The Revenue Attribution Fight at a Subscription Box
Parcel & Pantry is a subscription snack box with 420,000 email subscribers and $31 million in annual revenue. Finance reports $5.2 million in online revenue last quarter. GA4 attributes $4.1 million to email; the email platform reports $6.8 million; Meta reports a 4.8x return on ad spend; Google Ads reports 3.1x. The team uses last-click attribution in GA4, inconsistent UTM tagging, and discount codes shared on coupon sites. Apple Mail Privacy Protection hides opens for roughly 35% of subscribers, but the team still suppresses 'non-openers.' The list has not been cleaned in three years: 58% have not opened in 180 days, bounce rate is 4%, spam complaints are 0.14%, and the team sends three campaigns per week to the full list. The VP of marketing wants to cut frequency to one per week. The e-commerce lead fears losing revenue and says email is free.
1. Reconcile the three revenue numbers. What is the likely explanation, and what would you do first?
Read the worked answer
The numbers measure different things and are not reconcilable to a single truth without a defined question. Finance's $5.2M is likely total online revenue, not email revenue. GA4's $4.1M is last-click email, which undercounts assisted conversions and is distorted by inconsistent UTMs. The email platform's $6.8M is probably platform-attributed revenue using its own window, including discount-code orders and possibly revenue from users who clicked but would have bought anyway; it can exceed finance because of different attribution windows and because coupon-site leakage inflates code-based attribution. Meta's 4.8x and Google's 3.1x are platform-reported, modeled, and view-through-inclusive; they overlap with email and each other. First step: freeze a definitions document-date range, revenue definition (gross vs net of returns), attribution window, and inclusion of tax and shipping. Then build a reconciliation bridge from finance total to channel-attributed totals, showing overlap and unattributed revenue. Do not pick the highest number. Use finance as the source of truth for total revenue, and treat platform numbers as directional until incrementality tests are run.
2. Design a measurement framework that survives MPP, consent, and coupon leakage; include email engagement.
Read the worked answer
Use a layered framework. Layer 1: finance-grade total revenue and margin by cohort. Layer 2: channel contribution with a consistent attribution model-start with last-click for comparability but add a data-driven or Markov model for decisions. Layer 3: incrementality via geo holdouts, PSA tests, or conversion lift for email and paid social. Layer 4: behavioral diagnostics. For email under MPP, opens are unreliable; use clicks, site visits, add-to-cart, purchases, and reply rates. Segment engaged users by click and purchase recency, not opens. Use UTM governance: one naming convention, a registry, and validation in the email and paid teams. For coupon leakage, create single-use codes or affiliate-specific codes and exclude coupon-site traffic from email attribution. Run a holdout: suppress email for a random 5% of engaged subscribers for four weeks and compare revenue. Report a range, not a false-precision number.
3. Recommend frequency and list-hygiene plan; address the revenue fear with a test.
Read the worked answer
Cutting to one send per week is plausible but should be tested, not declared. The 58% inactive segment is not a revenue asset; it is a deliverability liability. Build engagement tiers from clicks and purchases: active (0-90 days), lapsing (91-180), inactive (181-365), and dormant (365+). Send full frequency only to active. Run a re-permission campaign to lapsing users with a clear value proposition; suppress inactive and dormant from regular sends. Sunset dormant users after a final winback, but keep a small holdout to measure whether suppression hurts revenue. Fix the dedicated IP by throttling volume, targeting engaged users first, and monitoring Google Postmaster Tools and spam complaints. The e-commerce lead's email-is-free claim ignores deliverability damage and list fatigue; the real metric is revenue per recipient and per 1,000 sends, not total email revenue. Test one vs three sends per week on a random 10% holdout for 60 days, with cancellation rate and revenue per subscriber as guardrails.