Marketing Analytics Monthly Review

Open GA4 DebugView and walk through each marked-conversion event end-to-end. The classic failure: a 'sign_up' event firing on email-blur instead of submit, inflating reported conversions 3-5x. Compare event counts to backend records (CRM new-contact count, Stripe checkout count) for a sanity check.

Pull the source/medium/campaign breakdown for the past 30 days. Flag any rows where source/medium drift from the documented convention — 'fb' vs 'facebook', 'email' vs 'newsletter'. Drift makes attribution reports uncomparable across months.

Load the site in an incognito window with a fresh consent state. Confirm Meta Pixel, LinkedIn Insight Tag, and any analytics tags do NOT fire before the OneTrust/Cookiebot prompt is accepted. Tag-management gating via consent state is the only correct pattern under GDPR/CPRA.

Compare HubSpot/Marketo contact counts and lifecycle stage transitions to Salesforce records for the past month. Investigate any sync errors, queued records, or field-mapping mismatches before they corrupt downstream attribution.

Export from Google Ads, Meta Ads Manager, LinkedIn Campaign Manager, and any active DSP (Trade Desk, DV360). Normalize to the same date range; platforms have different attribution windows by default and will not match without alignment.

Use the benchmark sheet — last quarter's blended CPA and channel-level ROAS targets. Flag any campaign more than 20% off target for review in the next step.

Cancelled ad sets keep serving for hours after pause — pause AND set frequency cap to 0 AND drop bid to floor as a belt-and-suspenders. Note any ad sets that need creative refresh vs. full sunset.

Send the pause list with CPA/ROAS context to the buyer (or agency contact). They need it before the next bid-pacing meeting so freed budget reroutes immediately rather than sitting idle.

Log every test that reached significance this cycle in the experiment registry — hypothesis, variant, lift, p-value, decision. Inconclusive tests still go in the registry so they are not re-run blindly next quarter.

Pull MQL → SQL conversion rate by source-medium and by campaign. A channel with cheap MQLs but a poor SQL rate is a quality problem, not a volume win — surface both rates side-by-side.

Use the GA4 Explore funnel exploration with the standard landing → product → pricing → signup steps. Compare to the prior month; investigate any step with a >5pp drop-off change.

Aggregate G2, Trustpilot, App Store reviews, and support-ticket NPS comments through your sentiment tool (Sprinklr, Sprout, or a simple GPT-based classifier). Pull the top three themes for the customer-insights summary.

Open Hotjar/FullStory and filter to rage-clicks and dead-clicks on the highest-traffic landing pages. Flag any UX issue that recurs across 3+ sessions for product-marketing handoff.

Use the position-based or data-driven model in HubSpot/Marketo, not last-touch. Last-touch undervalues TOFU content and overvalues branded search; the position-based view is closer to truth for B2B cycles longer than 30 days.

CAC ÷ monthly gross margin per customer. Anything past 18 months is a red flag for a SaaS business; for ecomm, target sub-3-month payback. Present alongside MER (marketing efficiency ratio) for a check.

Pull sourced + influenced pipeline by campaign from the CRM. Sourced answers 'who originated this opp'; influenced answers 'which campaigns touched it'. Both matter; report both, do not pick one.

If a high-spend channel (brand search, retargeting) shows strong attribution numbers but unclear true lift, an incrementality test belongs on the roadmap. Geo holdout is the cleanest design for paid social and display.

Specify the test channel, holdout DMAs, test duration (typically 4-6 weeks), pre-period baseline, and the lift KPI. Loop in analytics for power calculations before launch — underpowered tests are the most common reason incrementality reads come back null.

Move budget in 10-15% increments, not all at once — channel ROAS curves are not linear at scale, and over-funding a winner is a common way to crater its CPA. Document the new monthly allocation with reasoning.

Each experiment needs a hypothesis, success metric, sample-size estimate, and runtime. Vague experiments produce vague reads. Capture them in the experiment registry before kickoff.

Refresh the Looker/Tableau/Mode dashboard the leadership team reviews. Confirm every tile matches a number in this checklist run — discrepancies between the dashboard and the analyst's working numbers destroy trust quickly.