🗃Compiled automatically from public product data — the attributes on this page are assembled from the sources we hold, not written by a person or drafted by AI, and individual pages are not reviewed one by one. How we use AI →
Did product recalls really rise? Separating the trend from the data
Original analysis · updated 26 October 2026
Product recalls did not rise 41% over the last decade. That figure is what you get by counting every record we hold — but we began mirroring several registers part-way through the period, so the curve partly measures our own history. Counting only the 9 registers with continuous coverage every year from 2016 to 2025, the real change is +7.4% — with a 27% collapse bottoming out in 2021, and a recovery that has since passed pre-pandemic levels.
The two series side by side
Year
All records held (misleading)
Continuous registers only (comparable)
2016
11,619
10,762
2017
11,587
10,476
2018
11,114
9,767
2019
11,371
9,904
2020
9,477
7,853
2021
12,582
7,822
2022
12,501
8,100
2023
14,795
10,651
2024
15,513
10,786
2025
16,387
11,556
Why the two differ
A register only enters the right-hand column if we hold at least 50 records from it in every year of the window. France’s RappelConso, the UK’s OPSS and the US NHTSA all begin part-way through, so including them makes recalls appear to surge at the moment we started mirroring them. This is the most common way a recall chart goes wrong, and it is why we publish both columns.
Registers in the comparable series
EU Safety Gate (RAPEX)
Health Canada
US FDA — medical devices
US FDA — food
US FDA — drugs
US CPSC
UK DVSA
Australia ACCC
OECD GlobalRecalls — Japan
Method
Counting recalls by the year the issuing authority dated them. A register only enters the series if we hold at least 50 records from it in EVERY year 2016-2025 — otherwise the curve measures when we started mirroring a register rather than what regulators did. 2026 is excluded as incomplete.
Limitations
Counts are by the date the issuing authority gave the notice, not by when we ingested it.
2026 is excluded because the year is incomplete.
A recall is one published notice. Registers differ in whether related notices are grouped, so absolute counts are not perfectly comparable between registers — but each register is compared only against itself over time.
Reuse
These figures may be quoted freely with attribution to ProductGuru and a link to this page. If you find an error, please tell us — we will correct it and say so.
ProductGuru aggregates public product-safety notices from government registers. We reproduce no assessment of our own; always verify a notice at the issuing authority.