Research programme

Almost every claim in Amazon SEO is either Amazon's documentation or somebody's anecdote. There is very little in between. This is what we are doing about that, and the methodology we think any credible attempt has to meet.

Fact-checked 22 August 2026·Methodology and status·By Hymie Zebede

Why almost all published rank data is untrustworthy

Observed Amazon provides no organic rank data. Every rank figure in every case study you have read is third-party SERP scraping, and it fails in four specific ways that most write-ups never disclose.

Failure modeWhat it does to your data
Sponsored contaminationAround 30% of Amazon results are sponsored, and in 85% of queries a sponsored result appears above the first organic one. Tools differ in how they strip these to compute an organic position, so two trackers disagree about the same SERP by design.
PersonalisationResults are shaped by the observer's history. Any measurement taken in a working browser session is measuring the observer as much as the market.
Geography and devicePosition varies by delivery postcode and device class. A study that does not fix these is aggregating different questions.
Sampling cadenceRank moves intraday. Daily single-point sampling produces a series whose variance is mostly artefact.

Layered on top: since 10 August 2026 Sponsored Products reporting substitutes Amazon-inferred search terms for actual shopper queries in creator placements, which degrades the keyword-level data any ads-to-organic study depends on. Any before-and-after spanning that date needs to account for it.

The consequence

“We moved from position 14 to position 3” is not a finding unless the write-up states how sponsored slots were excluded, what location and device were fixed, how often it sampled, and what else changed on the listing during the window. Almost none do. That is why this field has thousands of case studies and almost no knowledge.

The methodology we hold ourselves to

  1. Fix the observation conditions. One marketplace, one delivery location, one device class, neutral crawl, no logged-in sessions.
  2. Separate organic from sponsored explicitly, and state the rule used. Report both, because the ratio itself is informative.
  3. Sample more than once a day, and report a trend with its variance rather than a point.
  4. Hold the listing constant. A rank study over a window in which price, images, copy, coupons or inventory changed is measuring the change, not the intervention.
  5. Use Search Query Performance as the ground truth wherever the question can be answered with it. It is Amazon's own funnel per query, and it is more trustworthy than any scrape — within its documented limits.
  6. Report the null results. The reason the ads-to-organic question is still open is that failed tests do not get published.
  7. State sample size, and do not generalise past it. Three ASINs is a hypothesis.

What we are running

ZBD Growth manages advertising and brand work across a portfolio of Amazon accounts, with daily keyword, placement and search-term data warehoused for the purpose. That is an unusual position from which to answer questions that need many accounts, many categories and a long window — which is most of the interesting ones.

Three studies are in progress. Results, methodology and the null findings will be published on this page.

On publishing negative results

If a study finds nothing, we will publish that it found nothing. The existing literature on Amazon ranking is badly distorted by the fact that only positive results are commercially useful to the people who run the tests.

If you want the analysis on your account rather than in the abstract

The research exists because the day job requires it. ZBD Growth runs Amazon advertising and brand management for sellers who would rather have this applied than explained.

Get a free account teardown →

Sources

Primary sources for the confirmed and published claims above. Observations and framework inferences are labelled as such in the text and are not sourced here.

  1. Sponsored is the New Organic — sponsored share of Amazon SERPs, arXiv — arxiv.org
  2. Amazon — Brand Analytics and Search Query Performance — sell.amazon.com
  3. PPC Land — inferred search terms in creator placements, August 2026 — ppc.land
  4. Amazon SEO Consultant — published PPC / organic split tests — amazonseoconsultant.com
ZBD Growth

Full Amazon channel management for established brands. Catalog, advertising, inventory and cases — run by a team you know by name, reported in depth.

© 2026 ZBD Growth. Selling on Amazon since 2012.Run by Hymie Zebede →