03CONVERSION
Traffic tripled. Sales fell.
An ecommerce conversion case study: sessions up 295% while conversion sat at 0.23%. Finding the four failures after the click, and the repair that took conversion up 409% in three months.
A made-to-order home goods brand with high-value custom orders watched conversion collapse in early 2026. February recorded 45,901 sessions, up 295%, converting at 0.23%, with sales down 27%. From the outside it looked like the ads had stopped working.
THE IMPACT
Fix the ads. New creative, new budget, keep tuning.
The visible symptom was ad performance: a March Meta snapshot attributed roughly $1,609 in spend against $1,684 in value, and the newest campaigns kept underdelivering. The obvious move was to keep working the media: another campaign, another creative round, another budget shuffle. That would have optimized the most visible part of a system that was broken somewhere else.
The ads were the most visible suspect and the least broken thing.
The order path underneath the spend was failing in four places at once. Inventory and feed issues were sending the ad platforms bad signals. A surge of low-quality traffic was flooding the funnel and distorting every rate underneath it. Site reviews documented misleading price displays and unstable custom-sizing paths at the moment of decision. And the custom-pricing calculator could accept orders at base price without the custom charges, an order-integrity defect on the highest-value products.
Every media dollar was pouring into an order path that couldn’t hold it.
Four failures had stacked underneath the spend: bad inventory signals feeding the platforms, a flood of low-quality sessions polluting the funnel, documented site defects at the moment of decision, and a custom-pricing calculator that could take orders at the wrong price. Any one of them would drag conversion. Together they made media spend unreadable.
Stop feeding the funnel. Repair it in order.
We recommended pausing Meta on May 8; the client approved. With spend dark, the repair ran in order: fix the inventory and feed signals, repair the calculator and adjust base pricing so mispriced orders couldn’t recur, and rebuild the documented site defects. The restart was staged: a $20/day warm catalog campaign first, cold prospecting only after checkout proved it could hold an order at the right price.
- 01PAUSE THE SPEND
Meta went dark on May 8, with client approval, while the diagnosis ran.
- 02REPAIR THE PATH
Inventory and feed signals fixed, the pricing calculator repaired, the documented site defects rebuilt.
- 03STAGE THE RESTART
A $20/day warm catalog campaign first. Cold prospecting only after checkout held.
- 04PROVE, THEN SCALE
Spend earns its way back as the order path proves it can hold what media sends it.
THE COLLAPSE
February: 45,901 sessions, up 295%, converting at 0.23%, with sales down 27%.
THE TURN
May: conversion at 1.17%, monthly sales up 86%, average order value up 34%.
THE HOLD
May through June: $66,241 in sales, up 48%, on 14% fewer sessions.
By the May review the client described performance as back on par with the prior year. The following two months held the level: monthly sales up 55% against February on 77% less traffic, with the recovery credited to the repair program as a whole rather than any single fix.
Why pause the ads instead of fixing them?
Because four stacked failures underneath the spend made media performance unreadable. Pausing Meta on May 8, with the client’s approval, made the repair measurable instead of guessable.
What was actually leaking after the click?
Bad inventory signals feeding the platforms, a flood of low-quality sessions, documented site defects at the moment of decision, and a custom-pricing calculator that could take orders at the wrong price.
How fast did it recover?
Conversion moved from 0.23% in February to 1.17% by May, up 409% in three months, with monthly sales up 55% against February on 77% less traffic.
THE POINT
More traffic ≠ growth. The ads were the most visible suspect and the least broken thing. When conversion collapses while sessions surge, the money is leaking after the click, and the first dollar spent should be on the order path, not the auction.
