--- title: "The agency’s ‘winning’ campaign was the biggest loser" description: "Attribution credited a campaign for sales it didn’t drive; budget reset to incremental channels. Marketing & margin audit: $53k/yr in wasted spend cut." url: https://marginfix.ai/cases/the-agency-s-winning-campaign-was-the-biggest-loser/ lastmod: 2026-10-24 --- The agency’s ‘winning’ campaign was the biggest loser. [Request an audit →](/book/) Ecommerce · Home & living # The agency’s ‘winning’ campaign was the biggest loser. This marketing & margin audit identified **$53k in wasted annual spend** in a DTC / Ecommerce business, evidenced, senior reviewed, and delivered in 7 days. $8–20M annual revenue Home & living Focus: Attribution −$53k wasted spend cut / yr −34% non incremental spend 7 days to findings Incremental ROAS: original campaign vs replacement channels Before audit 0.2× → After fix 2.1× ## The business A $14M home and living brand had scaled paid social aggressively over two years, and one campaign its agency had branded the "hero" was reported as the single biggest revenue driver in every monthly deck. Leadership trusted that number implicitly and kept feeding the campaign budget, treating it as the safest, most productive line in the entire media plan. ## What triggered the audit The founder asked a deceptively simple question the agency couldn’t answer cleanly: if we switched the hero campaign off tomorrow, how much revenue would we actually lose? The reporting had always assumed every attributed sale was caused by the ad. Nobody had ever tested that assumption, so the audit ran the experiment properly. ## What the audit found A matched market geo holdout test told a very different story from the dashboard. The "hero" campaign was almost entirely retargeting warm buyers who were already deep in the purchase journey and would have converted anyway; last click attribution simply stood at the finish line and claimed the credit. When the campaign was switched off in test markets, sales barely moved, which meant its true incremental contribution was close to zero. Yet it was absorbing roughly $53k a year of budget and, because it always "won" on attributed revenue, it was first in line for every increase. The brand had been scaling its least incremental spend and starving the channels actually creating new demand. ◉ How we produced this finding Behind the attribution finding is a deliberate test of cause and effect: MarginFix separated the spend that genuinely created demand from the spend that only claimed it, using its A.I Marketing Orchestrator, then had a senior auditor verify the evidenced read before it reached the client. Data sources: Spend by channel, campaign, creative and audience, joined to conversion and revenue data, plus a geo holdout test built to isolate what attribution genuinely caused rather than what it merely claimed. Key frameworks: Geo holdout incrementality testing, Marketing Mix Modeling (MMM) and attribution correction and inflation factor analysis, cross checked against Analytic Partners ROI Genome, Google Meridian and Meta Robyn. Human validation gate: Every incrementality read is rerun against your own data and signed off by a named senior auditor before it ships. No automated output ever leaves the building unreviewed. Verified against Ad platform exports Incrementality / geo tests GA4 / analytics Shopify / order data Finance P&L The wasted spend cut / yr was measured like for like over a matched period, reconciled to invoiced margin in the P&L, and signed off by a named senior auditor before publication. Client identity redacted to protect their commercial position. WORKING PAPER ████████ Home & living Representative Redacted Campaign Incremental ROAS Prospecting, cold 2.3× Retargeting, warm 1.9× "Hero" campaign 0.2× wasted spend cut / yr +$52,990 Recurring, recovered every year the fix holds, not a one off. Working paper: incremental ROAS of the original campaign versus replacement channels traced line by line and reconciled to invoiced margin in the P&L over a matched period. Line items representative and redacted; the recovered figure is the reconciled audit finding. ## What we changed ✓ Redirected the budget into two prospecting channels that each passed a geo holdout incrementality test before being scaled, so money follows proven cause and effect rather than attributed credit. ✓ Rebuilt the agency’s reporting around incrementality rather than last click, so “winning” now means demonstrably causing sales instead of standing at the finish line to claim them. ✓ Established a standing quarterly holdout test, so no campaign can quietly drift back into harvesting demand it didn’t actually create. ✓ Reset the budget allocation rule to weight incremental contribution, ending the automatic reward that high attributed revenue used to guarantee. ## The result The revenue band describes the same $14M business. The ROAS comparison is between the original campaign and the replacement channels. $53k a year moved from a campaign that caused almost nothing to channels that provably drive new customers, an annual figure equal to 9× the $5,950 Audit + Sprint fee. The founder had trusted that “hero” number for two years; only a holdout test could expose it, because attribution will never disprove itself. If you cannot answer “what would we lose if we switched this off,” you may be funding demand you already own. That gap widens every month it goes untested, and a fixed fee audit closes it in days, with a repeatable test you keep. From kickoff to signed off findings: 7 days. ![Portrait photograph of Victoria Miller](/assets/team/victoria-miller.jpg) Reviewed & signed off by: Victoria Miller Senior Auditor · MarginFix · 10+ years of auditing experience ✓ Anonymized to protect the client · senior reviewed findings · Audit completed between 2024 and 2026, contracted through Cristian Bragau's practice before the MarginFix name existed · Published in the MarginFix library July 21, 2026 · Last reviewed October 24, 2026 ## What the client said DTC / ECOMMERCE $53k a year of wasted spend cut in 7 days Incremental ROAS: original campaign vs replacement channels: **0.2× → 2.1×** “Our agency's hero campaign was the biggest loser. Victoria ran a geo holdout and sales barely moved when it was off. Two years of attributed revenue, almost none of it caused by the ad. The budget now goes to channels that pass that test.” **CMO · Ecommerce · Home & living** Written approval Under NDA 7 days to findings ![Portrait photograph of Victoria Miller](/assets/team/victoria-miller.jpg) **Victoria Miller** Senior Auditor · signed this audit off [Request an audit →](/book/) [Prefer to talk it through first? Talk to an auditor →](/book/call/) ## More DTC / Ecommerce findings [Free shipping was quietly deleting the margin on every third order, +$47k/yr](/cases/free-shipping-was-quietly-deleting-the-margin-on-every-third/) [Retargeting was paid twice for the same customer, −$31k/yr](/cases/retargeting-was-paid-twice-for-the-same-customer/) [Bestsellers were priced below their true landed cost, +$59k/yr](/cases/bestsellers-were-priced-below-their-true-landed-cost/) [Subscription churn was masked by aggressive new trial spend, +$55k/yr](/cases/subscription-churn-was-masked-by-aggressive-new-trial-spend/) [Bundles were cannibalizing full price hero SKUs, +$38k/yr](/cases/bundles-were-cannibalising-full-price-hero-skus/) [FBA fees and returns had overtaken the margin on the hero ASINs, +$61k/yr](/cases/fba-fees-and-returns-had-overtaken-the-margin-on-the-hero-as/) [The benchmark: what audits recover across 74 cases →](/benchmark/) [All case studies](/case-studies/) [Model a DTC / Ecommerce budget in the wasted ad spend calculator →](/estimator/ecommerce/)