WELO TRUST RESEARCH
How we collect evidence, assess companies and build clear, inspectable sector comparisons.
01
The directory contains published reports only. Recently Researched follows the publication date of the current public version. Categories come from the Research CMS. A draft or an unpublished company never enters a public ranking.
Most Viewed Today uses report views since 00:00 UTC. Popular This Week uses today and the previous six UTC calendar days. Trending Now compares the last seven complete UTC days with the preceding seven complete days. It requires at least 10 views in the recent period, 5 in the previous period, an increase of at least 5 views and growth of at least 50%. Eligible companies are ordered by proportional growth, then recent views. These thresholds prevent tiny changes from appearing as a meaningful trend.
Most Searched uses directory searches over today and the previous six UTC days. A search is recorded after submission or a pause of 1.2 seconds with at least two characters. Exact company names and domains take priority; otherwise the query must identify one published company unambiguously. Broad queries do not count as searches for every result. Research requests, card clicks and visits to external websites remain separate event types.
With analytics consent, a report view is recorded after the page has been visible for 1.5 seconds. Repeated views of the same report from the same browser are counted once per UTC day across English and Italian. Company searches and each type of company click are likewise deduplicated daily. Directory search totals deduplicate the same query from the same browser during a day. These are browser-based measures, not a verified count of people.
Known bots, prefetches, hidden pages, development previews and controlled QA sessions are excluded. Visitors who decline analytics or enable Do Not Track are not included. Raw search queries and contact details are not stored in the ranking tables. Short-lived daily hashes support deduplication; the directory reads aggregated counts refreshed at most once per minute. Umami receives separate consented event categories without raw search terms. Rankings are hidden when no qualifying data exists; we never substitute illustrative statistics.
02
Welo Trust Research helps readers understand a company before buying a product or subscribing to a service. We bring together the practical information that is often spread across terms of sale, return policies, payment pages, privacy notices and support centres. The aim is to make those conditions easier to find, compare and understand.
A report is an assessment of the evidence available at the stated research date. It is not a purchase recommendation or a promise that every order will go smoothly. Read the company-specific findings alongside the scores: a convenient payment method does not remove a restrictive return condition, and a high score in one area does not settle every question about a business.
03
We start with the company’s official website and the documents governing the transaction or service being reviewed. Depending on the business, these can include purchase terms, cancellation and refund policies, subscription conditions, delivery information, privacy and cookie notices, contact pages, product documentation and help-centre articles. Relevant external sources may add context; their origin remains visible.
Sources are stored with a URL, publisher, supporting excerpt and access date. Citations connect factual statements to the pages that support them. Multiple excerpts may come from the same page, so the number of evidence records in the editor can differ from the public source count. The public checked-source count counts distinct checked URLs, not every excerpt as a separate source.
A company’s statement establishes what it publishes or promises. It does not by itself prove that every customer receives that experience. We distinguish written policies, reported statements and information that has not been established, rather than presenting them as equivalent evidence.
04
Conditions can differ between countries, sales channels, products and customer types. We identify the scope of the policies we review and qualify a finding when that context matters. A return policy for an official European store should not be read as the policy of every reseller worldwide. A device privacy policy should not be used as evidence of how an online shop handles customer data.
For subscription services and SaaS, Returns covers cancellation, termination, renewal and refund conditions rather than physical shipping. This keeps the comparison useful across business models while preserving the details that differ. Readers should still check the conditions shown for their own order, location and chosen plan before committing.
05
The research workflow brings together source material, identifies relevant facts, prepares the company overview and evaluates the four rating areas. Source excerpts are checked against their original pages. Where a page cannot be read automatically, a recorded source review can establish the supporting evidence. An inaccessible page is not treated as proof that a policy or service is absent.
Before publication, the report is checked for source references, complete company information, language consistency and the evidence behind its ratings. English and Italian are published together from the same report version. The numerical assessment is shared across both languages; translation does not create a different score.
Draft changes remain separate from the public report until publication. Earlier versions remain in the editorial history. Updating a score or its supporting evidence requires a new review, so a previous approval is not silently reused for changed findings.
06
Each assessed area receives an integer score from 1 to 100. The score is our editorial evaluation of the documented protections, practical conditions and material limitations relevant to that area. It is not a percentage of satisfied customers, a probability that a company is safe, a legal compliance verdict or a measured service-performance statistic.
Overall is calculated, not rated separately: (Returns + Payments + Privacy + Support) ÷ 4. Each area has equal weight of 25%, and the result is rounded to the nearest whole number. Overall is shown only when all four area ratings have sufficient evidence. Strong performance in one documented area can therefore coexist with a lower Overall result when another area has material limitations.
A policy does not earn a high score simply because it exists. We consider what it actually says, how usable the stated arrangements are and which restrictions matter to a customer. The supporting rationale is retained with the assessment and its evidence in the editorial record.
Eligibility, cancellation or withdrawal window, return costs and refund procedure, timing and method. For SaaS: cancellation, renewal and refund terms.
Payment choices, fees, charge timing, eligibility and documented payment protections.
Data collection and minimisation, purposes, third-party sharing, optional tracking, user rights and controls. Website and service scope must be explicit.
Contact options, help resources, published availability and complaints handling. Published targets are not measured performance.
07
The same rating bands guide assessments across companies. The exact number within a band reflects the balance of documented strengths and limitations; it is not the output of a customer survey. Small numerical differences should be read cautiously, especially where companies operate under different policy scopes.
1–20 indicates substantial restrictive practices supported by specific evidence. 21–40 indicates documented material limitations. 41–60 indicates basic protections or mixed conditions. 61–80 indicates clear, useful protections with identified limits. 81–100 is reserved for unusually comprehensive and favourable documented protections across the area’s factors.
Missing information is not automatically a negative finding. It reduces the certainty of an assessment and may prevent us from showing a score. We do not turn an unreadable policy, an unfinished review or a missing source into a zero. A score near the top of the scale is also not a guarantee of future conduct.
08
Tech groups electronics and software. Home & Appliances groups household goods, furniture and electrical appliances. Fashion & Accessories, Beauty & Care, Food & Beverage and Travel each have their own broad comparison group. Ecommerce groups general online retailers. Marketplace separately groups platforms connecting buyers with third-party sellers.
Classification follows the company’s main offering, rather than the mere presence of a checkout. An electronics brand that sells directly through its own website remains Tech. A general online marketplace is classified as Marketplace. The industry stored in the CMS determines the comparison group, and unclassified businesses are not silently added to an unrelated benchmark.
As reviewed companies are published, their eligible ratings automatically become part of the appropriate sector sample. Tech and Ecommerce are not combined. A change of classification affects the comparison only when the updated report is published.
09
For each chart area, the sector AVG is the arithmetic mean of the eligible ratings of other published companies in the same sector and methodology version. The company being viewed is excluded. Each company domain counts once, using its current published report, so duplicate reports and historical versions do not increase the sample.
An average appears when at least 5 other companies have an eligible rating for that area. This threshold is an editorial minimum, not a claim of statistical representativeness. The sample consists of companies researched by Welo, not every company in the market.
Eligibility is checked separately for each area. If a company has sufficient evidence for Payments but not Privacy, it can contribute to Payments without being counted in Privacy. Overall AVG uses the complete Overall scores of the eligible peer companies. Missing values are excluded, never substituted with zero.
The mean is rounded to one decimal place after calculation. For example, peer ratings of 60, 64, 68, 72 and 76 produce an AVG of 68. This is a calculation example, not data from a live company report. Because the subject is excluded, the peer AVG can differ slightly between two companies in the same sector.
10
The darker blue bar is the company’s rating; the lighter blue bar is its sector AVG. The tooltip identifies both values for the selected area. Overall, Returns, Payments, Privacy and Support all use the same 1–100 scoring scale, with a zero baseline on the chart to keep bar heights comparable.
Overview cards repeat the individual ratings, show the sector AVG, and display the difference in points. A positive difference means the company’s rating is above the displayed peer average; a negative difference means it is below. It does not mean that the company is a corresponding percentage safer or better.
The four factual cards near the headline serve a different purpose: three concise company-specific facts and a checked-source count. They are sourced facts rather than ratings. The source count is calculated from the evidence library and is not used as a shortcut for a higher score.
11
The research date shows when the report was assessed. Source access dates and the recorded rating review provide additional context. For a public rating to contribute to a benchmark, the report, review and relevant evidence must be no more than 180 days old. Older or incomplete assessments are excluded until reviewed again.
When a sector has too few eligible companies, the individual company ratings can still be shown while the average remains unavailable. We do not borrow an average from another sector, invent a comparison bar or reuse incompatible historical scores. If a data request temporarily fails, that state is kept separate from an insufficient research sample.
The benchmark is calculated from the current eligible published reports. It can change when another company is published, updated, reclassified or becomes too old for the comparison. Its sample details show the included companies and dates so readers can understand the basis of the comparison.
12
The certification badge answers whether an official Welo certification is present in Welo’s records. A research score answers a different question: how we assess the published policies and information examined for that report. Neither a high rating nor inclusion in a sector average grants certification.
An absent certification should not be read as a finding of fraud or misconduct. Equally, the research report should not be used as a substitute for checking the exact certification status, scope and relevant official records.
13
Policies and business practices can change. If you find a factual error, a changed policy or a source that no longer supports a statement, contact Welo with the report, the relevant passage and a current source URL. Specific evidence makes it easier to identify what needs rechecking.
A correction is evaluated against the sources and the same methodology used for other companies. New information can change a factual card, a finding, an area score or the comparison sample. Publication preserves a new version rather than silently rewriting historical evidence.
Use the Contact Welo form on a report to raise a question or request further research. Companies can also use it to submit supporting documents or clarify a policy. A request or commercial enquiry does not itself change a research result or certification status.