Methodology · 8 of 8

Every number,
labeled and explained.

How this audit was measured, which source answers which question, and the plain-language meaning of every attribution term used in these pages.

01Data Sources

Four sources, four jobs.

Triple Whale (API)The source of truth for all business performance: revenue, orders, ROAS, CPA, channel mix, customer metrics. Pulled: 90-day summary with daily series + order-level attribution for 20,299 orders (May 28 – Jun 10).
Google Ads (API, read-only)Both accounts. Used only for what platforms uniquely know: spend, budgets, impression share, Quality Score, ad policy status, search terms, settings. Windows: 30d (May 12 – Jun 10) and 90d (Mar 13 – Jun 10), 2026.
Field researchCompetitor sites, search results, Meta Ad Library, FDA/FTC public records — June 11, 2026. Facts labeled OBSERVED / INFERRED / UNVERIFIED; unverifiable numbers are not quoted.
Google Keyword Planner (API)Search-volume and bid-range data behind the advertorial keyword research (US, June 2026).
02Attribution, in Plain Language

The four terms used in this report.

TermWhat it meansWhen we use it
Last clickThe whole order is credited to the last ad the customer clicked before buying.Default for campaign tables — it matches how day-to-day campaign decisions are made.
Linear (multi-touch)The order's value is split evenly across every ad touch in the customer's journey. Example: a $90 order that touched an AppLovin ad, a Google search ad and an email gets $30 credited to each.To see which channels participate in journeys even when they don't close them. A channel whose linear number is much higher than its last-click number is a journey starter.
Deduplicated channel viewEach order counted once across all channels (credit shared when several channels touched it). Channel numbers add up to total revenue.Channel mix and budget allocation (Summary page). Google = $323.5K / ROAS 2.05 in the 14-day window on this view.
Platform-comparable viewGoogle gets full credit for every order where it had the last platform click — the same rule Google's own tag uses. Numbers from different channels must NOT be added up in this view.Only for the tracking audit: comparing what Google's tag reports ($259K) against what Triple Whale says Google actually drove ($611K) → the 42% capture finding.

Why two views of the same Google number is honest, not contradictory

"How much credit does Google deserve in the mix?" (deduplicated: $323.5K) and "is Google's conversion tag seeing everything it should?" (platform-comparable: $611K vs $259K reported) are different questions. Each table in this report states which view it uses.

03Reconciliation Method

How the 42% capture figure was measured.

14-day window (May 28 – Jun 10), chosen to end 2+ days before the pull so late-arriving conversions don't skew the comparison. Both sides aligned to order date (Triple Whale's convention; Google's "conversions by conversion date" metric). Purchase conversions only. Joined campaign-by-campaign using the campaign IDs Triple Whale records on each order's ad touches. Result: Google's tag reported $259,380 of conversion value; Triple Whale credited Google-last-click orders $611,467 → 42% capture, worst on brand search (13%). Most likely causes: the click ID getting lost across the funnel's domain hops (advertorial → store → external checkout) and conservative data-driven attribution. The repair plan and its success metric live on the Search page.

04Limitations · Stated Plainly

What this audit cannot see (yet).

Subscriptions

New vs recurring orders can't be split per channel yet

Triple Whale's subscription metrics read from its Recharge integration, which shows only 2 active subscriptions — clearly not the real recurring base for a store where 52% of orders come from returning customers. Recurring billing runs through the external checkout, invisible to Triple Whale as "subscriptions". Fix: action 6 on the Action Plan.

Channel split

Per-channel revenue is a model

Triple Whale's blended totals are ground truth (real orders); the per-channel split depends on its pixel and attribution rules — the best available view, but a model. Product-level revenue exists only platform-side (used for relative comparisons only, labeled where shown).

Correlation ≠ causation

The demand analysis is correlational

AppLovin spend moving with brand searches (r = 0.68) is strong evidence, not proof. That's exactly why the YouTube proposal pre-registers a lift measurement instead of assuming the effect.

Field data

Competitor pages change fast

All field research dated June 11, 2026. Meta Ad Library per-brand ad counts were not reliably retrievable and are not quoted. Anything unverifiable is labeled UNVERIFIED or excluded.

← PreviousAction Plan Back to startSummary & Business