How we test dog harness
We are a shop, not a laboratory. What follows is exactly what we do and do not do, so you can weigh our recommendations accordingly.
Our criteria
- We read the primary research, never the summary. Every study cited on this site was opened and read at the source — the journal page or its PubMed Central record — not taken from a search-engine summary or another shop's blog. Where a paper states a limit on its own finding, we repeat that limit. Shih et al. (2021) found their effect with food rewards and not with toys; we say so on the home page rather than quoting only the half that helps us.
- We reject numbers that have no study behind them. While researching this site we repeatedly found the claim that a front-clip harness cuts pulling by 70–90%, and a second claim of a 45% reduction. Both trace back to commercial blogs citing nothing. They are not on this site. The figure we do publish — that a front-clip harness showed no significant difference from a metal prong collar across 23 dogs — comes from Johnson & Wynne, PeerJ 12:e18131.
- We cross-check size charts against real buyer measurements. Published charts are written conservatively. We read every review that named both a size and a real measurement — a chest circumference or a body weight — and compared them. For the front-clip harness, two buyers fitted 35 kg Labradors into a size L that the chart caps at 30 kg, both reporting spare adjustment; one buyer at 38 kg found XL loose. That is why our advice is to size down when you are between bands, which is the opposite of the usual.
- We keep the critical reviews. We read the one-star and four-star reviews as carefully as the five-star ones, and the recurring complaints appear in our FAQs rather than being quietly dropped. The hardware question is the clearest example: several buyers describe the rings as plastic, others describe metal at the chest. Both versions are on the product pages.
- We record where every product comes from. Each harness we sell has a written sourcing record: supplier, product identifier, store name, the cost we paid, and the date it was checked. It is how we can tell you that two listings we compared were the same physical product — their supplier images matched byte for byte — rather than guessing from the photographs.
The limits of our method
We have not run our own instrumented pulling trial, and we are not going to claim one. We have not yet stress-tested the hardware to failure. We have not measured this specific harness against a named competitor. When we write that a front attachment redirects a pulling dog, that is the published research on front-attachment harnesses in general, not a measurement of this product. Every order carries a 30-day money-back guarantee, so the risk is on us, not you.
The studies this site relies on
Five papers underpin every claim we make about pulling. Here they are in full, so you can check them directly.
| Study | Sample | What it found |
|---|---|---|
| Shih et al., Frontiers in Veterinary Science 8:735680 (2021) | 52 shelter dogs, 16–43 kg | Peak leash tension 16.87 N on a back-clip harness vs 6.97 N on a collar; time spent pulling roughly doubled |
| Johnson & Wynne, PeerJ 12:e18131 (2024) | 23 dogs, four equipment types | Front-clip harness cut pulling vs a martingale (p = 0.001) and matched a metal prong collar |
| Pauli et al., JAAHA 42(3):207–211 (2006) | 51 eyes of 26 dogs | Intraocular pressure rose significantly through a collar, not through a harness |
| Grainger, Wills & Montrose, J. Vet. Behavior 11:60–64 (2016) | 30 pet dogs | No measurable stress-behaviour difference between collar and harness walking |
| Townsend et al. (2021), cited in Johnson & Wynne | Owner survey | 82.7% of owners report their dog pulls on the lead |
How the size cross-check works
For each harness, we read every review that named both a size and a real measurement — a chest circumference or a body weight — and tabulated them against the published chart. The sample is small by research standards and we publish its size rather than implying a study: seven usable data points for the front-clip harness, two for the personalised vest.
What that sample showed was consistent enough to change our advice. The chart runs generous at the top end, so between two sizes we recommend the smaller one. That recommendation contradicts the published instruction, which says to take the larger size in an overlap.
Reading the reviews
We read the one-star and four-star reviews first. Recurring criticism goes into the FAQ of the relevant page, in the buyer's own terms. A complaint that appears once is noise; a complaint that appears four times across two separate listings is a product characteristic.
Photographs are matched to their own review by identifier, never paired by position in a list, and deduplicated by file hash so one buyer cannot appear twice under two names.