Our review process is designed around repeatable, category-specific criteria rather than one generic star rating for every type of product.
1. We define what matters for the category
Before ranking products, we define the characteristics that matter for the category. A robot vacuum may be evaluated on cleaning, navigation, obstacle avoidance, mopping and maintenance, while a television may require an entirely different set of criteria.
2. We build a candidate set
We research current products, important models, major price tiers and products that serve different use cases. Discovery may use manufacturer information, retailer catalogs, published research and automated research tools. A product being considered does not mean it will be recommended.
3. We evaluate evidence and product data
We review available specifications, features, documented capabilities, pricing and other relevant evidence. When hands-on testing or original measurements are available, they can be incorporated into the evaluation. When they are not, we avoid implying that we personally tested the product.
4. We score against category criteria
Each category can have its own weighted review criteria. Product scores are calculated from those criteria so that the reasoning behind a recommendation can be inspected instead of being hidden inside an unexplained overall score.
5. We consider value separately from raw performance
A higher-priced product may offer stronger performance while a lower-priced model may be the better value for a particular buyer. That is why our guides can include awards such as Best Overall, Best Value, Best Budget or other use-case-specific recommendations.
6. We compare retailer offers
When retailer data is available, we compare current offers from multiple stores. Prices and availability can change quickly, so the retailer's checkout page is always the final source of truth for the transaction.
7. We update published content
Product lineups, prices and availability change. We may revise rankings, replace discontinued products, update scores, add new models and refresh retailer information when new evidence becomes available.