A clear verdict.
An open methodology.
You should be able to question the recommendation—and understand where every number came from.
Evidence comes first.
Catalog entries use identified manufacturer or retailer sources. Regional offers are observations from connected stores. Imported listings keep their source excerpts and await model and variant review. Specifications describe a product; they do not establish satisfaction or reliability.
Ownership responses start as unverified and enter a review queue. Only approved responses contribute to community statistics. Review does not verify a purchase or ownership. A statistic remains unavailable when its sample is too small.
Separate demo pages contain fictional sample statistics, including owner counts, failure rates and timelines. They demonstrate the interface and are never used as production evidence. This private site is excluded from search indexing.
The Buy Score: 0–100. Version 1.0.
We multiply each factor score by its weight, add the contributions, then round once. A higher “low ownership costs” score means fewer additional costs. Every factor needs adequate evidence before a full score is published. Missing evidence is never treated as zero, and remaining factors are not silently given extra weight.
Evidence coverage is the sum of weights with supported factors. For example, a known satisfaction factor supplies 25% coverage; it does not imply a score of 25. A full Buy Score needs 100% coverage. Features and competition provide context without extra hidden weights.
What makes a factor available?
Satisfaction uses the percentage of approved respondents choosing “yes” to buying again. “No” and “not sure” both remain in that denominator. Reliability and value use mean ratings from 1–10, multiplied by 10. Each needs at least 30 approved responses.
Long-term use is the percentage still using the product among approved respondents who report at least one year of ownership. This group needs its own minimum of 30 responses. It is a self-reported snapshot, not a prediction of how long a new purchase will last.
Price requires an in-stock offer checked within 15 minutes and at least 14 distinct observed days spanning 30 days within the last 90 days. History must match the exact retailer, model variant, country and currency. We take the median price within each day, then the median of those daily medians, so repeated checks do not give one day extra influence.
The price factor is 80 × observed median ÷ current item price, limited to 0–100. Shipping, duties and checkout discounts can change your final cost. Observed history is not a claim that we captured every price offered by the retailer.
Ownership costs need a documented 24-month total of required extra costs for the exact product, market and currency. Optional extras are shown separately and excluded from the factor. The factor is 100 − 100 × required extra costs ÷ current item price, limited to 0–100. Even a total of zero needs an explicit source assessment; an empty costs list does not establish that ownership is free.
Repairability uses a published score on a 0–10 or 0–100 scale, normalized to 0–100 with the publisher, source excerpt and date visible. Both assessments must match the canonical product and any relevant SKU and country, with a check date within the last year. Without those sources, the factors stay unknown. AI cannot fill the gaps by generating a plausible number.
A good product is not always a good purchase.
Score of 80+ and at least 10% below the observed median.
Score of 75+ with no warning rule triggered.
More than 10% above the observed median, even if owners like it.
Enough evidence, but a score below 75 with no other rule triggered.
Score below 50 or reliability below 40. This rule takes priority over price.
These rules apply only after all factors are supported. AVOID takes priority over other rules, followed by WAIT for an above-median price, GOOD DEAL, BUY, and MAYBE. Missing evidence produces “Not enough data yet,” not a recommendation to buy or avoid.
Confidence is separate from the score. Incomplete evidence is insufficient; a complete score with 30–99 approved respondents has limited confidence. A complete score with 100 or more has moderate confidence. Unverified ownership is never described as high confidence. These thresholds do not eliminate selection bias.
Owner evidence, with its limits visible.
Overall “still using” covers all reported ownership lengths and is not a two-year retention estimate. Long-term use has a separate denominator limited to respondents reporting at least a year. Failure, repair and return reports describe this survey sample, not the risk faced by all purchasers.
Demo topic sentiment uses topic mentions as its denominator. Its ownership timeline uses different respondents in each age group, not one group followed over time. Real longitudinal timelines and automated topic summaries need more evidence before publication.
Trust is earned, not implied.
Verified purchase, verified ownership, unverified ownership, and visitor opinion must remain distinct. Current contributions are unverified. Receipt verification is not available, and personal receipts or order information should never be included in comments.
Account writes are limited on the server, and one ownership response is stored per account and product. Edits return a response to the review queue. Moderators can approve, reject or requeue a response; private review history records those actions. Account owners can export or delete their data. A merchant cannot pay to change a Buy Score. See privacy and data controls.