Skip to content

mbfinotti/advertising-skills / skills/ad-campaign-consolidation

ad-campaign-consolidation

Verified

Evaluate a fragmented paid-ads account and recommend which campaigns and ad sets to merge, which splits to preserve, and how to migrate without resetting learning across the account - it plans the consolidation, it never executes it. Use whenever the user mentions too many campaigns or ad sets, merging or restructuring an ad account, ad sets stuck in learning or learning limited, budget spread too thin, account simplification, or campaigns competing with each other - even if they never say 'consolidation'. Covers B2B and B2C on search, paid social, video, and professional-network platforms. Do NOT use to find out why the account underperforms in the first place - use mbfinotti/advertising-skills@ad-account-diagnostic instead.

FileKindSize
SKILL.mdskill md33.1 KB
evals/evals.jsonconfig23.9 KB
references/worked-invocation-example.mdmarkdown925 B

evals/evals.json

View on GitHub
{
  "skill_name": "ad-campaign-consolidation",
  "evals": [
    {
      "id": 1,
      "prompt": "I run growth at Pinemarten Goods, a DTC cookware brand. Our Meta CPA jumped from $38 to $81 over the last ten days and the CEO is furious. We have 11 ad sets; two of them re-entered the learning phase last week after our media buyer tweaked bids. The CEO wants me to merge everything into a single campaign by Friday to stop the bleeding. Can you give me the merge plan? Happy to pull whatever exports you need.",
      "expected_output": "A refusal to restructure mid cost-spike: the response declines to deliver a Friday merge plan, explains that a learning reset often costs more than the spike, routes to root-cause diagnosis first, states what evidence would clear the gate, and stays a recommendation only.",
      "files": [],
      "expectations": [
        "Declines to deliver the requested merge plan on the Friday deadline; no immediate restructuring plan is produced",
        "Names the active cost spike / panic-driven request as the specific reason to delay restructuring",
        "States that resetting learning through a restructure often costs more than the spike itself",
        "Cites the two ad sets currently in the learning phase as an additional reason not to restructure now",
        "Recommends diagnosing the root cause of the CPA jump before any consolidation work",
        "States concretely what would clear the way to a plan (specific exports or access, additional days of data, or accumulated per-branch spend) rather than a vague 'need more data'",
        "Pushes back on the single-campaign end state, warning that one undifferentiated pool lets the algorithm park on its easiest slice",
        "Frames all output as a recommendation the account owner applies; never claims to have made or scheduled account changes",
        "Asks at least one structured clarifying question before analysis (e.g. why each of the 11 ad sets was split out, or where the target CPA comes from)",
        "Does not present consolidation as a guaranteed fix for the CPA spike; treats it as a hypothesis to test"
      ]
    },
    {
      "id": 2,
      "prompt": "We're Kestrel Shield, a B2B cybersecurity SaaS ($30K ACV, 4-month sales cycle). We spend $7K/month across LinkedIn and Google Search, 11 campaigns total, optimising for demo requests, and we get about 6 demos a month. Every LinkedIn campaign shows terrible delivery and the Google ones never leave Learning. I've read that consolidating campaigns fixes this. How should we combine ours so the algorithms finally get enough data?",
      "expected_output": "A plan that runs the merged-structure arithmetic first, says plainly that 6 conversions/month cannot clear any threshold even fully merged, promotes moving the optimisation event up-funnel to a higher-volume proxy as the first move, offers manual bidding as the alternative with an explicit value/effort ordering, and uses only each platform's documented thresholds.",
      "files": [],
      "expectations": [
        "States that even the fully merged structure at roughly 6 conversions/month cannot clear the platform learning threshold",
        "Says plainly that consolidation alone will not fix this account; merging cannot create conversions that do not exist",
        "Recommends moving the optimisation event up-funnel to a higher-volume proxy as the first move, ahead of pruning or merging",
        "Names concrete proxy events the platform can actually observe (e.g. qualified visit, form open, MQL)",
        "Requires feeding CRM outcomes back to keep the proxy event honest against real pipeline outcomes",
        "Presents manual or volume-independent bidding, judged on leading metrics, as the alternative to the proxy-event move",
        "Ranks the proxy-event move above manual bidding on value (it restores an optimisation signal; manual bidding only stops pretending there is one) while acknowledging its higher effort",
        "Names the condition for choosing manual bidding first (no tracking/CRM owner to coordinate with, or an effort ceiling that rules the proxy work out)",
        "States that LinkedIn documents no learning-phase conversion threshold, only minimum spend levels (around $10/day and $100 lifetime)",
        "Does not apply a 50-conversions-per-week or Meta-style 50-per-7-days threshold to LinkedIn",
        "Grounds the Google side in a documented figure (e.g. 15 conversions/30 days for Search) or explicitly flags that the live help page must be re-verified; imports no undocumented threshold",
        "Does not promise that combining campaigns will by itself exit Learning or fix delivery"
      ]
    },
    {
      "id": 3,
      "prompt": "Loamside Home Goods here, roughly $22K/month across Google and Meta. Our new consultant wants us to merge everything, brand search, non-brand search, prospecting, retargeting, into one campaign with a shared budget, and judge it on account-level ROAS after 30 days. For context: brand search converts at about $9, non-brand at $52; retargeting reports 8.4x ROAS, prospecting 1.7x. Sounds efficient to me. Anything we should watch out for before we do it?",
      "expected_output": "A rejection of the full merge: brand and non-brand stay in separate campaigns with distinct bid strategies, prospecting and retargeting stay structurally separate, pooled budgets are predicted to concentrate into the cheapest branches, and the verdict comes from incrementality measurement against the account's own baseline rather than 30-day account ROAS.",
      "files": [],
      "expectations": [
        "Rejects merging brand and non-brand search under one campaign and one bid strategy",
        "Recommends keeping brand and non-brand in separate campaigns with distinct bid strategies",
        "Recommends excluding brand terms from any blended or automated campaign type",
        "Predicts that a pooled budget concentrates roughly 70-90% of spend into the structurally cheapest branch or two",
        "Names brand search and/or retargeting as the branches that will eat the pooled budget, given the $9 vs $52 cost gap",
        "Warns that blended ROAS will look healthy while non-brand growth starves and the account goes blind where it matters",
        "Keeps prospecting and retargeting structurally separate rather than pooled into one entity",
        "States that reported retargeting ROAS overstates true incremental ROAS, citing a magnitude around 40-70%",
        "Recommends judging the consolidation by incrementality rather than platform-reported ROAS",
        "Recommends a platform lift study as the default verdict method",
        "Names the escalation condition for a geo holdout (a large share of the account's budget moved, or the platform grading its own homework being exactly what is disputed)",
        "Restricts any budget pooling to comparable cost-per-conversion branches, with minimum/maximum spend floors where a strategically required branch would starve",
        "Rejects the consultant's 30-day account-ROAS success criterion in favour of comparing against the account's own pre-merge baseline, not industry benchmarks"
      ]
    },
    {
      "id": 4,
      "prompt": "Rivertown Outfitters (outdoor gear, Meta only, about $40K/month). We're going from 8 campaigns to 3. The plan my team wrote: next weekend we edit the 3 keepers in place with new targeting and switch their optimisation event from add-to-cart to purchase, move 100% of the budget over on Sunday, then delete the other 5 on Monday to keep the account tidy. It's November 15th and we want this done before Black Friday. Sanity-check the plan?",
      "expected_output": "A rewrite of the migration: in-place edits are flagged as learning-resetting, merged campaigns get built new alongside, retired campaigns are paused never deleted, the pre-peak timing is rejected, the migration is phased with a written no-touch window and pre-defined rollback trigger, budget moves in small steps, and the relearning dip is forecast.",
      "files": [],
      "expectations": [
        "Flags that editing targeting and the optimisation event inside the live keeper campaigns resets their learning",
        "Recommends building the 3 merged campaigns new alongside the existing structure instead of editing keepers in place",
        "Instructs pausing, never deleting, the 5 retired campaigns",
        "States that deletion destroys learning history and reporting, or explicitly converts the delete request into a pause-or-archive decision",
        "Notes that pausing does not reset learning, keeping the old structure available for rollback",
        "Rejects executing the restructure in the run-up to Black Friday; recommends off-peak timing because the relearning dip costs most at peak volume and auction prices",
        "Phases the migration one merge group at a time, highest-confidence merges first, rejecting the single-weekend cutover",
        "Requires a written, pre-committed no-touch window signed off before phase one",
        "Specifies the no-touch window length: at least 7 days as a floor, around 14 for a reliable read, up to 30 or a full cycle for the verdict",
        "Explicitly forecasts a transient performance dip while the merged campaigns relearn",
        "Moves budget in steps of roughly 20% or less every few days, rejecting the 100%-on-Sunday shift",
        "Keeps a control during migration (legacy structure at reduced budget, or the platform's experiment tooling)",
        "Defines a rollback trigger before phase one, e.g. blended cost per conversion worse than roughly 1.5x baseline after the no-touch window matures"
      ]
    },
    {
      "id": 5,
      "prompt": "Fernbox is a B2C subscription snack box on Meta. Target CPA is $40, derived from our contribution margin. Over the last 30 days: 'Broad prospecting' spends $180/day and got 41 purchases. 'Interest stack 2023' spends $45/day and got 6 purchases, though our buyer switched its attribution from 7-day click to 1-day click 12 days into the window. 'Premium tier' spends $70/day and got 11 purchases; that product carries about 3x the margin of the standard box so it has its own target. 'Spring 2024 launch' spends $25/day and got 0 purchases, and honestly nobody remembers why it's still separate. 'Podcast listeners test' spends $3/day and got 0 purchases. Which of these should we merge and which should we keep?",
      "expected_output": "A per-branch classification using the four states: the budget floor near $286/day is computed and applied independently of the volume test, the $3/day test is insufficient evidence rather than merge, the 2024 leftover is the archaeological prune/merge, the premium tier keeps its split on unit economics with its cost named, the attribution-switched branch's counts are refused as-is, and the target structure concentrates budget until surviving branches clear their floors.",
      "files": [],
      "expectations": [
        "Computes the per-branch daily budget floor from target CPA x exit threshold / threshold period, landing near $286/day for a $40 CPA on Meta's ~50 events per 7 days",
        "Runs the budget feasibility test as a separate check from the volume test, noting even Broad prospecting at $180/day sits under the floor despite being the biggest branch",
        "Uses Meta's documented ~50 optimisation events per ad set per rolling 7 days as the volume threshold (a note to re-verify the live help page is acceptable)",
        "Classifies the Podcast listeners test as insufficient evidence rather than merge, because its ~$90 spend is below roughly 3x the $40 target CPA, so zero conversions is plausibly noise",
        "States what would clear the Podcast listeners verdict: more days, concentrated budget, or accumulated spend past the ~3x-CPA mark",
        "Treats the Spring 2024 launch's zero conversions as real evidence, since its ~$750 spend comfortably exceeds 3x the target CPA",
        "Identifies the Spring 2024 launch as an archaeological split (nobody can justify it) and makes it the first, cheapest prune or merge",
        "Keeps the Premium tier split, citing the unit-economics preserve rule (3x margin, its own target CPA)",
        "Names the cost of keeping the Premium tier split: its own budget floor or the learning-signal price of the separation, stated in the plan",
        "Flags the mid-window attribution change on Interest stack 2023 and refuses to count its conversions alongside the others without re-windowing or excluding it",
        "Assigns every branch exactly one of the four states (merge, keep, keep but re-parent, insufficient evidence) with a stated justification",
        "Requires each surviving branch in the target structure to clear its budget floor, concentrating spend into fewer branches",
        "Applies prune mechanics to the unjustified branch: pause it and move its budget to survivors in steps of roughly 20%",
        "Does not propose a whole-account big-bang rebuild"
      ]
    },
    {
      "id": 6,
      "prompt": "I'm the marketing manager at Brightpath, a B2B HR software company ($15K/month across paid social and search). On the social side we run 9 ad sets, one per buyer persona: HR director, CFO, IT admin, and so on, each with about a dozen stacked interest and job-title filters, all serving the same product demo ad. On Google we're holding off on Target CPA because we haven't hit 15 conversions in 30 days yet, and I've been told we should raise each ad set's daily budget to 10x CPA before merging anything. One wrinkle: our co-marketing contract with Medlow guarantees $2,000/month of ad spend on their line, and our CMO reviews spend and leads per persona every week. I want to consolidate but I don't want to lose control over who sees which message. What's the plan?",
      "expected_output": "A plan that merges the persona ad sets and moves segmentation into per-persona creative variants, debunks the 15-conversion Target CPA gate and the 10x-CPA budget rule as folklore, keeps the contractual $2,000 line on its own budget with finance/client approval required to change it, and preserves per-persona reporting through a re-parent arrangement rather than nine structural splits.",
      "files": [],
      "expectations": [
        "Recommends merging the 9 persona ad sets rather than preserving each as a structural split",
        "Moves the segmentation into creative: one creative variant per persona inside the merged structure, with delivery matching variant to viewer",
        "Names the current setup as the anti-pattern: generic identical creative propped up by a stack of narrow targeting filters, producing a small audience that all see the same undifferentiated ad",
        "Advises spending the persona knowledge on the message rather than on the targeting filter",
        "Keeps the contractually guaranteed $2,000/month Medlow line on its own per-branch budget",
        "States that campaign-level budget pooling will not honour the contractual spend guarantee, and that removing the guarantee needs explicit finance or client approval",
        "States that creative variants cannot substitute for structural separation where the budget mandate applies",
        "Preserves the CMO's weekly per-persona reporting through a keep-but-re-parent arrangement (separate reporting under a pooled budget and bid strategy) or verifies the report survives the merge before merging",
        "Debunks the 'raise daily budget to 10x CPA before merging' rule as a practitioner extrapolation, not platform documentation",
        "Debunks the Target CPA gate: the platform's own documentation says the strategy works with no conversion history, and the 15-conversion figure is an evaluation sample, not a prerequisite",
        "Does not advise waiting to accumulate a conversion minimum before using automated bidding",
        "Tests each split against preserve rules and treats 'keeping control' alone as comfort, not a business reason that protects a split",
        "Distinguishes the deliberate split (the contract-mandated Medlow line) from the unprotected persona splits"
      ]
    }
  ],
  "trigger_queries": [
    { "query": "We have 40 campaigns in our Google Ads account and I think it's way too many. Help me figure out what to merge", "should_trigger": true },
    { "query": "Our ad sets are all stuck in learning limited, should we combine some of them?", "should_trigger": true },
    { "query": "How do I consolidate my Meta campaigns without resetting learning?", "should_trigger": true },
    { "query": "The agency left us a spaghetti account: 25 campaigns with tiny budgets each. Where do I start simplifying?", "should_trigger": true },
    { "query": "Is it a bad idea to merge our brand and non-brand search campaigns into one?", "should_trigger": true },
    { "query": "Our ad budget is spread too thin across ad sets. How should we restructure?", "should_trigger": true },
    { "query": "Two of our campaigns seem to be bidding against each other for the same audience, should we combine them?", "should_trigger": true },
    { "query": "Plan a migration from 18 campaigns down to 5 without tanking performance", "should_trigger": true },
    { "query": "When does it make sense to keep separate campaigns per country versus one global campaign?", "should_trigger": true },
    { "query": "My boss wants everything in one single campaign. Should I push back or just do it?", "should_trigger": true },
    { "query": "We get 30 conversions a month split across 12 ad sets and none of them ever exit learning. What now?", "should_trigger": true },
    { "query": "Simplify our LinkedIn ads account structure", "should_trigger": true },
    { "query": "Should each product line really have its own campaign, or should we combine them?", "should_trigger": true },
    { "query": "How do I merge two ad sets without losing the audience segmentation?", "should_trigger": true },
    { "query": "What's the right way to restructure a fragmented PPC account?", "should_trigger": true },
    { "query": "Is a Hagakure-style restructure worth it for our search account?", "should_trigger": true },
    { "query": "Do we have too many campaigns for our budget?", "should_trigger": true },
    { "query": "We keep launching a new campaign for every promo and now nothing spends properly", "should_trigger": true },
    { "query": "Each of our 9 personas has its own ad set. Is that hurting delivery?", "should_trigger": true },
    { "query": "Should we combine our TikTok ad groups so they can finally exit learning?", "should_trigger": true },
    { "query": "The account has campaigns from three former managers and half of it nobody understands. Clean-up plan?", "should_trigger": true },
    { "query": "Which splits are worth keeping when collapsing an over-segmented ads account?", "should_trigger": true },
    { "query": "Ad account simplification best practices", "should_trigger": true },
    { "query": "We merged campaigns last month and performance dipped. Did we break it, and should we roll back?", "should_trigger": true },
    { "query": "How many campaigns should a $10K/month ad account actually run?", "should_trigger": true },
    { "query": "Power 5 says simplify the account. How far should we take it?", "should_trigger": true },
    { "query": "Should we fold our retargeting ad sets into prospecting to pool the budget?", "should_trigger": true },
    { "query": "Our CPA doubled and someone said fewer campaigns fixes the learning phase. True?", "should_trigger": true },
    { "query": "consolidate google ads campaigns", "should_trigger": true },
    { "query": "Draft a plan to collapse 60 single-keyword ad groups into a few themed campaigns", "should_trigger": true },
    { "query": "Everything in our account shows learning limited. Restructure or wait it out?", "should_trigger": true },
    { "query": "Our client mandates separate budgets per product line but the account is starving. What are our options?", "should_trigger": true },
    { "query": "Should we rebuild the whole ad account from scratch in one go or merge things gradually?", "should_trigger": true },
    { "query": "Too many ad sets and not enough conversions to feed them all. What would you cut or combine?", "should_trigger": true },
    { "query": "Can I merge campaigns mid-quarter without resetting the algorithm?", "should_trigger": true },
    { "query": "We run 14 search campaigns and 9 paid-social ad sets on $9K/month with about 40 conversions. What should we merge?", "should_trigger": true },
    { "query": "Why did our CPA double last month?", "should_trigger": false },
    { "query": "Our ads stopped converting. Figure out what's wrong with the account", "should_trigger": false },
    { "query": "Meta says 120 conversions but our CRM shows 70. Reconcile the numbers", "should_trigger": false },
    { "query": "Choose between Target CPA and maximize conversions bidding for our new campaign", "should_trigger": false },
    { "query": "We're overspending against our monthly ad budget. Flag pacing issues before month end", "should_trigger": false },
    { "query": "How should I split $50K/month between Google and Meta?", "should_trigger": false },
    { "query": "Build me a layered targeting plan with prospecting, lookalikes and retargeting pools", "should_trigger": false },
    { "query": "Design a retargeting sequence with recency windows for cart abandoners", "should_trigger": false },
    { "query": "Write 10 ad copy variants for our spring sale", "should_trigger": false },
    { "query": "Is our hero video ad fatigued or is it just seasonality?", "should_trigger": false },
    { "query": "Set up a creative test matrix with per-cell budgets and kill rules", "should_trigger": false },
    { "query": "Which negative keywords should we add from this search terms report?", "should_trigger": false },
    { "query": "Verify our conversion pixel fires once and deduplicates before launch", "should_trigger": false },
    { "query": "What's a healthy ROAS for a DTC skincare brand?", "should_trigger": false },
    { "query": "When can we scale our winning campaign's budget, and how fast?", "should_trigger": false },
    { "query": "Which ad platforms fit a $3K/month B2B budget?", "should_trigger": false },
    { "query": "Map the buying committee for our data-warehouse product", "should_trigger": false },
    { "query": "Pick the seed customer list for a lookalike audience", "should_trigger": false },
    { "query": "Audit the landing page our ads send traffic to", "should_trigger": false },
    { "query": "Set our maximum allowable CAC and a kill-switch threshold", "should_trigger": false },
    { "query": "Merge duplicate contacts in our CRM", "should_trigger": false },
    { "query": "Consolidate our marketing tech stack, we have too many overlapping SaaS tools", "should_trigger": false },
    { "query": "Combine our three email nurture sequences into one", "should_trigger": false },
    { "query": "Merge our two Google Analytics properties into one", "should_trigger": false },
    { "query": "Should we consolidate our five brand Instagram accounts into one handle?", "should_trigger": false },
    { "query": "Merge our Google Ads account into our new agency's MCC manager account", "should_trigger": false },
    { "query": "Restructure our marketing team, we have too many specialists per channel", "should_trigger": false },
    { "query": "Combine departmental budgets into one marketing budget for next year", "should_trigger": false },
    { "query": "Consolidate three overlapping SEO keyword clusters into one pillar page", "should_trigger": false },
    { "query": "Our Salesforce has duplicate campaigns from two imports. Clean them up", "should_trigger": false },
    { "query": "Write a creative brief for the new UGC campaign", "should_trigger": false },
    { "query": "Which ad format should we use for a B2B awareness push?", "should_trigger": false },
    { "query": "Score the hooks in these three video ad scripts", "should_trigger": false },
    { "query": "Promote our CEO's LinkedIn post as a thought leader ad", "should_trigger": false },
    { "query": "Build a swipe file of our competitors' ads", "should_trigger": false },
    { "query": "How do I break into paid media as a career?", "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