A competing bottle appears above yours, costs less per bottle, and has more reviews. Should you change your price, rewrite your listing, or investigate a different audience? Those observations do not yet answer the question. The result might be sponsored, the price might require buying two bottles, and the reviews might describe another size.
This guide is for Amazon sellers building a competitor analysis they can use to make a specific operating decision. It explains how to choose comparable products, organize keyword and offer evidence, interpret review samples, and turn findings into a next step. The worked example is fictional; it is not a report on an actual seller or an OpenMax customer result.
Quick answer: compare the buying decision, not just the listing
Amazon competitor analysis is a structured comparison of the products and offers a shopper could choose instead of yours. Start with one marketplace, buyer task, and selected product version. Build a small comparison panel, record the source and date of each observation, and separate organic visibility, sponsored visibility, offer conditions, and estimated sales. Finish with an action whose reasoning another person can check.
Use six steps: define the decision, select comparable competitors, collect like-for-like evidence, explain differences, identify what remains unknown, and assign one next investigation or test. If the evidence compares different pack sizes, periods, or product versions, repair the comparison before acting on it.
For an occasional comparison, a worksheet and source links can be enough. A larger catalog may need a research tool and a repeatable review process. AI can help organize supplied evidence, but it cannot turn an unobserved competitor conversion rate into a fact.
Define whether you are comparing a brand, a product, or an offer
ASIN stands for Amazon Standard Identification Number, the product identifier used in Amazon's catalog. It identifies the product being compared, not a seller's private sales record. See Amazon's ASIN guide.
“Competitor” can mean several different objects. Mixing them produces reports that look detailed but answer incompatible questions. In this guide, the comparison panel is the set of products you deliberately choose to examine; a seller offer is the purchase option presented by a seller for a product.
Swipe horizontally to view all table columns.
| Comparison object | Question it helps answer | Record explicitly | Common mistake |
|---|---|---|---|
| Brand | Which proposition or assortment competes for this audience? | Brand, relevant range, buyer task | Treating every product from a large brand as a direct substitute |
| Product family | Which related versions cover the buyer's requirements? | Family relationship where identifiable, sizes and formats | Counting variants as independent brands or adding overlapping totals |
| Selected child product | How does this specific size, color, or configuration compare? | ASIN, selected attributes, pack count | Using one variant's price with another variant's specifications |
| Seller offer | What purchase conditions are being presented for that product? | Seller, condition, fulfillment and displayed delivery context | Treating a seller-offer change as a new product competitor |
Amazon describes variations through parent-child relationships, with buyable child products related by attributes such as size or color. That distinction is a reason to record the selected version, not permission to assume every measurement covers the whole family. See Amazon's explanation of product variations.
For a private-label product decision, comparable products and their positioning usually matter first. For a reseller comparing offers on the same product, the offer conditions are more relevant. Name that choice at the top of the report. One analysis can contain both, but each row needs an object type.
Build a comparable panel from the buyer's task
Start with a sentence that can exclude products
“People buying bottles” is too broad. “A commuter buying one 750 mL insulated bottle that fits their bag” creates useful boundaries. A small children's bottle and a two-bottle household bundle can still provide context, but neither should silently become the benchmark for that single-bottle decision.
Write the use case, essential attributes, marketplace, and observation date before collecting ASINs. Add exclusion reasons: wrong capacity, different material requirement, replacement accessory rather than complete product, or an unrelated use case. The Amazon niche research guide explains how to define substitutes before estimating a market.
Use more than one discovery route
Search buyer-language queries, inspect relevant products encountered during research, and use available native or third-party reports to find additional candidates. Look for overlap across relevant queries rather than choosing only the first products you see. Do not describe a handful of search results as the entire market.
A practical first exercise is five comparable products across three relevant queries. These are manageable review limits, not a statistically representative sample or a recommended maximum. Add products when the initial set misses a meaningful format, price proposition, or buyer requirement. Remove a product only with a recorded reason; otherwise the comparison can become biased toward the conclusion you wanted.
Keep a stable core panel for repeated observations and a separate list of newly discovered products. Replacing the entire panel between reviews makes it difficult to tell whether competitors changed or your selection changed.
Record six evidence fields before writing conclusions
A useful note has more than a number. For each observation, preserve:
- Identity: marketplace, ASIN or offer object, and selected product attributes.
- Context: query, placement type, seller or delivery context when relevant.
- Value: the visible fact or the provider's reported metric, with units.
- Time: observation date or report period, including timezone when timing matters.
- Source: a retrievable link, permitted export, or internal evidence reference.
- Interpretation boundary: what the observation supports and what it does not establish.
For example, “sponsored placement observed for query X on September 8” is a different record from “organic position reported by provider Y for period Z.” Neither should be shortened to “rank 1” without its definition.
Mark missing data as missing. A blank sales estimate is not zero sales; an unavailable offer is not proof that the product was discontinued. When two sources disagree, retain both definitions and periods until you can reconcile them. Averaging incompatible values produces a tidy cell, not better evidence.
Analyze competitor keywords without confusing visibility with demand
Create a query-to-product matrix
Group queries by buyer intent: the core product, use case, essential attribute, and problem being solved. Record which comparable products appear for each group and whether the observation is organic, sponsored, or a provider-reported association.
An attribute query can reveal a different competitive set from the broad category phrase. A product repeatedly relevant to your specific buyer task deserves attention even if it is not the most visible result for a much broader query. Conversely, appearing for a popular phrase does not make every result a relevant substitute.
For eligible accounts, Amazon describes Brand Analytics dashboards including Search Query Performance and Top Search Terms. Its public documentation lists a Professional selling account and Brand Representative status for an enrolled brand as access requirements. Check the current account experience and report definitions rather than assuming access or complete competitor coverage. See Amazon Brand Analytics.
Amazon describes Search Query Performance around queries used to find your brand and associated shopping-funnel measures, while Top Search Terms provides a broader view of search activity and associated leading products. Use the former to investigate your brand's query context and the latter as another discovery input. Neither description promises a complete competing seller sales ledger. Preserve the exact report denominator before comparing a share with a ranking or an estimate.
Use reverse-ASIN results as sourced observations
Reverse-ASIN research starts with a product identifier to discover associated keyword information in a provider's dataset. For example, Helium 10 documents Cerebro fields for detected organic rank and sponsored-rank observations. Preserve the provider, marketplace, report date, and metric definition when using those fields. See Cerebro's field definitions.
Do not promise that a report exposes every term a competitor targets or every search that generated a sale. A keyword absent from the export might be outside the dataset or filters. It is not automatically an untapped opportunity. First ask whether the term accurately describes your product and whether the underlying result is comparable.
Keep sponsored and organic observations separate
Amazon explains that Sponsored Products can appear within shopping results and on product detail pages. A visible placement therefore cannot be treated as organic rank merely because it appears high on the screen. See Amazon Ads' sponsored advertising guide.
An observed ad also does not reveal its budget, profit, exact targeting setup, or the reason the advertiser selected it. Record the placement, then investigate your own relevant query performance separately. Avoid translating “competitor appeared here” directly into “we should bid on this term.”
Compare listing clarity and review themes, not only counts
Ask which buyer question the listing answers
Compare how each listing explains capacity, dimensions, compatibility, included items, and usage boundaries. Record what is actually communicated, then identify an unresolved question for your own product. The goal is to clarify a truthful proposition, not to copy another seller's images or wording.
For a bottle, a capacity label may answer one question while leaving another unanswered: will the bottle fit the intended pocket or holder? Your next task might be to verify the dimensions and produce a clearer image for your own product. That is a more defensible action than claiming a competitor wins because its images “look premium.”
Define the review sample before summarizing it
Record the number of reviews examined, how they were selected, their date range, and any identifiable product versions. Separate comments about the product from delivery, seller service, or an older design. Keep an ambiguous comment in an uncertain category rather than forcing it into the most convenient theme.
If four of twenty sampled comments mention the lid, the statement is about those twenty comments. It does not establish that 20% of buyers experience a lid problem. Reviewers are not a complete or randomly selected set of customers, and your selection method can introduce further bias.
Use themes to formulate questions for your own product: is the opening method clear, are the dimensions verified, or does the instruction omit an important step? Do not publish a claim that a rival product is defective based on a small collection of comments. A material allegation requires a different standard of evidence and appropriate review.
Interpret price and sales signals on their own terms
Keep checkout commitment separate from unit comparison
Record the selected version, pack count, displayed price, conditional discounts, and any relevant charges that are actually known. If a coupon's eligibility is not verified, show the displayed price and the unresolved condition separately. Do not silently subtract an assumed discount.
Dividing pack price by the number of identical items can help compare unit amounts. It does not make a two-pack interchangeable with a single item: a buyer may not want the larger purchase commitment. Capacity, accessories, and use requirements can also make a simple unit comparison incomplete.
Do not turn rank into a private sales ledger
Amazon distinguishes Best Sellers Rank from organic search rank. BSR describes sales performance relative to other products in a category; it is not an exact units-sold report for a competing seller. See Amazon's BSR explanation.
Third-party sales estimates require their own definitions, period, coverage, and uncertainty. A family-level estimate cannot be assigned to every child product and then added together. Nor can a product estimate automatically be allocated to a particular seller offer. The market-size estimation guide covers scope and aggregation errors in more detail.
Treat price changes, visibility changes, and estimated sales changes as separate observations. When they move together, that suggests a question to investigate; it does not prove which change caused another.
Worked example: three bottle offers do not imply one pricing answer
The following records are invented solely to demonstrate comparison logic. Prices are illustrative US-dollar amounts before any additional charges not specified here. No actual competitor or Amazon account was analyzed for this example.
Swipe horizontally to view all table columns.
| Record | Selected product | Displayed amount | Useful interpretation | Unresolved issue |
|---|---|---|---|---|
| Our product | One 750 mL bottle | USD 24 | Single-bottle purchase commitment | Whether our dimensions are clearly explained |
| Competitor A | Two 750 mL bottles | USD 36 per pack | USD 18 per bottle; USD 36 pack commitment | Whether our target buyer wants two |
| Competitor B | One 500 mL bottle | USD 19 | Lower displayed amount, smaller capacity | Coupon condition not verified |
Dividing USD 36 by two gives USD 18 per bottle. Compared with USD 24, that is USD 6 less per bottle, or 25% less using USD 24 as the denominator. But the pack requires USD 12 more than buying our single bottle before unspecified charges. Both statements can be true because they describe different decisions.
Competitor B's USD 19 amount is not a like-for-like 750 mL comparison. Converting every bottle into a price per milliliter would still not answer whether its shape, capacity, or intended use fits the same buyer. A normalization can support a comparison without settling it.
Suppose the analyst also notes one sponsored appearance for A and a few comments about bag fit across the sample. Those observations do not demonstrate that A's lower unit amount caused stronger sales, or that bag fit is the category's dominant purchase driver.
The justified next step is narrower: verify our product dimensions and review whether the listing clearly answers the single-bottle commuter's fit question. Keep the bundle-value hypothesis as a separate research question. An immediate price reduction is not established by this evidence.
Turn findings into an action with an explicit missing link
Use a short decision record: observation, possible explanation, alternative explanation, missing evidence, next action, owner, and review date. This forces the report to separate what happened from why the team thinks it matters.
For the fictional bottle example, the observed difference is pack size; whether our fit information is clear remains an open question. One hypothesis is that clearer fit information could help relevant shoppers evaluate our product if the current explanation is inadequate. An alternative is that the comparison group targets a different audience. The missing evidence is the actual dimension verification and relevant feedback about our own page.
The first action is to inspect those facts, not to launch multiple simultaneous changes. If a later content test is appropriate, define the expected observable outcome and avoid changing the product, price, and main image together while claiming to isolate one cause.
Prioritize findings by decision relevance, evidence strength, and reversibility of the next step. A weakly supported high-impact claim often deserves verification before execution. Do not manufacture a numerical opportunity score when the inputs are subjective or unavailable.
Choose the lightest workflow that preserves the evidence
Manual worksheet for an occasional decision
Start with a small panel, source links, and the six evidence fields. The output is one comparable record per observation plus a decision note. Stop collection when identities cannot be reconciled; adding more uncertain rows will not fix the mismatch. This method fits a narrow question and a person who can inspect every source.
Native reports for questions the account can answer
Use accessible Amazon reports for their stated purpose and retain the report period and definitions. If an account cannot access a dashboard, note the gap and use a narrower method; do not populate its missing values from unrelated estimates. Native data about your own brand should not be relabeled as a complete view of a competitor's business.
Rules or scripts for stable, permitted exports
Where a team already has an appropriate data export, deterministic checks can flag missing ASINs, inconsistent units, duplicated records, or mismatched periods. Review the input schema before each changed export format. Route failed records to a separate queue rather than dropping them silently or replacing them with zeros.
Agent assistance for explanations and handoffs
Give the assistant a bounded evidence packet and ask for source-linked observations, conflicting definitions, and unanswered questions. Review whether each sentence is supported by the supplied record. Reject inferred competitor spend, motives, or sales figures that the packet does not contain. The expected output is a draft analysis, not permission to change a live listing or campaign.
A shared operating process for repeated analysis
For several analysts or markets, define field ownership, change history, and an escalation route. Preserve an earlier snapshot when the panel or metric definition changes. Choose the review cadence around the decision: a one-off positioning exercise does not need the same observation schedule as a time-sensitive offer investigation. Avoid promising that a weekly or daily cadence is universally sufficient.
Where OpenMax fits—and where a research tool is still needed
OpenMax publishes this guide and describes its product as a human-agent collaboration workspace. That makes the coordination problem relevant: a team may have evidence but still struggle to pass a clear question, source packet, and decision history between people. See OpenMax's product overview.
A proposed workflow is to prepare one approved evidence packet, have an assistant draft the comparison and unresolved questions, and have a responsible person decide the next action. Evaluate that workflow against the actual OpenMax setup and available data connections before relying on it. This is an implementation proposal, not documentation of a native Amazon competitor-analysis connector.
OpenMax should not be presented here as a source of hidden competitor sales, guaranteed keyword coverage, or autonomous Amazon repricing. If you mainly need an ASIN dataset or keyword research interface, evaluate a dedicated source first. The Amazon product research tools comparison addresses that separate choice.
For a single occasional comparison, keep the worksheet if it already works. Consider the collaboration layer when ownership, source traceability, and repeated review—not a lack of raw data—are the bottleneck.
Limits: evidence quality comes before automation volume
Use only data your team is entitled to access and process. This article does not authorize scraping, access-control bypasses, or the reuse of another seller's creative assets. Obtain appropriate human review for the actual collection method, platform terms, and any sensitive data handling; a generic workflow cannot clear those obligations.
Do not include unnecessary personal information in review-analysis packets. Public comments can help identify product questions without copying reviewer identities into an internal narrative. The exact handling requirements depend on the data and operating context and should be reviewed by the responsible person.
Commercial actions also need their own checks. A competitor comparison is not a margin model, legal assessment, or approval to make product claims. Keep the analysis in draft when a material assertion cannot be verified. “Insufficient evidence to act” is a useful conclusion when it prevents an unjustified change.
Start with one decision your team can actually resolve
Choose one buyer task, five comparable products, and three relevant queries as an initial exercise. Record the selected versions and observation conditions, then produce one action note with its sources and missing evidence. Expand the scope only when the first comparison remains interpretable.
If the resulting handoff is the difficult part, bring that small packet to an OpenMax conversation and evaluate the proposed human-agent process. The question is not whether AI can produce a longer report. It is whether your team can trace a recommendation back to evidence and decide what to do next.
Frequently asked questions
How do I find my real competitors on Amazon?
Define the buyer's task and essential product attributes, then compare products found through relevant queries and available research sources. Keep direct substitutes separate from adjacent formats. Record why each product belongs in the panel instead of selecting only the first search results.
Is another seller on my ASIN a product competitor?
It is an offer-level comparison for the same identified product, rather than necessarily a different product proposition. Record the seller and purchase conditions separately from product attributes so that an offer change is not mistaken for a new product competitor.
Can I see exact Amazon competitor sales?
A public ranking or third-party estimate is not an exact competing seller sales ledger. Label the provider, period, object scope, and uncertainty. Do not assign a family estimate to each variant or assume product-level estimates belong to one seller.
How do I find competitor keywords?
Start with relevant buyer queries and comparable ASINs. Where available, use Amazon search reports or reverse-ASIN research, retaining their definitions and dates. Distinguish detected organic and sponsored information; missing keywords are not proof that a competitor never appears for them.
Can I do Amazon competitor analysis for free?
You can begin with a small manual comparison of available information and a worksheet. That provides limited observations, not complete historical coverage or private sales data. Check account eligibility before assuming native dashboards are included in your available tools.
How often should I update the analysis?
Match the cadence to the decision and the pace of relevant changes. Keep a stable panel and record material changes in definitions or selected products. A positioning review and an active offer investigation need different levels of freshness; there is no universal interval.
Can AI perform the entire competitor analysis?
AI can help structure supplied records and draft explanations, but missing data, ambiguous variants, and unsupported causal claims still require resolution. Keep source references with the output and have a responsible person review recommendations before making consequential changes.
Sources and editorial scope
Current public sources were checked on September 8, 2026: Amazon product variations, Brand Analytics, Amazon BSR, Amazon Ads, Helium 10 Cerebro definitions, and OpenMax.
The comparison method and bottle example are editorial explanations, not first-hand testing, measured keyword research, or an audit of a named competitor. Product access and data coverage should be checked against the account and provider being used.