Value Packs / Retail / Conversion & Checkout Funnel Radar / Wisdom

Conversion & Checkout Funnel Radar · Wisdom notebook

Ask the follow-up.

Agentic conversational BI · one thread across every paid surface, from the ones outside their band to the one whose orders come back.

Leaving the agent funnel aside — which paid surfaces are converting outside their own band this month, and is it the same cause each time?

RK

Scored every advertiser tier × department × campaign objective against its own seasonal attribution band, then, for each one outside it, pulled the merchandising and marketplace signal on the same slice to see what the ads were pointing at.

Four surfaces are outside their band and no two of them for the same reason — which is why a media review would have found none of them. Three are being let down by something downstream, and one is outperforming and being held back:

Paid surfaces outside their own attribution band

Attribution vs planWhat the ads point atCause
Apparel_Womens−32% iROASstock not sellingmerchandising
Beauty + Household · Conversion74% of plana pickup that failsstore operations
Home · mid-tier advertisers−18%orders that come backmarketplace
Beauty · mid-tier advertisers6.2× iROASinventory it cannot getyield cap

Mass-tier Beauty returns 3.9× against the mid tier's 6.2×, so the last row is a yield question rather than a fault. The other three each need a different owner — which is the point of scoring the surfaces separately rather than reading a blended iROAS.

Beauty and Household conversion ads at 74% of plan — is that the creative, or the pickup those ads are sending guests to?

RK

Held the campaigns constant and read the fulfilment leg instead: curbside on-time rate for the store cohort the impressions were served against, compared with the rest of the chain over the same Cyber Week, and the attribution each of them delivered.

The pickup. Fourteen stores in the bottom-remodel quartile fell to 62% curbside on-time during Cyber Week against a chain running at 97% — and the conversion-objective ads served into those markets are the ones at 74% of plan. The creative is identical in both columns:

Cyber Week · the click-to-pickup loop, and the ads that depend on it

0%27.5%55%82.5%110%62%74%97%100%Curbside on-timerateConversion-adattribution, % of plan
Bottom-quartile remodel stores, Southeast
The rest of the chain

Stockouts at the same cohort ran 1.8× baseline in the same week and guest NPS in the region sat 18 points below its own — a guest who clicks through to collect, cannot collect, and says so. The lever is the 4-to-7pm labour mix in fourteen named stores, not the campaign.

Mid-tier Home advertisers are down 18% — is that fewer orders, or orders being taken back?

RK

Separated attributed orders from attributed revenue for the tier, then traced the reversals to the marketplace cohort they were placed against and read that cohort's own return and shipping numbers.

Taken back. The orders were placed; a marketplace seller cohort onboarded last quarter is returning them at a rate that reverses the credit, and the advertiser sees the restatement rather than a weak week:

The cohort behind the restatement · week of 2026-01-19

Mid-tier Home attribution−18%
Marketplace return rate, this cohort31% vs ~12%
On-time ship rate, this cohort78% vs a 92% SLA
Share of the move from sellers rated under 3.561%

Nothing about the media changed. Worth telling the affected mid-tier advertisers the attribution will be restated before they find it in their own reporting — the reversal is real, and it is not theirs.