Open methodology
A score you can question.
DoxaIQ uses deterministic rules—not a generative AI model—to organize public evidence. Every conclusion is designed to be inspectable and reproducible.
We are
A post-purchase evidence scanner for recurring public ownership patterns.
We are not
A lab test, representative owner survey, or generated AI answer engine.
Best used with
Professional tests, specifications, safety notices, and your own priorities.
1. Evidence collection
Queries combine the product name, model aliases, and ownership-language terms. Public videos, available transcripts, comments, dates, engagement metadata, and explicit sponsorship markers are collected.
2. Cleaning and deduplication
Canonical platform IDs, normalized URLs, near-identical captions, and repost fingerprints prevent the same claim from being counted repeatedly.
3. Deterministic classification
Curated keyword and phrase taxonomies tag failure modes, support experiences, buyer regret, would-buy-again language, owner age, and disclosure signals. Negation and proximity rules reduce false positives.
4. Weighting
First-person owner evidence receives the highest base weight. Recency, ownership duration, corroboration across independent creators, and source completeness adjust it. Sponsored or loaned-product content remains visible but is scored separately.
5. Reliability score
The 0–100 score combines issue severity (35%), issue prevalence (25%), persistence over time (15%), support outcomes (10%), buyer-regret signals (10%), and evidence confidence (5%). A low-confidence report cannot present a precise score as definitive.
6. Product identity checks
Brand, model number, and suffixes such as Pro, Max, Plus, Mini, or XL must match. Conflicting variants, accessories, and near-duplicate results are rejected before scoring.
7. Trend detection
Rolling 90-day issue share is compared with the preceding 90 days. A trend appears only after a minimum evidence threshold and cross-source corroboration.
Important limits
Public social evidence is not a random sample of all buyers. People may post disproportionately about excellent or poor experiences. Classification can miss context, transcripts may be unavailable, deletions create survivorship bias, and platform disclosures are incomplete. DoxaIQ therefore shows confidence, sample size, source mix, and the underlying evidence.