Customer Behavior Analysis Checklist

Export the prior 90 days of sessions, orders, and refunds from GA4 and Shopify Analytics. If you run on Amazon as well, pull a Business Reports export (Detail Page Sales and Traffic by ASIN) so the marketplace side of the funnel is in the dataset rather than missing.

Compare last-click GA4 numbers against MTA / first-click views in Triple Whale, Northbeam, or Hyros. Post-iOS-14 last-click chronically under-credits Meta; if the gap between platforms is over 20%, flag it before drawing conclusions about channel mix.

Filter out internal IPs, known bot user-agents, and duplicate orders from payment retries. A common gotcha: subscription dunning retries inflate order counts in raw Shopify exports — dedupe by transaction ID before computing AOV.

Verify the OneTrust / Cookiebot / Termly export reflects consent state for the dataset window. EU/UK and California sessions without analytics consent should be excluded from cohort modeling, not silently included.

Bucket by paid social, paid search, organic, email, SMS, direct, and marketplace. Tag first-order vs. repeat sessions. Cohort by first-order month so retention curves are measurable later in the workflow.

Use 90-day, 180-day, and 365-day LTV windows by acquisition cohort. Lifetimely or Polar Analytics can do this natively against Shopify; compute net of refunds and discounts, not gross.

Standard RFM tiers: Champions, Loyal, Potential Loyalists, At Risk, Hibernating, Lost. Klaviyo's predictive analytics block scores these natively if you have at least five orders per profile in scope.

Pull the Recharge or Smartrr cancellation report and filter for failed-payment churn vs. voluntary cancellation. Failed-payment churn is recoverable via Stripe card-updater and smart retry; voluntary cancellation needs a retention message.

Awareness, Consideration, First Purchase, Repeat, Advocate. Use behavior signals (browsed-but-didn't-add, added-but-abandoned, post-purchase, review-left) to bucket — not survey self-reports.

Cross-tabulate fast-mover SKUs with cohort behavior to identify which audiences pull which products. This feeds creative direction in Section 3 and the merchandising decisions at the end of the workflow.

Pull funnel metrics from Shopify Analytics: session → product view → add to cart → checkout → purchase. A drop greater than 60% between checkout-initiated and completed signals a checkout problem (shipping cost, payment options, mobile bug) rather than a top-of-funnel one.

Break abandonment by checkout step: contact → shipping → payment. Shipping-step abandonment usually means surprise cost; payment-step abandonment usually means missing wallet (Apple Pay, Shop Pay) or a card-decline UX.

Pull open, click, and placed-order rates for welcome, browse-abandon, cart-abandon, and post-purchase flows. Welcome flow placed-order rate under 5% on a healthy DTC list usually means the offer or product mix in the welcome series needs a refresh.

Look at MER (total revenue / total ad spend) and per-campaign ROAS, but cross-check against Northbeam or Triple Whale rather than relying on platform-reported ROAS alone. Watch for budget exhausted before noon — daypart caps may be required.

Filter Clarity (or Hotjar) for rage clicks, dead clicks, and excessive scrolling on PDPs. Watch ten replays of mobile checkout abandons; the qualitative signal often beats funnel numbers for diagnosing UX issues.

Pull the prior 90 days of reviews from Yotpo, Stamped, or Junip. Tag by category — sizing, quality, shipping, packaging, expectation mismatch. A spike in any single tag against the prior period is a signal for the product team.

Use Gorgias (or Re:amaze, Zendesk) tag reports to count tickets per intent: WISMO, sizing, returns, defect. WISMO above 30% of total volume usually means tracking or shipping ETA messaging needs work upstream.

Trigger NPS 14 days post-delivery, not at order placement. For Amazon orders, do not use buyer-seller messaging to solicit reviews — Amazon ToS restricts this; use Amazon's Request a Review button or the Vine program instead.

Pull mentions from Sprout, Brandwatch, or a manual Instagram/TikTok scan. Track sentiment direction week over week; a single viral negative post can move the average more than the cumulative ticket data suggests.

Synthesize from reviews, tickets, NPS verbatims, and social. Capture the three highest-impact pain points and indicate whether any rises to a product-team escalation (defect pattern, sizing, mislabeled item).

Use Klaviyo's predictive churn risk score, or a Recharge / Smartrr-driven model, to bucket subscribers as low / moderate / high risk. Captured here so retention triggers downstream can branch on the result.

Use Inventory Planner, Cogsy, or your OMS forecasting (NetSuite, Cin7) to project velocity. Layer in known events — a paid push, a Meta launch, Q4 seasonality — and apply a conservative lead-time buffer for any SKU with overseas manufacturing.

Run an A/B price test in VWO or Optimizely on a top-five SKU. Hold the test long enough for statistical significance (typically 2-3 weeks at typical DTC traffic). Watch contribution margin, not just conversion rate — a higher CR at lower price can still be margin-negative.

Build a Klaviyo + Postscript flow targeting the high-risk segment with a value reminder, not just a discount. For Recharge subscribers nearing skip-or-cancel events, time the message before the next billing run, not after a cancellation has happened.

Calendar a 30-minute review with the product manager and ops lead. Bring the tagged review verbatims, ticket counts, and any photos. The goal is a decision on whether the issue is a one-off, a sizing or labeling fix, or a manufacturing-spec change.

Final read-out to leadership. Capture the headline insight, owner per action item, and any artifacts (slides, dashboard exports). Sign-off closes the analysis cycle and starts the clock on the next month's run.