How We Compare Baby Products
CompareBabyStuff continuously compares baby products using manufacturer documentation, official manuals, safety databases, recall notices, retailer data, and structured product records. Every comparison on this site is a deterministic difference calculation over structured records — not a collection of opinions, and never a claim of hands-on testing.
Last updated: 2026-08-24. This page describes the exact method used on live comparison pages such as Joolz Aer² vs Bugaboo Butterfly 2.
The pipeline, step by step
- Product discovery. Candidate models are identified per category (strollers, car seats, monitors, carriers, high chairs, cribs, bassinets, pumps) from manufacturer lineups and authorized-retailer catalogs.
- Structured records. Each product becomes a record of named fields — weight, folded dimensions, weight limits, recline range, harness type, power, connectivity, price — where every field stores three things: the value, the source it came from, and the date it was retrieved.
- Normalization. Units are converted to one system, model variants are separated (an Aer² is not an Aer² Plus), and regional versions are kept distinct so US specs are compared to US specs.
- Deterministic spec diff. The two records are compared field by field. Identical values are marked equal. Fields with no verifiable source are shown as UNKNOWN — never estimated, never borrowed from a similar model.
- Safety pre-publish gate. Weight limits, installation rules, recall status, and certification claims are checked against primary sources (manual, manufacturer, CPSC/NHTSA). If primary sources conflict, no definitive claim is published — the conflict itself is what gets published.
- Verdict + independent verification. A transparent scoring model turns the diff into situational recommendations (below). A separate verification pass re-checks every factual claim against its cited source before anything goes live.
What "deterministic" means here
The same two product records always produce the same comparison output. Spec differences are computed, not narrated into existence: if one stroller weighs 14.3 lb and the other 16 lb according to their manufacturers' published figures, the table shows exactly those numbers with those sources and retrieval dates, and the lighter one is marked as the winner of that row. Nothing in the table depends on prose style, mood, or marketing language.
Category-specific decision criteria
Fields are weighted differently per category because families optimize for different things:
| Category | Deciding criteria | Typical tie-breakers |
|---|---|---|
| Strollers / travel strollers | Weight, folded size, newborn usability (full-flat recline / infant-car-seat mode), seat weight limit | Price, suspension, carry-on compatibility |
| Infant & convertible car seats | Rear-facing height/weight limits, installation method, harness positions, expiration/lifespan rules | Price, cover washability, side-impact documentation |
| Baby monitors | Video/audio privacy model (local vs cloud), range specification, battery, total multi-year cost incl. subscriptions | Resolution, night vision, sensor extras |
| Baby carriers | Weight range, hip-healthy positioning documentation, back-carry support, carrier weight | Price, airflow, newborn insert requirement |
| High chairs | Harness type, recline/adjustability, footprint, conversion options | Price, tray dishwasher compatibility, fold/storage |
| Cribs & bassinets | Juvenile-product compliance documentation, mattress size, conversion availability, weight/age limits | Price, footprint, accessory ecosystem |
| Breast pumps | Suction strength (documented), battery life, closed-system design, insurance/price model | Noise spec, parts availability, app features |
Transparent verdict logic
Verdicts on this site follow a fixed structure, visible on every comparison page:
- Row winners come from rules, not adjectives. Lower weight wins the weight row (within the same product class); longer rear-facing limits win the limits row; lower verified price wins the price row. Where a spec difference is trivial or unverifiable, the row shows no winner.
- "Buy X if…" maps situations to criteria. The recommendation paragraphs translate the computed diff into family situations: frequent flyers get fold-and-weight logic; daily sidewalk users get ride-and-seat logic; budget-first shoppers get verified-price logic.
- Tie-breakers are stated explicitly. When the computed diff is close, the page names the single deciding question (e.g., "child under 6 months now?" or "25+ flights a year?") rather than hedging.
- Verdicts never override safety rows. A recall flag or a conflicting-sources note stays visible regardless of which product the model favors.
How the picker quizzes work
Comparison pages include a short quiz ("Still deciding? Answer three questions."). Its mechanics are fully inspectable in the page source:
- Each answer option carries fixed numeric weights toward each candidate product — defined at build time from the same computed diff, not generated per visitor.
- The quiz runs entirely in your browser. Answers are not transmitted, stored, or profiled anywhere.
- The result names the product whose documented strengths match the priorities you selected, with the reasoning repeated in plain text (specs, not sentiment).
- If answers point both ways, the quiz says so and returns the same tie-breaker text as the written verdict.
What this method cannot tell you
Document-based comparison has hard limits, and pretending otherwise would make every other claim suspect. This site does not measure push-force, cushion softness, fabric feel, hinge longevity, or real-world battery drift — no hands-on testing is performed, so no such claims appear in verdicts. Where ride quality or comfort matters, pages say which manufacturer-documented design features (suspension type, wheel size, recline geometry) plausibly relate to it, and stop there. For physical impressions, consult retail return policies and try products in person.
Fail-closed defaults, conflicts stated
The pipeline is fail-closed by design: a field with no verifiable primary source publishes as UNKNOWN — never estimated, never borrowed from a similar model — and a page whose safety-critical fields cannot be verified does not publish at all. Where approved sources disagree (manual versus manufacturer page versus CPSC/NHTSA record), the rule is that conflicts are stated: no definitive claim is published, and the conflict itself appears on the page — “Current sources conflict. Verify the current product manual before use.” — instead of a quietly chosen winner.
No invented experts; disciplined safety language
No invented experts appear anywhere in this method: no fake pediatricians, fabricated CPSTs, phantom engineers, or manufactured parent testimonials. If no human expert reviewed a comparison, none is claimed — verdicts come from documented specification analysis, which is exactly what the AI methodology page describes. Safety language is held to the same discipline: medical promises (“prevents SIDS”), superlatives (“safest car seat”), guarantees (“guaranteed safe”), and unsupported crash-performance claims are rejected at the pre-publish gate. Pages describe what manufacturer documentation and regulator records actually establish, and stop there.
See the method applied: Joolz Aer² vs Bugaboo Butterfly 2 — filterable spec diff, situational verdict, quiz, and a sources list with retrieval dates. Related transparency pages: AI methodology, data sources, corrections.