How we score durability

FirstSqueak asks owners how their things are doing months and years after buying them, not just in the first week. This page explains exactly how those answers become the numbers on product pages. The method is the same for every product, and nobody can pay to change it.

What owners tell us

Each owner writes a review, then checks in 1, 6 and 12 months after the purchase month, and yearly after that. They can also report a problem at any time. Each check-in records:

• whether the product is working fine, has problems, or is broken;
• if there were problems, what went wrong (from a fixed list of issue types) and whether it was repaired;
• whether they would buy it again, and how satisfied they are.

Only the purchase month is recorded, so time is measured in months since the start of that month.

When a product "breaks"

A product counts as failed when an owner reports it broken and not repaired. If a later check-in shows it working again, it doesn't count as failed: repairable products get credit for being repairable. "Has problems" is not a failure; it shows up in the issue breakdown instead.

Owners whose product hasn't failed count as still working up to their most recent check-in. We don't assume anything about the time after that. (Statisticians call this right-censoring.)

Survival estimate

We estimate the share of units still working at 1, 2 and 3 years using the Kaplan–Meier estimator, the standard method for survival data where people join and leave at different times. At each point in time where failures were reported, the estimate is multiplied by the share of units still being followed that didn't fail then.

• Weights. Owners' answers are weighted by how much we trust the account: verified and long-standing accounts count more than brand-new ones (see "Trust" below). Weights are bounded, so no single account can dominate a product's score.
• Uncertainty. Every percentage comes with a 95% range, computed with Greenwood's formula on a log–log scale so the range always stays between 0% and 100%. With weights we use Kish's effective sample size: unequal weights carry less information than the same number of equal ones, and the ranges honestly get wider. While no failures have been reported yet, the lower end of the range comes from the exact binomial bound, so "no failures among 6 owners" doesn't look like certainty.
• Enough owners. A time point is only shown once at least 5 owners have been followed that long.

When a product gets a score

A product gets a durability score once at least 5 owners have checked in 6 months or more after buying it. Until then its page says "No long-term data yet", with how many owners have reviewed it and how many are still in progress.

Other numbers

• Median lifespan: the time by which half the units had failed, shown only if the estimate actually drops that low.
• Repair rate: of owners whose product broke, the weighted share who got it working again. Shown once at least 3 owners reported a breakdown.
• Would buy again: each owner's most recent answer, weighted. Shown once at least 3 owners answered.
• Common issues: the most reported issue types, as a share of owners who reported any problem.

Every number is shown with the number of owners behind it. We never show a headline number without its sample size.

Ranking ("most durable")

Category pages list scored products first, ranked by the lower end of the 1-year survival range. A product with 40 owners and 95% still working ranks above one with 5 owners and 100%, because we're less sure about the second.

What doesn't count

• Reviews that are held, hidden or removed by moderators, and accounts that are banned.
• Contributions quarantined by our abuse checks, until a moderator clears them.
• Comments and "helpful" marks never affect durability scores.

When moderation decisions change, scores are recomputed from the raw data. Scores are also refreshed shortly after every new check-in and in full every night.

Trust

Every account starts as an unverified reviewer, whose reports count at a reduced weight. Adding a passkey verifies the account. Trusted reviewers earn their status automatically through months of consistent, on-time follow-ups plus a good trust score. Helpful marks from other trusted members add a small, capped bonus that can't be farmed. Each owner's weight stays between 0.1 and 1.5, so no single account can dominate a score. Details: how trust works.

No money involved

FirstSqueak has no affiliate links, doesn't sell placements, and has no brand accounts. Shop links on product pages are plain links with our tracking removed.