How we count owner reports
The short answer
Every number here is a count of owner reviews: collected from retailers, de-duplicated, and read by an automated classifier checked against a sample where a person decided every disputed call. Promotional reviews never count toward failures, and a problem needs at least 10 owner reports before it gets a page.
This site analyzes what dishwasher owners report. It isn't a review site: we don't test, review or rate dishwashers. This page explains exactly how the counts are made, and what they can and can't tell you.
Where the reports come from
Owner reviews from Home Depot (6,805 reviews collected). For the 100 Series, 300 Series and 800 Series we haven't collected every review yet. Our sample includes extra one- and two-star reviews and reviews that mention error codes or words like "leak" and "drain", so for those series the numbers describe what complaints are about, not how often the dishwasher fails overall. Each page says which kind of count it shows.
SaferProducts.gov incident reports are not included yet: the public API was unavailable when this data was built.
Removing duplicates
Manufacturers often copy the same review onto several retailer sites. Counting each copy would inflate the numbers, so we match identical and near-identical reviews (same rating, posted within a few days, almost the same words) and count each review once. Of 6,805 reviews collected, 3 were copies, leaving 6,802. 4,083 of those were originally written on another site.
Reading each review
Each review is read by an automated classifier that answers a fixed set of questions: does it describe the dishwasher failing (rather than a dislike), is the problem the dishwasher itself (not delivery or installation), was the review part of a promotion, how long did the owner have it, which part failed, and what fixed it.
Reviews collected through promotions or free products are left out of every failure count. They only count toward design complaints.
Classifier: jev-1.13.0. Question set v2.3. Its answers have been checked against a sample of reviews: a second AI model read each one without seeing the classifier's answers, and a person decided every review where the two disagreed, plus a random sample of the ones where they agreed.
Download the counts
Every published count is in one file: dishwasher problem counts (CSV). You can reuse it under the CC BY 4.0 license if you credit Appliance Problems and link here. It holds the same numbers as the pages, and nothing below the minimum.
Energy and water figures
Energy use, water per cycle and capacity come from the US Department of Energy's certification database, where manufacturers must file official test results for every dishwasher sold in the US (captured 2026-10-02). The ENERGY STAR mark is shown only for models on the current certified list (captured 2026-10-02); a model missing from that list may have been certified in the past, so we never say it wasn't. If the official records for a model disagree, we show no figures for it.
Safety recalls
Series and model pages check our model numbers against the 22 dishwasher recalls in the Consumer Product Safety Commission's recall database (captured 2026-10-05). The CPSC lists recalled models by how the model number starts, so that's what we match on. Some recalls only cover certain serial numbers, which we can't check, so every page sends owners to cpsc.gov/Recalls to check theirs.
How strong is the evidence?
Data pages show an evidence-strength meter. It measures how much evidence sits behind a number, not how good or bad the dishwasher is:
- Limited: 10 to 24 owner reports. Enough for a page, but treat it as an early signal.
- Moderate: 25 to 49 owner reports.
- Strong: 50 or more owner reports, read by a classifier that has been checked. Without that check, nothing is rated strong.
Strong evidence still describes what owners complain about. Where our sample includes extra low-star reviews, the meter says so: a count of complaints is not a failure rate.
Ratings without promotional reviews
Many retailer reviews are written in exchange for a free or discounted product, usually within weeks of installation. They rate higher than other reviews and come too early to say much about reliability, so we never count them as breakdowns. We also show what a series' star rating would be without them.
Because we read more one- and two-star reviews than a random sample would include, averaging the reviews we read would understate the rating. Instead, for each star level we measure what share of the reviews we read were promotional, remove that share from the retailer's own count of reviews at that star level, and average what's left. It's an estimate, and we only show it when we've read at least 20 reviews at every star level.
When we publish a number
A problem only gets its own page when at least 10 owners report it for that series, out of at least 50 reviews we read. Below that, we don't publish a page at all.
How to read the counts
- An owner report count is the number of separate owners who described that problem, after duplicates are removed.
- We always show the count next to what it's out of, never a percentage on its own.
- Timing is what owners say ("stopped after 14 months"), not a measured lifespan.
- Counts describe reports about a series, not a defect finding about any product.
Data last updated Oct 2026.