Growth-Today/claude-skills / gt-linkedin-ads
gt-linkedin-ads
VerifiedGT LinkedIn Ads — expert LinkedIn Ads strategist for B2B companies, built by Growth Today (growthtoday.co). Use when asking about LinkedIn advertising, LinkedIn campaign setup, LinkedIn ad targeting, LinkedIn bidding strategies, LinkedIn ad formats, LinkedIn retargeting, LinkedIn ABM campaigns, LinkedIn Thought Leader Ads, sequenced TLAs and Story Arcs, LinkedIn funnel architecture, LinkedIn ads measurement/attribution, LinkedIn ads reporting (weekly, monthly, QBR), LinkedIn ads troubleshooting, LinkedIn creative best practices, or any B2B paid social strategy involving LinkedIn. Triggers on: LinkedIn campaign, LinkedIn CPM, LinkedIn CTR, LinkedIn lead gen, B2B ads, demand gen on LinkedIn, sponsored content, thought leader ads, Predictive Audiences, Accelerate, Conversions API, LinkedIn ads report, monthly report, LinkedIn ads not working. Do NOT use for LinkedIn organic content (use gt-linkedin-content) or LinkedIn outbound messaging (use gt-linkedin-outbounder). More GT skills: growthtoday.co.
resources/references/reporting-playbook.md
View on GitHub# LinkedIn Ads Reporting Playbook
The recurring-report reference: what to pull, how to define each metric, how to grade a period, and what a report must refuse to claim. Pair with `benchmarks.md` for the numbers and `measurement-attribution.md` for the tracking that produces them.
---
## 1. The reporting contract
Settle these before pulling data. A report written without them will be argued about instead of acted on.
| Item | Why it matters |
|------|----------------|
| Audience | In-house exec wants the verdict; a client wants the work shown; a board wants spend-to-pipeline only |
| Cadence | Determines which sections are legitimate (see §5) |
| Date window + comparison window | Equal-length windows, same weekday count |
| Objective per campaign | A brand-awareness campaign graded on CPL is a broken report, not a bad campaign |
| Currency and spend | Benchmarks are USD; state the conversion if not |
| Targets | Without stated targets a report grades against benchmarks only. Say so explicitly |
| Data access | What exists: platform export, ZenABM, CRM, native post analytics |
---
## 2. Data sources and pull order
1. **ZenABM MCP** — `https://app.zenabm.com/api/mcp`. Per-ad and per-format metrics, account-level ABM views, and the benchmark dataset the grading in `benchmarks.md` is anchored to. Preferred when connected: it removes the manual export step and makes the report automatable.
2. **LinkedIn Campaign Manager export** — ad-level CSV over the window. Ask for the export, not screenshots; screenshots lose the columns that matter (landing-page clicks, dwell, per-format splits).
3. **CRM** — HubSpot or Salesforce for leads, MQL/SQL, opportunities, pipeline. Loop and model details in `crm-attribution.md`.
4. **Native post analytics** — required for Thought Leader Ads. Ads Manager under-reports follower and save actions on TLAs; the creator's own post analytics is the control source. Where the two disagree, name both numbers and say which one you trust.
**Non-negotiable:** a metric that cannot be sourced is reported as `n/a` with the reason. Never interpolate, never estimate to fill a table, never carry a number forward from a previous period as if it were current.
---
## 3. Metric definitions
Ambiguity here is where most reporting arguments come from. Define once, use consistently.
| Metric | Definition | Trap |
|--------|-----------|------|
| Spend | Platform-billed spend in the window | Billing lag on the last 1-2 days of a window |
| Impressions | Served impressions | Not reach |
| Reach / frequency | Unique members / impressions per member | Target ~3 impressions/week/person; above ~5-6 expect fatigue |
| Clicks (total) | All chargeable clicks: landing page, social actions, profile, company page, carousel swipes | Inflates CTR heavily on TLAs and carousels |
| Landing-page clicks | Clicks that actually reached the destination URL | The only click that can convert |
| CTR | Total clicks / impressions | Compare per format only; see grading rule below |
| CPC (raw) | Spend / total clicks | Flattering on formats with high social-action volume |
| **Effective CPC-to-LP** | **Spend / landing-page clicks** | **The honest cost of traffic. Use this for every link-driving format** |
| CPM | Cost per 1,000 impressions | The real lever on awareness campaigns |
| Dwell time | Seconds spent on the ad | Best single read of whether the body copy held attention |
| Leads / CPL | Platform or CRM lead count; spend / leads | State which source. Platform lead-gen-form leads and CRM leads are different numbers |
| MQL / SQL | CRM-defined only | Never platform-derived |
| Pipeline | CRM opportunity value influenced in window | Quarterly metric, not monthly (§5) |
| Pipe-to-spend | Influenced pipeline / spend | The only number some execs read. Give it a window and a model |
---
## 4. Grading rules
- **Grade per format, never blended.** Median CTR runs from 2.68% (Thought Leader Ads) to 0.02% (Text Ads) in the same dataset. A blended account CTR grades format mix, not performance.
- **Link-driving formats are graded on effective CPC-to-LP, not raw CPC.** Derived from the LP-clicks-per-$1K figures in `benchmarks.md`:
| Format | Median CPC (raw) | LP clicks per $1K | Effective CPC-to-LP |
|--------|-----------------|-------------------|---------------------|
| Thought Leader Ads | $2.29 | 327 | ~$3.06 |
| Single Image Ads | $13.23 | 71 | ~$14.08 |
| Video Ads | $15.61 | 62 | ~$16.13 |
TLAs stay the cheapest route to a landing page, but the gap narrows once non-LP clicks are stripped out. Report both columns so nobody is surprised later.
- **CTR is not a pipeline proxy.** Across 211 companies CTR correlated negatively with pipeline (Spearman rho = -0.170). A CTR improvement is a delivery finding, not a business result. Never headline a report with it.
- **Grade against the campaign's own objective.** Awareness on dwell and CPM, mid-funnel on landing-page conversion, education on view/completion rate, retargeting on CPM, bottom-funnel on cost per opportunity.
- **Never grade an account down for not adopting an optional or beta feature.**
- **Minimum sample before reading a change:** ~$100 spend per creative, ~1,000 impressions per ad set, and a full 7 days outside the learning phase. Below that, report the number and label it insufficient.
- **Normalise period-over-period.** Compare cost and rate metrics, not raw totals, when spend or day count differ. A 40% impression rise on a 40% budget rise is not an improvement.
---
## 5. What each cadence is allowed to claim
| Cadence | Legitimate sections | Do not claim |
|---------|--------------------|--------------|
| **Weekly** | Delivery hygiene: spend pacing, frequency, delivery drops, disapprovals, obvious breakage | Creative verdicts, audience verdicts, anything about pipeline |
| **Monthly** | Format performance, audience performance, creative winners and losers, leads and CPL, budget reallocation | A pipeline verdict on a 30-day window |
| **Quarterly** | Everything, plus pipeline, pipe-to-spend, attribution review, incrementality reads | Certainty. State the model and its window |
B2B cycles run 3-6 months. Monthly reporting exists to manage delivery and creative; quarterly reporting exists to judge the channel.
---
## 6. Report structure
1. **Exec summary** — verdict first, 3-5 sentences. What happened, whether it met target, the one decision being asked for.
2. **Spend & delivery** — spend vs planned, pacing, impressions, reach, frequency, anything that broke.
3. **Performance by format** — per-format table with the benchmark column beside the actual.
4. **Performance by audience** — audience or account segment, with match rates and exclusions in effect.
5. **Creative winners and losers** — top and bottom by effective CPC-to-LP and dwell, with the copy or angle named so the pattern is reusable.
6. **Conversions & pipeline** — leads, CPL, CRM stages; pipeline only on a quarterly report. Include the attribution-gap sentence.
7. **Benchmark grading** — each objective graded against `benchmarks.md`, with the format and geography the benchmark applies to.
8. **Decisions for next period** — each with owner, expected effect, and how it will be measured. No more than five.
Close with **"What this report cannot tell you"**: unmeasured view-through impact, the 70-80% of conversions the platform does not capture, cross-device gaps, and any `n/a` in the tables.
---
## 7. Reporting traps
- Headlining a blended CTR improvement.
- Quoting raw CPC on Thought Leader Ads without the LP-click column.
- Treating platform lead-gen-form leads and CRM leads as the same count.
- Declaring a topic or creative dead on under $100 of spend.
- Comparing an unequal window without normalising.
- Reading pipeline monthly.
- Presenting a vendor-claimed performance figure as measured. Label claims as claims.
- Silently dropping a campaign from the table between periods. If it stopped, say when and why.
---
## 8. Automating the monthly report
A monthly report is a good scheduled-task candidate once the data pull is programmatic (ZenABM MCP or a stable CSV export path):
- Fixed window: previous calendar month, compared against the month before.
- Refuse to generate on missing data rather than filling gaps — an automated report that estimates is worse than one that fails loudly.
- Keep a running snapshot of prior periods so period-over-period does not depend on re-pulling history.
- Human review before it reaches a client or a board. Automate the assembly, not the verdict.
---
*Benchmark dataset: ZenABM 2026 (median across 161,256 ads / 211 companies), as used in `benchmarks.md`. Metric definitions follow LinkedIn Campaign Manager reporting fields. Accessed 2026-08-31.*
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| 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: 472k | Updated | Claude CodeCodex | Unverified | Install |
| 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: 322k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| 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 available | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |