1. Lock the exact product
We start with the exact ASIN or catalog identifier, brand, model, part number, variation, pack count, size, color, and image. Closely named products are not merged when their specifications differ.
2. Prefer primary evidence
Manufacturer product pages, manuals, specification sheets, warranty terms, regulators, and recognized standards bodies are preferred. Retailer fields are used for exact-offer identity and are labeled when they conflict with primary documentation.
3. Read real owner experiences
For each completed review, we search multiple accessible owner-review sources for recurring use contexts, positives, complaints, failures, and delivery issues. Exact-model feedback is preferred. Family or variant feedback is identified separately. Incentivized reviews, pooled variants, AI-generated retailer summaries, suspicious repetition, and small samples are disclosed rather than hidden.
4. Synthesize, do not copy
Owner comments are summarized as patterns. A single anecdote is not turned into a product fact, and review averages are not reproduced as TopFiveFind test scores. User reports cannot prove safety, certification, material composition, measured performance, or long-term reliability.
5. Reconcile conflicts and gaps
Conflicting dimensions, electrical ratings, capacities, model names, warranty terms, and feature claims remain visible. Missing evidence is stated as missing. We do not fill gaps with a related model or the most favorable number.
6. Add decision support
Each completed product page explains likely fit, important limitations, checks before buying, direct answers to common questions, and the sources used. Guide pages stay out of search indexing until all five selected products pass the completion gate.
7. Use automation transparently
Catalog collection, matching, consistency checks, research organization, and drafting are AI-assisted. The system is useful for comparing large evidence sets and flagging contradictions. It is not presented as a human expert, owner, engineer, doctor, technician, or laboratory. Publication checks cover identity, sourcing, unsupported claims, metadata, structured data, internal links, redirects, and desktop/mobile rendering.
8. Understand the score
A catalog score is a comparison aid based on feature fit, specification fit, use-case fit, value signals, and evidence quality. It is not an average of customer ratings, a laboratory grade, or proof that a product is objectively best.
9. Correct and update
Substantive updates should be tied to new evidence, corrected identity, changed documentation, or completed editorial work. Dates are not changed merely to appear fresh. Readers can submit a correction with the page URL and supporting source through the contact page.