Your Prime Day wrap-up deck is probably telling you a comforting story. Here's why GMV, ROAS and SOV mislead, and the seven metrics that measure what actually happened.
Key takeaways
Right about now, a wrap-up deck is being built somewhere in your organization. It has a GMV number, a ROAS number, a year-over-year comparison, and probably a share of voice chart. It will tell a clean story.
It will also probably be fiction. Not because anyone's lying, but because all three headline metrics were designed to evaluate a campaign that no longer exists. They'll tell you that you won when you lost, and lost when you won.
Because GMV captures topline volume while hiding whether you bought it with margin you can't afford. Omnia Retail's pricing data makes the trap visible: during Prime Day 2025, 54.9% of products showed no price drop at all, while 45.5% were actually priced higher than the week before. The brands moving the most volume run one to three hero SKUs as loss leaders subsidized by a long-tail margin harvest. A GMV number can't tell you whether that architecture held or whether you discounted yourself into a record-breaking loss.
Because last-click attribution structurally undercounts the upper-funnel work that drove conversions and overcounts the branded-search clicks that would have happened anyway. In a 96-hour event where Impact.com clocks the average purchase journey at over seven days, the click that gets credit is rarely the touch that did the work. A blended ROAS over the event is an average of signals that contradict each other.
Because the surface it claims to measure no longer exists. With AI-mediated answers and audience-targeted sponsored placements, two shoppers entering the same query see different products in different positions with different context. Mars Agency's research, reported in The Drum, found only 22% of page-one organic products also appear in the AI assistant's recommendations. Your SOV report is a synthetic average of personalized auctions that no individual shopper experienced.
Seven things. Not in pursuit of comprehensiveness, but because GMV, ROAS, and SOV systematically lie about three specific failures: margin discipline, audience output, and post-event value. The scorecard is organized around those failures.
1. Audience output. Net-new addressable audiences built, cohort entry volume, and modeled 90-day value. The audience graph is the asset you monetize for the next year.
2. New-to-brand cohort economics. Actual cost per NTB acquisition, modeled 30, 60, and 90-day repeat rates, and contribution margin per cohort, return-adjusted.
3. Hero SKU margin discipline. Did the long-tail margin harvest cover the hero loss leader? For portfolio brands, this single number validates or invalidates the whole strategy.
4. BSR halo decay. How long did the organic ranking lift persist? A 14-day halo means the deal worked. A 2-day halo means you bought a spike, not a position.
5. Inventory health. Percentage of spend that landed on in-stock ASINs across the four days, Buy Box loss hours on hero SKUs, and proximity to low-inventory fees on restock.
6. Subscribe & Save conversion from NTB. For consumables, this is the metric that converts a one-time event into recurring revenue. If you're not measuring it, you're not running a post-event phase.
7. Spend efficiency by AMC audience. ROAS by audience rather than by campaign. The dispersion between your top and bottom deciles is the planning signal that replaces SOV.
Only if you present it that way. The practical implementation: the analytics team runs all seven during the post-mortem and writes the narrative, then the two or three numbers that materially shifted strategy get elevated into the executive readout. The point isn't to replace the one-pager. It's to make sure the one-pager is grounded in something that exists.
This is also where judgment earns its keep. Measurement has gotten harder precisely because execution has gotten more automated, and the differentiator now is knowing which signals deserve a decision. A comforting deck built on three legacy metrics outsources that judgment to numbers that stopped meaning anything two summers ago.
The full scorecard, with scoring guidance for each of the seven metrics, is in the 2026 Prime Day Playbook.
It is an analytical approach requiring AI to show its work, providing full transparency into the data retrieved, the entities resolved, and the logical evidence chain behind every recommendation, ensuring every number is traceable.
Generalist models often sound authoritative while using incorrect data. In retail media, a recommendation built on wrong numbers leads to wasted spend and faulty budget reallocation, making accuracy the only true performance metric for your stack.
Evaluate tools using a benchmark framework that prioritizes retrieval accuracy, entity resolution, and visible reasoning rather than just writing quality. A trustworthy tool must allow you to verify the logic and data behind every specific recommendation it makes.
No fluff. Just what's working.
