Two score bases answer different questions
For a new store, we first check whether the exact domain can be corroborated as the official website of an established brand and its accountable operator. The established-brand identity route records 100/100 when that relationship is supported by the required official source and separate corroboration. A familiar name, copied logo, old domain or attractive design does not establish ownership.
That 100/100 concerns official-brand identity. It does not grade product quality, value, delivery, service or refunds. Relevant attributed customer complaints and restrictive buying conditions remain part of the written review. A genuine official website can still receive serious complaints.
Other future stores use the weighted evidence-confidence route when valid observations are available, or remain unscored where the evidence or scope cannot support a number. An identity score and an evidence-confidence score are not comparable measures of purchase reliability. Neither is a customer average, statistical probability, certification or guarantee.
Version 1.4 applies to future new stores
The new weighted method measures the confidence supported by recorded evidence across a fixed rubric out of 100. Missing evidence adds no supported confidence. It remains unknown rather than becoming a finding of failed deliveries, poor quality, dishonest conduct or an illegitimate business.
A lower number can therefore reflect limited evidence, actual adverse findings, or both. The explanation must distinguish those reasons. When a low number mainly reflects information we could not establish, we say so rather than implying that customers were defrauded.
Earlier versioned assessments used their own recorded arithmetic, including normalization over assessed evidence. Existing reviews keep their scores, method, dates and evidence. This version release does not rescore them or turn earlier publisher opinions into results of a method they did not use. Deliberate corrections are separate editorial decisions with retained history.
What we examine about an online store
The weighted route examines the identity linked to the exact storefront, delivery commitments and charges, usable returns and refunds, payment conditions, attributed customer outcomes, and the consistency of material claims. These questions are assessed separately because they establish different things.
A clear shipping policy supports an observation about the stated promise. It cannot demonstrate that parcels arrive. A refund deadline does not prove refunds are paid. Polished policies must not outweigh well-supported adverse outcome evidence simply because the retailer explains its promises clearly.
Our internal worksheet retains the criterion ratings, calculation, qualifications and source links. Public store pages display one final overall number and explain the substantive supporting observations, concerns and unknowns in ordinary language. The method is an editorial framework and has not been calibrated as a predictive risk model.
Match sources to the exact business
We begin with the supplied domain and check whether a store review already exists. Existing stores are skipped by the new-review workflow. Similar names, regional storefronts and redirects are recorded rather than silently treated as the same merchant.
A retailer page establishes what the retailer publishes. A registry can establish facts about a registered entity or domain. Independent original reporting and attributed customer accounts may establish other observations when the business, storefront and relevant period match. We explain the inference separately from the source fact.
More links do not necessarily supply more independent evidence. Repeated announcements, copied complaints and aggregator summaries may share one underlying origin. External platform ratings are not averaged into our score. An inaccessible source is a research limitation, not a clean check or evidence of wrongdoing.
How customer accounts affect an assessment
We record specific product-description, delivery and completed-refund accounts with their original post dates and source references. Positive and negative accounts follow the same inclusion rules. We retain contrary context and relevant seller responses; a reply alone does not prove resolution.
For version 1.4, sufficiently distinct underlying customer accounts may come from one platform when their provenance and possible cross-posts have been reviewed and documented. Different websites do not prove different customers. Copies of the same underlying account are not extra corroboration. If account independence is unreliable, the reports remain attributed context rather than graded outcome evidence.
These are self-selected, attributed accounts, not transactions authenticated by StraightReview, verified unique purchasers or a measured complaint rate. We do not infer a completed failed refund from an open return discussion. Older reports and gaps in accessible evidence remain clearly qualified.
Historical reviews retain their original report rules. New account-handling rules do not silently reinterpret earlier research or change existing merchant scores.
Domain age and security have narrower meanings
Our registration check uses primary RDAP data for the exact domain and records the returned registration event and actual check date. A registration date concerns the domain object; it does not establish continuous trading, current ownership or order reliability. If no usable primary date is available, that limitation stays visible. Age alone adds no automatic bonus or penalty.
HTTPS describes an encrypted connection to a host. It does not prove that products match descriptions or that a merchant resolves complaints. Redacted registration details, shared hosting or a new domain are not, by themselves, findings of misconduct.
As of 25 October 2026, our review pipeline has no external malware or blocklist API configured and does not automatically run URLVoid scans. We do not report an unchecked service as having cleared a store. Any future threat-data integration would require permitted access, current results, correct attribution and careful treatment of false positives; an absent listing would still not guarantee fulfilment.
What the research actually does
Bounded collection can retrieve readable retailer pages and primary registration metadata. The research agent or editor then matches identity evidence and customer accounts, checks current policy scope and dates, considers contrary observations, and records the evidence. Deterministic code validates structured inputs and calculates the selected version of the method.
Customer provenance and factual interpretation still require editorial review. Publication is a deliberate decision; this is not a fully unattended crawler issuing verified verdicts. Access failures, limited samples and unresolved redirects are retained, and some stores remain unpublished while material questions need follow-up.
We do not place test orders, receive deliveries, contact support or complete returns as part of this assessment approach. We do not claim to have tested product authenticity, service quality or refund performance. Published policies and other people’s accounts must not be presented as our own purchase experience.
Why a result may be higher or lower
For a future weighted assessment, stronger relevant support across the defined questions can increase supported editorial confidence. Specific mixed or adverse findings can reduce the points supported by an assessed question, and serious unresolved concerns limit favourable conclusions. Those are editorial decisions, not scientific risk thresholds.
Missing information and observed problems require different explanations. We may be unable to corroborate an operator or find suitable customer accounts. Alternatively, current sources may directly conflict about who is responsible or how a return works. The first is an evidence gap; the second is a specific observation requiring context and investigation.
A lower number must not be described as proof of wrongdoing, and a higher number does not establish a safer individual order. For the separate official-brand identity route, explain the domain-to-brand relationship and the customer outcomes that its 100 leaves outside the score.
Editorial responsibility, responses and corrections
StraightReview remains responsible for the wording, attribution and opinion it publishes. Ownership, affiliation, sponsorship and other material relationships need accurate disclosure. Automation does not remove those responsibilities or justify a claim that a test or verification happened when it did not.
A correction should identify the exact statement and domain, provide an appropriate source, and explain what changed and when. A factual correction differs from disagreement with an opinion. We preserve the earlier assessment and explanation when an authorized change is made. Payment, pressure or superficial redesign must not purchase a favourable result.
Read the evidence before relying on the number
Start with the score basis, method, domain examined and source-check date. Then read the attributed findings, purchase conditions and limitations that matter to your destination and product. A page cannot resolve facts that have changed since its research was checked.
Useful original analysis, accurate sourcing and transparency guide the work. Google’s review guidance does not approve our score formula or guarantee rankings. FTC guidance also explains why review collection and presentation must not distort customer opinion. Businesses can read our separate transparency guide for practical improvements that make information and outcomes easier to assess.
Clear sources. Useful context. A considered view.