        "Deletes per-role campaigns from the menu by name because a single generalist marketer cannot run six campaigns, rather than ranking them last as a stretch goal",
        "States the precision-versus-reach dispute is unresolved: no controlled test compares role-differentiated creative against a single strong message",
        "Steelmans the precision side (bounded reach in small named-account markets, high-ACV economics, activation for the in-market share) instead of dismissing it",
        "Gives effort in orders of magnitude: per-role campaigns about several weeks, the merged map about a few days",
        "Includes the flip check: tightening further is justified only if CPM inflation and frequency on the narrow audiences are not degrading cost per opportunity",
        "Deletes both list-based targeting rungs for the roughly 40,000-company market and plans on function plus seniority",
        "Directs the role knowledge into proof-point variants inside the single campaign rather than into audience splits",
        "The committee summary states which precision-vs-reach posture the map takes and why",
        "The output carries a named list of every option a stated constraint deleted from a menu",
        "States as a caveat that per-role creative outperforming a single strong message is unproven, framing the campaign KPIs as the test of which side of the dispute this market is on"
      ]
    },
    {
      "id": 5,
      "prompt": "Metriq Systems sells a $150K ACV supply-chain planning suite to enterprise manufacturers; about 300 named target accounts. Two problems: deals keep dying late when the security review shows up in month five and reopens everything, and an embarrassing share just fizzle out with no decision at all. Leadership wants an ad push that creates urgency - countdown-style 'cost of waiting' messaging aimed at CFOs and security reviewers - with everything gated behind forms so sales gets the leads. We have call recordings and win/loss notes from about 30 deals. Build the committee map and the campaign sequencing.",
      "expected_output": "A committee map with sequencing that replaces urgency with risk-reduction messaging for economic buyers and gatekeepers, reaches security before validation, and runs ungated content instead of form-gated capture.",
      "files": [],
      "expectations": [
        "Rejects urgency and countdown-style messaging for the economic buyer and gatekeepers, recommending risk-reduction and proof instead",
        "States 40-60% of deals end in no-decision and about 56% of those are lost to indecision (fear of messing up) rather than status-quo preference, which is why amplifying urgency backfires",
        "Recommends reaching security, legal, and procurement before the validation stage, noting a security reviewer added late can reopen requirements and reset agreed work",
        "Calls reaching validation-stage roles before validation the cheapest insurance in the map",
        "Models the journey as six looping buying jobs that roughly 90% of buyers revisit, not a linear funnel",
        "Rejects gating the content behind forms, citing that fewer than 30% of eventual buyers ever fill a form on the winning vendor's site",
        "Cites that roughly 70% of the journey completes before any seller contact and/or that 81-84% of deals go to the first vendor contacted, so ungated content and paid reach must carry the pre-contact phase",
        "Maps arrival order: end users and the initiator at problem identification, the economic buyer at the business case, gatekeepers at validation",
        "Cites that 79% of purchases require CFO approval when placing the economic buyer",
        "Assigns the security gatekeeper an ungated compliance/security documentation pack, delivered before the validation stage",
        "Judges per-role creative defensible under the thresholds, since about 300 named accounts and $150K ACV clear both",
        "Gates committee-wide multi-role targeting on account-level intent or engagement (surging or engaged accounts), not a cold industry-wide list",
        "Includes a sequencing-notes section mapping which roles to reach at which buying job and where risk reduction replaces urgency"
      ]
    },
    {
      "id": 6,
      "prompt": "I do marketing for Hearthside Climate, a regional HVAC installer. An average full-system replacement runs $14K. From our reviews it's clear one partner does all the research and the other one mostly worries about getting locked into a bad contract. Two questions: first, can we target each spouse separately with different ads - I heard connected-TV household graphs can do that; second, should we build the same decision-maker mapping for our $89 duct-cleaning tune-up offer? We have about 220 customer reviews to work from.",
      "expected_output": "A household map using the consumer role model, role-differentiated creative to one shared household audience instead of per-member isolation, and a refusal to apply committee mapping to the routine $89 offer.",
      "files": [],
      "expectations": [
        "Uses the consumer five-role vocabulary (initiator, influencer, decider, buyer, user) or an explicit subset of it for the household map",
        "States what carries over from organizational buying: the user often is not the approver, each decider needs a different message, and one unconvinced member can veto",
        "States at least two things that do not transfer: no procurement, legal, or security review; no career risk (the fear is personal and financial); personal money rather than company money; reviews and peers instead of RFPs",
        "Recommends role-differentiated creative shown to one shared household audience instead of trying to isolate each partner",
        "States household-graph and shared-device targeting is imprecise and unquantified",
        "Declines to build a decision-maker map for the $89 tune-up offer, stating routine purchases have one decider and the mapping is overhead",
        "Confirms the $14K system replacement qualifies as a high-consideration category with a genuine second decider",
        "Grounds the roles in the 220 reviews and labels each role evidenced or hypothesis",
        "Gives the researcher partner and the contract-worried partner distinct messaging angles, neither of which could run unchanged against the other role",
        "Does not import B2B committee statistics (such as 6-10 decision makers or CFO-approval rates) into the household map",
        "Weights the contract-worried partner's angle toward risk reduction, such as warranty terms or contract safeguards, drawn from the stated fear of lock-in",
        "Fills per-role fields including a messaging angle and a proof/offer type for each surviving household role"
      ]
    },
    {
      "id": 7,
      "prompt": "Freightlane, $60K ACV logistics-visibility platform, mid-market shippers. Our champions love us but deals stall in committee - our main contact can't seem to get the rest of the group moving. Sales wants ad campaigns aimed at 'influencers' inside the target accounts, and our CEO insists every ad ends with 'Book a demo'. We hold SOC 2 Type II already. Someone also floated producing a big industry benchmark report, though we don't have any proprietary data. We've got two years of call recordings. Map the committee with messaging and offers per role.",
      "expected_output": "A role map that rejects the influencer label, reframes the stuck-champion problem as needing a consensus-driver, replaces demo-everywhere with friction-matched offers led by the compliance pack, and deletes the benchmark report.",
      "files": [],
      "expectations": [
        "Rejects 'influencers' as a role label, stating it yields no targeting attribute and no messaging angle",
        "Distinguishes champion, coach, and mobilizer, and states a champion who cannot move the group calls for a consensus-driver, not a louder champion",
        "Rejects ending every ad with a demo CTA and matches offer friction to role warmth instead",
        "Cites the conversion bands: low-friction offers around 10-15% versus cold demo or trial requests around 1.5-4%",
        "States cold gatekeepers do not book demos at all",
        "Sets the default offer rung per role to an ungated low-friction asset, escalating only after that role's proxy audience engages twice, never on the first touch",
        "Promotes the compliance/security pack given the SOC 2 Type II already held (about an hour to assemble), noting it removes a validation-stage veto",
        "Deletes the benchmark report from the offer menu because no proprietary data exists, rather than ranking it last",
        "Notes a demo or trial is a standing job because each one spends sales capacity again",
        "Runs the swap test on angles: flags, rewrites, or merges any angle that could run unchanged against another role",
        "Sources verbatim customer language from the two years of call recordings for the messaging angles, stating exact phrases beat polished descriptions",
        "Fills all six per-role fields: measured on, personal risk, likely objection, messaging angle, proof/offer type, targeting proxy",
        "Marks any role or objection seeded from generic patterns as a hypothesis to verify against the user's own call notes, noting fine-grained per-role objection maps are weakly evidenced in published research"
      ]
    }
  ],
  "trigger_queries": [
    { "query": "Map the buying committee for our $80K ACV compliance platform selling to mid-market fintech", "should_trigger": true },
    { "query": "Who's actually involved in the decision when a hospital buys scheduling software, and what do we say to each of them?", "should_trigger": true },
    { "query": "who signs off on a $200k security tool purchase and how do I reach them with ads", "should_trigger": true },
    { "query": "We need different LinkedIn ad messaging for the CFO vs the end users - help me structure it", "should_trigger": true },
    { "query": "who else decides when a family buys solar panels and what do we say to each", "should_trigger": true },
    { "query": "My champion is sold but the deal is stuck - who else do I need to be advertising to?", "should_trigger": true },
    { "query": "build me a buying group map for our HR software", "should_trigger": true },
    { "query": "Which job functions and seniorities should we target for each member of the buying committee?", "should_trigger": true },
    { "query": "our ads only reach end users but finance keeps killing the deals - how do we get in front of the budget owner", "should_trigger": true },
    { "query": "decision makers vs influencers vs blockers in enterprise deals - how do I target each on paid social?", "should_trigger": true },
    { "query": "Should we run one broad ad campaign or separate campaigns per stakeholder role?", "should_trigger": true },
    { "query": "What's the right messaging angle for procurement and security reviewers in our ad campaigns?", "should_trigger": true },
    { "query": "husband researches the car, wife has the final say - how should our dealership ads handle that", "should_trigger": true },
    { "query": "I keep hearing B2B purchases involve 6-10 people. How do I figure out who they are for my product and reach them with ads?", "should_trigger": true },
    { "query": "map out who's in the room when a mid-market company buys payroll software", "should_trigger": true },
    { "query": "per-role ad creative for an ABM campaign - which roles and what does each need to hear", "should_trigger": true },
    { "query": "we're launching ABM ads - help me figure out which personas at each account need which message", "should_trigger": true },
    { "query": "who do I target ads at for a $500k ERP deal besides the CIO", "should_trigger": true },
    { "query": "Economic buyer vs champion - how should our ad targeting treat them differently?", "should_trigger": true },
    { "query": "how do committees buy martech and when does each stakeholder show up in the deal", "should_trigger": true },
    { "query": "what ad audiences reach a VP of Engineering vs their security team for the same account list", "should_trigger": true },
    { "query": "the IT guy loves our product but someone above him keeps blocking the purchase - how do we advertise to whoever that is", "should_trigger": true },
    { "query": "stakeholder mapping for paid media - our deals involve like 5 different people", "should_trigger": true },
    { "query": "when in the buying process should our ads hit legal and procurement?", "should_trigger": true },
    { "query": "Give each buying role its own message and tell me how to actually target them on ad platforms", "should_trigger": true },
    { "query": "how do families decide on private school and how should our ads speak to each parent", "should_trigger": true },
    { "query": "I need an ad plan that covers everyone involved in the purchase, not just our main contact", "should_trigger": true },
    { "query": "our product is used by analysts but bought by their directors - how should the ads differ", "should_trigger": true },
    { "query": "buying group targeting on LinkedIn - which functions and seniority per role?", "should_trigger": true },
    { "query": "Who are the hidden stakeholders killing our deals at the end, and can ads pre-empt them?", "should_trigger": true },
    { "query": "one message for the whole committee or tailored ads per role - what's right for a $40k ACV product?", "should_trigger": true },
    { "query": "help me identify the decision making unit for our fleet management software and craft ads for each member", "should_trigger": true },
    { "query": "DMU mapping for B2B ads", "should_trigger": true },
    { "query": "what does the CFO need to see in an ad before approving a six-figure software purchase, vs what the daily user needs", "should_trigger": true },
    { "query": "we sell to schools - superintendent, principal, IT director all weigh in. How do we split the ad messaging?", "should_trigger": true },
    { "query": "sequence ads to different committee members across the buying journey for our cybersecurity platform", "should_trigger": true },
    { "query": "my ACV is $8k - do I even need to worry about a buying committee in my ads?", "should_trigger": true },
    { "query": "targeting the economic buyer on paid social without wasting spend on everyone else", "should_trigger": true },
    { "query": "couples shopping for kitchen remodels - who do we aim ads at and with what message", "should_trigger": true },
    { "query": "roles and messaging matrix for our enterprise ABM ads, deal size $250k", "should_trigger": true },
    { "query": "Build our layered ad audience plan - cold, lookalike, retargeting tiers with budgets per tier", "should_trigger": false },
    { "query": "Which customers should we upload as the seed list for a lookalike audience?", "should_trigger": false },
    { "query": "Design a retargeting sequence for cart abandoners with frequency caps", "should_trigger": false },
    { "query": "Write 10 headline variants for our LinkedIn ads targeting CFOs", "should_trigger": false },
    { "query": "We want to promote our CEO's LinkedIn posts as paid ads - set up the campaign", "should_trigger": false },
    { "query": "Our CPA doubled last month and I don't know why - audit the ad account", "should_trigger": false },
    { "query": "Is a 3.2 ROAS good for B2B SaaS? What CAC can we afford?", "should_trigger": false },
    { "query": "Should we advertise on LinkedIn or Google Ads with our $20k budget?", "should_trigger": false },
    { "query": "How do I split $50k a month across our search, social, and retargeting campaigns?", "should_trigger": false },
    { "query": "Write a creative brief our designer can execute for the new brand campaign", "should_trigger": false },
    { "query": "Find micro-influencers to promote our skincare brand", "should_trigger": false },
    { "query": "How much should we pay an Instagram influencer for a sponsored post?", "should_trigger": false },
    { "query": "Map the org chart at this account so our AE knows who to call about the open deal", "should_trigger": false },
    { "query": "Build a battlecard on who our main competitor sells to and how they position", "should_trigger": false },
    { "query": "Which conversion events should fire before we launch the campaign?", "should_trigger": false },
    { "query": "Our ad platform reports 300 conversions but the CRM shows 90 - reconcile the gap", "should_trigger": false },
    { "query": "Define our ICP - firmographics, company size, industry filters", "should_trigger": false },
    { "query": "Write cold email sequences for the different personas at our target accounts", "should_trigger": false },
    { "query": "Score which of these five video hooks deserves test budget", "should_trigger": false },
    { "query": "Our ad sets are stuck in learning limited - should we consolidate campaigns?", "should_trigger": false },
    { "query": "Set our maximum allowable CAC and a kill-switch threshold for paid spend", "should_trigger": false },
    { "query": "When can we scale the winning campaign from $5k to $20k a month?", "should_trigger": false },
    { "query": "Build a negative keyword list from this search terms report", "should_trigger": false },
    { "query": "Which ad format should we use for a brand awareness push - carousel or video?", "should_trigger": false },
    { "query": "Write a UGC script for a customer testimonial ad", "should_trigger": false },
    { "query": "Plan the A/B test cells and budget for our new creative batch", "should_trigger": false },
    { "query": "The landing page gets clicks but no conversions - audit it", "should_trigger": false },
    { "query": "Track whether we're on pace to spend our Q4 media budget", "should_trigger": false },
    { "query": "Should we pick target CPA or maximize conversions for the new campaign?", "should_trigger": false },
    { "query": "What ads are our competitors running right now - build a swipe file", "should_trigger": false },
    { "query": "How do I get hired as a media buyer at an agency?", "should_trigger": false },
    { "query": "Interview questions for hiring our first performance marketer", "should_trigger": false },
    { "query": "Which PPC newsletters and podcasts should I follow to stay current?", "should_trigger": false },
    { "query": "Create buyer personas for our website messaging and brand voice", "should_trigger": false },
    { "query": "Who at our company should approve ad spend increases?", "should_trigger": false },
    { "query": "Segment our email list by job title for the newsletter", "should_trigger": false },
    { "query": "Help me respond to an RFP from a hospital procurement team", "should_trigger": false },
    { "query": "Build lead scoring rules so sales knows which contacts to call first", "should_trigger": false },
    { "query": "My champion at the account went silent - draft a re-engagement email", "should_trigger": false },
    { "query": "Set up sponsored answers inside AI chat assistants for our product", "should_trigger": false }
  ]
}
reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.Stars: 472kUpdated Claude CodeCodexVerifiedInstall
diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness.Stars: 323kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 323kUpdated Claude CodeCodexVerifiedInstall
executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is availableStars: 323kUpdated Claude CodeCodexVerifiedInstall