How products earn a place on the list

Cradlebase is not a catalog. Products enter the list only by passing four filters, in order, and most don't make it. The catalog was built by filtering a real 99-item baby registry down to what families genuinely used — it grows only when a new product passes all four filters, never for volume.

Filter 1 — Lived, not tested

Every product was used by at least three families through a complete 0-to-24-month cycle. A weekend of testing tells you how a product demos; two years tells you how it holds up to daily use, machine washes, a second child, and a move. If a product has been on the market for less than 18 months, it doesn't qualify yet — no matter how good the launch reviews look.

Filter 2 — Reviews as filter, not gospel

We analyze 40,000+ verified reviews per category to detect systematic failures. We never look at the average — everything has 4.7 stars. We look at the distribution and the temporal curve: if month-18 reviews drop relative to month-3 reviews, there is a durability problem the average is hiding. That pattern has disqualified more products than any other signal.

Filter 3 — AI last, not first

We use language models (Claude, by Anthropic) to process review volume no human could read and surface thematic patterns: the strap that frays, the motor that weakens, the zipper that fails in cold weather. The AI identifies patterns; humans decide. Cradlebase is not an AI app — the AI is a lens, not a judge.

Filter 4 — One recommendation per category

If there are three reasonable options, we pick one. The other two appear on the product's page under “What we considered and rejected,” each with the concrete reason it lost and who it would still be right for. A list with three “great options” per slot is a catalog with extra steps.

One more thing: no brand pays to appear here. Our only revenue is affiliate commissions, paid by retailers at no cost to you, and they never affect the picks — the rejects on every product page are the proof.