The method behind the number.

Public Trust Score: a provisional launch index

The circular Public Trust Score identifies either a historical source index with documented risk ceilings or a separately identified editorial assessment. It is separate from the LootGrade composite score below. It is not a verified current rating, a service-specific test, or an overall recommendation.

Convert each supplied Trustpilot snapshot from 0–5 to 0–100 by multiplying by 20. A historical editorial rating already on 0–100 keeps its scale. Add the available inputs, divide by their count, and round once to the nearest integer. Missing values do not become zero. When no input exists, display Not rated.

Example of the arithmetic only: a 4.0/5 snapshot becomes 80/100; with a separate 90/100 editorial input, the mean is 85/100. These example numbers do not describe a company.

Excellent is 85 to 100, Good 70 to below 85, Mixed 50 to below 70, Poor 30 to below 50, and Bad below 30. All displayed scores use these boundaries. The tag describes the index band. It does not establish safety or classify a business model. A marketplace label alone does not trigger an adjustment. A documented seller-dependent fulfillment assessment can limit confidence in a platform-wide score.

Historical index and editorial assessment

Historical source indices follow the calculation above. Editorial assessments are assigned values, separately labelled and dated, rather than outputs of that average. Each profile identifies the assessment scope, historical inputs and sources. Neither a historical index nor an editorial assessment is a paid-test composite.

Empirical scores require recorded purchases and the qualifying evidence described below. Price comparisons establish the quoted result for their stated sample; delivery and protection outcomes require their own records.

Risk ceilings take priority over historical averages

Version 1.1 of the provisional index applies two editorial confidence limits. A sourced review-platform notice that suspends a rating for review-integrity reasons caps the displayed score at 29/100, in the Bad band. A documented assessment of seller-dependent marketplace fulfillment, where a uniform execution standard has not been established by our research, caps it at 69/100, in the Mixed band.

These are published editorial rules, not measured probabilities or failure rates. Start with the historical average and apply the lowest relevant ceiling. A category label, unsupported allegation, payment, or ownership relationship cannot activate a rule. Each finding identifies the exact company domain, source, date, reason, and limitations. The same conditions apply to every provider.

Recheck risk findings within 30 days. An expired finding withholds the score until review; it does not silently restore a favorable historical rating. Sort lists by displayed score from high to low, place missing scores last, and break equal-score ties alphabetically. Neither this order nor a risk-adjusted tag qualifies an empirical winner or a provider link.

Each profile shows the historical base, risk ceiling when applicable, final displayed score, inputs, source date where known, receipt or capture date, and limitations. The supplied Trustpilot extraction lacks original records and extraction dates. Other publisher ratings may use different methods and populations; their independence has not been established. An equal-weight mean does not repair those limitations. Traffic estimates, domain age, SEO visibility, reported purchase counts, and unrelated market averages receive no points.

Protection, Transparency, and Red flags remain evidence fields. A policy check records what a provider publishes, not a measured outcome. No investigation means Not assessed, not zero problems. Repeated use of the same source material is not another independent observation.

The risk-adjusted provisional index cannot fill the empirical or public-evidence components below, establish a winner, or award a provider link. A current assessment needs fresh exact-domain sources, dated service-relevant evidence, and editorial review. Read the marketplace buyer checklist.

Two evidence components. One explicit formula.

The LootGrade composite score combines 70% empirical testing with 30% fact-checked public evidence. Every score belongs to a defined company, service, game, region, and assessment date. A company-wide reputation signal cannot establish which provider delivers the best service in a particular game.

Composite score = (empirical score × 0.70) + (public-evidence score × 0.30)

Both components use a 0–100 scale. Retain unrounded criterion and component values, then round the final score once to the nearest whole point. For illustration only: an empirical score of 80 and a public-evidence score of 70 produce 77/100. This is a hypothetical example, not a company result.

These percentages are an editorial policy, not a search-engine ranking factor. They preserve the value of original measurements while allowing documented public experience to influence the conclusion.

70%: what our own purchases establish

The five criteria below divide the empirical component. Their percentages add to 100% of that component, which contributes 70% of the final score. They do not add another 100% to the public-evidence component.

Criterion rubric and order aggregation

Assign 0 for a documented failure, 25 for major documented shortfalls, 50 for partial fulfillment, 75 for meeting the agreed requirement with a stated limitation, and 100 for fully meeting the documented requirements. Link the evidence and explain each value. An untested requirement remains blank. The terms criterion can assess the clarity of saved policies; refund performance stays unmeasured when no refund was exercised.

For each eligible order, calculate the weighted sum of the five criterion values. The empirical component is the arithmetic mean of those order scores. Publish the order count, individual results, range, incidents, and limitations alongside the mean. Do not discard a failed purchase or replace it with a successful one.

Before publishing an empirical component, require at least three ordinary paid orders for the same provider and service scope, on at least two different dates. Every scored criterion needs evidence. This minimum is a pilot sample, not a representative estimate of all customers or future orders.

30%: public evidence, checked rather than assumed

Publicly available information can establish advertised prices, published policies, domain identity, and documented reputation signals. It cannot, by itself, establish the delivery time or handling of a purchase we did not make. Trustpilot and other review sites are inputs to this component after a source and claim check; recognition or a high star rating alone does not qualify a source.

A reproducible fact-check record

1. Match the business. Confirm the company domain and profile identity. A similarly named business is not a substitute.
2. Record the source and scope. Save the URL, author or publisher, publication date when available, capture date, relevant service and game, and the exact claim being assessed.
3. Sample without picking a verdict. Use up to 50 consecutive newest customer reviews from the previous 12 months, stopping at the cutoff. Include every rating in that window. Record the available count and exclusions; do not sample only the best or worst reviews.
4. Classify the evidence. Record verification labels as supplied by the platform. A Trustpilot label is not a LootGrade purchase verification. Public reviews may concern an interaction rather than a purchase; private receipts we have not inspected remain unverified by us.
5. Check corroboration and conflicts. Compare specific claims with primary records and qualifying editorial findings. Check duplicate or syndicated content, shared operators, affiliate relationships, relevant company replies, and unresolved contradictions.
6. Publish the outcome of the check. Label a statement as supported, contradicted, or unresolved and link its basis. An allegation stays a reported allegation unless evidence establishes the underlying fact. Include corrections and the assessment date.

Public-evidence eligibility and rubric

A public component requires at least 20 eligible service-relevant customer reviews from that sample, one qualifying review publisher, and a dated primary-source check. The customer platform and review publisher must have different operators and no shared ownership with the company being assessed; commercial links must be disclosed. General reviews with no identifiable service scope can inform company context but cannot fill a game-specific score. If the scope or source minimum is unavailable, publish a qualitative evidence summary and leave this component unscored.

Assess each public criterion on the same 0, 25, 50, 75, or 100 rubric: documented failures, major shortfalls, mixed or partial fulfillment, supported fulfillment with a limitation, or consistently supported fulfillment. Record which agreed requirement each item addresses, the dated sample, the supporting and conflicting findings, and why that evidence supports the assigned value. More reviews or a recognizable publisher do not themselves earn a higher performance score. Calculate the weighted sum only when all three criteria have adequate evidence.

Capture inputs within 90 days of assessment and label older events separately. Imported historical star ratings and unverified editorial numbers remain attributed context until this protocol has been performed; they are not automatically converted into public-component points. Sources from another service or game do not fill a missing criterion, and the same order cannot be counted as both our purchase and external customer evidence.

Trustpilot explains its Verified review labels and reviewer guidelines. We assess the evidence for a claim, without claiming access to another platform’s private verification records.

Randomized purchases within comparable conditions

This is the protocol for future purchases. No randomized purchase campaign or paid test order has been documented by LootGrade yet. Randomization applies to ordinary purchase timing and eligible order options within a fixed comparison scope; it does not make a small sample representative of every customer.

1. Set the scope before ordering. Fix the game, region, service type, route, currency, budget, allowed options, and purchase window across the providers being compared.
2. Register the sampling plan. List the eligible purchase slots and options before seeing outcomes. Randomly select them and save the selection method, seed or draw record, and scheduled slots. Explain any constraints.
3. Buy through the ordinary flow. Pay for at least three eligible orders across at least two dates. Providers cannot choose which orders enter our sample or supply a special review order. Sponsored, gifted, or specially discounted review orders are separately labeled and excluded from this ordinary-purchase component.
4. Retain the original records. Save the quoted price, checkout total, amount paid, receipt, terms, payment time, first action, completion, support messages, and incidents, with time zone.
5. Keep failures and deviations. Record cancellations, unavailable routes, delays, and refunds. Log any deviation from the draw and its reason; never quietly replace an unfavorable result.
6. Review and publish. An identified editor checks the redacted evidence, criterion calculations, public-source assessment, limitations, and commercial disclosures before a score is released.

Public reviews show reported patterns over a wider set of experiences. Controlled purchases answer a different question: what happened when we bought this service under these conditions. This is why public research and empirical testing both belong in the model.

What can be published before both components exist?

A sourced public-evidence assessment may be published before our purchases are complete, clearly labeled as public research. An empirical result may be published with its own scope and limitations. Neither is a composite score. Missing evidence is not zero, and a missing 70% or 30% is never redistributed to the available component.

The composite score and overall recommendation remain unpublished until both components and their criteria qualify. Current research files contain no documented LootGrade paid orders and no public assessments completed under version 2.0; the existing imported snapshots do not fill those gaps. Inclusion in our catalogue is not a recommendation.

A first ranking requires at least five comparable services with eligible evidence, identified reviewers, and the same scope. Show the two component scores, the final calculation, sample sizes, assessment dates, and direct evidence links next to every conclusion. A score for one game or service is not a universal company score.

Limits, ownership, retesting & corrections

Only a documented comparison winner receives a provider link to the relevant game and service. A price-only award identifies its quoted configurations and result, without implying a paid-test verdict. Overall service recommendations require the composite evidence. Other providers are identified by domain and internal evidence files. Ties and missing evidence receive no recommendation link. Our winner-only links policy explains source citations, review dates, and commercial disclosures.

Our weights are fixed before collecting results and apply equally to every provider. Shared ownership between projects, sources, or companies must be disclosed. Different domains do not establish independent ownership. Affiliate links, sponsorships, and paid placement cannot contribute points or override evidence.

Check published sources monthly. Material changes to terms, prices, ownership, or fulfillment trigger a new assessment and, when needed, new purchases. A missing source is not an adverse result: mark the affected component unavailable until it can be checked. Keep the previous calculation, version, date, and reason for every revision.

A documented order does not establish universal safety or future delivery. A protection score cannot override a game publisher’s rules. Consult the current Riot Games terms and Steam Subscriber Agreement.

Corrections and evidence submissions follow the editorial contact process once configured. We apply the same standards to favorable and unfavorable findings and provide space for a documented response.