eCommerce GrowthAnalytics & Data

How We Build the eCommerce Ad Reports: Data Rules and the Standard Test Card

First published Nov 11, 2026Updated November 11, 20264 min read
Valentin Radu, Founder and CEO of Omniconvert
Valentin Radu
Founder & CEO, Omniconvert · Author, The CLV Revolution
Published: Nov 11, 2026Updated: Nov 11, 2026
Reviewed by Cristina Stefanova, Head of Content
Quick Answer
Every Omniconvert eCommerce ad report counts the ads brands ran in a month, from the eCommerceBenchmark Ad Library. It shows presence and share, not sales or profit. Each report ends with a test card; this page holds the setup, length, stop and win rules that every card shares.

Each month we publish eCommerce ad reports by market and by industry, such as the October 2026 eCommerce ad report and the October 2026 US creative report. They share one method and one test card. This page explains both once, so each report can spend its words on its own findings.

Where the numbers come from

All counts come from the eCommerceBenchmark Ad Library. Each report states its own scope at the top and again in its data note: the market, the industries, the number of brands and ads, and the two months it compares.

  • Active, not new. A share counts every ad active in the month, not only the ads launched in it.
  • Presence, not performance. The library shows what brands ran. It has no sales, spend or profit data, so a large share does not mean an ad earned money.
  • Figures can change. The library re-measures past months as it labels more ads and updates the stores it tracks. A month's figures can differ from one edition to the next. The newest edition has the current read.

How to read the figures

  • Three separate decisions. Every ad is labelled by its hook, angle and concept. This split is Omniconvert's own framework, not a library finding. Compare hooks with hooks, angles with angles and concepts with concepts.
  • Concepts rest on fewer ads. Fewer ads carry a concept label, so compare concepts only with each other.
  • Sixty-day figures are a separate sample. Staying-power numbers and the named long-running ads come from a separate longevity-ranked ad sample. Never add them to the monthly shares.
  • Our notes are ours. The "what this means" notes, the named brief and the thirty-minute check are Omniconvert's reasoning, not library findings.

Long-running ads are a useful signal in their own right; see the ads still running after a year. When an ad that ran well starts to slip, check for ad creative fatigue before you blame the concept.

The standard test card rules

Every report ends with a test card for one ad idea. The card gives the hypothesis, the one thing to change and what to keep the same. The rules below apply to every card. They come from general A/B testing practice, not from the library.

Standard rules for every test card in the ad report series.
Funnel stageSay first whether the ad is for prospecting (cold) or retargeting (warm). A deal in the first line is often right for warm audiences and wrong for cold ones, so test it where the ad will really run.
SetupOne control and one variant, same audience, same placement, even budget split. Do not edit either ad while the test runs.
How long and how muchUntil each version has about 50 purchases, or two full weeks, whichever comes later. Meta's own learning-phase guideline is about 50 results per ad set per week. Too few results and the gap is noise; see sample size.
Success metricCost per purchase, not clicks, unless the card names a different metric.
Stop rulePause a variant that has spent about 3 times your target cost per purchase with no purchase. If it is close, keep it running.
What counts as a winThe variant beats the control on the metric by an amount you would act on, and still does in a second week.

Write your numbers down before you launch. A target set after the results arrive is not a target.