gtmskills.com
GitHubContribute
LinkedIn ads & ABM audit

Reference file

Data playbook: which ZenABM tool to call for each section

data-playbook-which-zenabm-tool-to-call-for-each-section.md

Data playbook: which ZenABM tool to call for each section

Tool names below are shown unprefixed (e.g. get_creative_performance). At runtime they carry the ZenABM connector's server prefix — match by the trailing name. Two windows matter everywhere: cur = last 30 days, prev = the 30 days before that. Pass explicit startDate/endDate (YYYY-MM-DD) so "previous period" is unambiguous, rather than relying on period for the comparison window.

Probe first: call get_linkedin_metrics for the last 30 days. If it errors or is empty, the connector isn't ready — stop and help the user finish connecting their LinkedIn ads account in ZenABM. Never invent numbers.

1. Account scorecard (spend, ads, CPC, CPM, CTR, eff-CPC-to-LP)

  • get_linkedin_metrics for cur and prev → impressions, clicks, engagements, cost. Derive CPC, CPM, CTR (see the metrics reference).
  • get_ad_spend for cur → summary.costInUsd, adSetCount, and byAdSet for the spend breakdown.
  • Landing-page clicks come from get_creative_performance (sum landingPageClicks) → account effective CTR/CPC to LP.

2. Are you running the right number of ads?

  • list_creatives with pageSize: 100, paginate via nextCursor; count currently serving ads (serving state), and note format of each. Exclude TEXT_AD from the click-budget count.
  • Feed cur spend + account CPC into the ad-count model in the metrics reference. Classify too few / about right / too many and compute per-ad daily spend + health tier.

3. Where your budget goes (format mix + spend share)

  • get_creative_performance with groupBy: "format", cur window → one row per format with summed impressions, clicks, landingPageClicks, cost, engagements. Spend share = format cost / total cost. Ad counts per format come from the list_creatives catalog in step 2.

RULE — classify every ad by its adFormat / format field, never by campaign or ad-set NAME. Users name campaigns freely, so a campaign called "TLA" can be full of Single Image ads and a "Bootcamp" campaign can be TLAs. Always read the actual adFormat (TLA, STANDARD_UPDATE=Single Image, SINGLE_VIDEO, CAROUSEL, SPONSORED_UPDATE_NATIVE_DOCUMENT=Document, TEXT_AD, SPOTLIGHT) from the creative/format data. If you mention a campaign in the report, state the format its ads actually are, not what its name implies.

4. What's influencing your deals (needs CRM/HubSpot)

  • list_deals with influenceFilter / abmCampaigns to get influenced deals and amounts. If deals return with the influencing campaigns/ad sets, map those to formats (via list_creatives / get_creative_performance) to attribute deals→format. get_company_deals can confirm which ads touched a given account's deal.
  • If list_deals is empty or CRM isn't connected: skip this section and note that connecting the CRM in ZenABM unlocks it. Don't guess influence without deal data.

RULE — quarantine test / sandbox deals; never count them as pipeline. CRM connections often include a developer or sandbox org seeded with fake opportunities, and these are usually the largest amounts, so they silently dominate "influenced pipeline." Treat a deal as test data (exclude it from headline pipeline, or show it clearly labelled "test/demo") when any of these hold: the externalUrl points at a sandbox/dev host (contains develop.my.salesforce.com, sandbox, --dev, orgfarm, test, or similar); the deal or company name is an obvious placeholder (Acme, Test, OPPORTUNITY TEST, Demo, Example); the company on the deal doesn't match the account it's filed under (e.g. an opp named "Acme Robotics" attached to company "GrowthMentor"); or a round, outsized amount ($25k/$30k/$40k) sits on an account with only a handful of real impressions. When in doubt, keep the deal in the report but flag it as likely test data and base the verdict on the real, production-CRM deals. Prefer HubSpot production deals over Salesforce dev-org rows when they conflict.

5. Benchmark grade by format

  • Reuse the groupBy: "format" result from step 3 for cur. Compute CTR, CPC, effective CTR to LP, effective CPC to LP, CPM (+ dwell if present) per format and grade each against the benchmarks reference.

6. Reallocate the budget

  • Derived from steps 3 + 5 (+ 4 if CRM). No new calls.

7. Campaign trends

  • list_campaigns for cur and prev → per-campaign impressions, clicks, engagements, cost, CPC, CTR; compute deltas and the efficiency verdict.
  • list_abm_campaigns + get_abm_campaign_overview → ABM campaign performance; combine with list_deals (abmCampaigns) to find the top ABM campaign by pipeline/deals.
  • Best campaigns / best campaign types = rank the list_campaigns rows by clicks/CTR/efficiency.
  • CTR spike/drop explained: when a campaign's CTR jumps or craters vs prev, pull its ads with get_creative_performance (campaignUrn filter, sort by clicks/impressions) and name the specific ad(s) driving it.

8. Ad insights

  • Best/worst ads by clicks: get_creative_performance cur, sortBy: "clicks", sortOrder DESCENDING then ASCENDING.
  • Best/worst formats: the groupBy: "format" result.
  • TLA effective CTR + effective spend: get_creative_performance with adFormat: "TLA" → per-TLA landingPageClicks/impressions and cost/landingPageClicks. Flag TLAs with no link (0 LP clicks) as n/a.
  • Decaying ads (needs weekly series): call get_creative_performance for ~4 consecutive weekly windows (e.g. last 4 Sundays) and track each ad's weekly CTR. Flag any ad whose CTR fell three weeks running, or that is below its format's pause threshold (see the benchmarks reference) after 1,000+ impressions. list_companies with includeWeekly: true also returns weekly buckets if you need account-level weekly trend.

9. Accounts to go after (company trends)

  • Top engaged: list_companies cur, sortBy engagements/clicks DESCENDING.
  • Surging: list_companies with includeWeekly: true — compare each company's latest week vs prior weeks; rank the biggest positive jumps in clicks/engagements.
  • Newly reached: companies present in cur with no prior-period activity (diff cur vs prev company lists).
  • Moved to Interested: list_abm_stages to find the "Interested" stage id, then get_abm_stage_companies_entering with that id + cur window. Link the reader to the filtered stage view in ZenABM if you can.
  • Impression hogs: list_companies cur, sortBy impressions DESCENDING. Only recommend impression-capping / excluding a company when its impression share is lopsided versus the rest (e.g. one account eating a large multiple of the median). Otherwise just list them.

Pagination & efficiency

  • Prefer groupBy: "format" and account-level tools over looping per-ad where possible — fewer calls, cleaner totals. Only drill to per-ad/per-campaign detail for the specific spike/decay/best-worst callouts.
  • Respect page sizes (max 100). Paginate list_creatives by cursor; paginate list_companies/list_deals by page. Stop once you have enough for the top-N lists — you don't need every row.