How we pick

1. Compile every published specification

For each of the 30machines on this site we read the manufacturer’s own published specification and its own product listing, on a dated visit, and record every dimension, weight, capacity, rail length, seat height, monitor and warranty figure they print.

Where a manufacturer publishes a separate specification page we cross-check the two. Concept2’s figures agree across both sources; that is not universally true, and where they disagree we say so rather than picking one silently.

2. Record what is not published

This is the part that makes the tables useful. Our spec type makes “not published” a first-class value rather than an empty cell, so a missing figure renders as Not published because the data says it is missing — not because somebody remembered to check.

Two findings from that, both of which every other publisher in this category quietly discards:

• Not one of the 17 manufacturers here publishes a decibel figure for any machine. Every noise number in circulation was produced by a reviewer with a phone at an unstated distance in an unstated room.
• 18 of 30 machines do not publish a maximum user weight, including one that costs over $1,700.

3. Compute the floor space

Published length is the machine, not the floor. Concept2 is the only manufacturer we found that publishes both: a 96 in by 24 in machine and a recommended 9 ft by 4 ft space. That is +12 in of length for handle travel at the catch and +24 in of width for elbows at the finish.

We apply that published relationship to every machine that publishes usable assembled dimensions, and we label the result as our arithmetic everywhere it appears. It is a calculation on a sourced figure, not a measurement, and nothing on this site describes it as one.

4. Compute the cost of ownership

Live machine price plus the manufacturer’s own published subscription rate, over 12, 36 and 60 months. Of the 27 machines we can price, 20 have no recurring cost at all; the rest run $29 to $50 a month.

We deliberately do not model electricity, shipping, mats, seat pads, water tablets or resale value. Each would need an assumption we cannot source, and including them would make the number look more precise while making it less true. The tables are here.

5. Rank, and say who should not buy

Rankings are editorial and argued in prose. We publish no numeric score, because a score implies an evaluation we have not performed — it would dress spec-reading up as testing.

Every machine carries a “skip this one if” line, including the ones we recommend most strongly. If we cannot name who a machine is wrong for, we do not understand it well enough to recommend it.

When we order a roundup, these are the criteria and roughly the weight each carries.

1. Fit for the page’s buyer (roughly 30%). A roundup for apartments ranks on footprint and noise mechanism. A budget roundup ranks on what is published at that price. The same machine can top one page and be absent from another.
2. Published specification quality (roughly 25%). Does the manufacturer publish a rail length, a capacity, a resistance figure? A brand that publishes them has had them rated, and it lets a buyer check rather than trust.
3. Five-year cost (roughly 20%). Machine plus subscription. A cheaper machine with a $50 membership frequently loses to a dearer one without.
4. Durability signals (roughly 15%). Warranty length, parts availability, frame construction, and whether a used market exists.
5. Monitor usefulness (roughly 10%). Whether the numbers it produces mean anything outside that machine.

These are weights, not a formula. There is no spreadsheet producing a score, because there is no measurement to feed one.

Every price on this site comes from a live Amazon feed, refreshed daily. There is no cached fallback and no hand-entered price anywhere in the codebase — the product type has no price field at all, so a price physically cannot be authored into a page.

Prices carry a hard expiry of 48 hours. If the refresh stops running, every number on this site disappears on its own: buy buttons revert to “Check price”, and cost tables stop rendering rather than showing stale arithmetic. The failure mode is silence, not a wrong claim.

Every rendered price carries the date it was captured. Amazon prices change without notice and frequently.

We write “Not published” and leave it there.

We do not estimate a decibel figure. We do not infer a weight capacity from a similar model. We do not take a specification from another review site without a manufacturer source behind it. And where a listing’s dimension field is plainly a shipping carton rather than the assembled machine, we discard the value rather than adjusting it — we reject numbers, we never alter them.

One documented normalization: where a listing reports dimensions in an order that would make a rower wider than it is long, we treat the largest figure as the length. That is stated here rather than applied silently.

Where a first-party source was unreachable at the time we looked, we say so and name the secondary source. Peloton’s All-Access price is the current example: Peloton’s own membership page did not return a figure to us on 24 July 2026, so the $49.99 is attributed to CBS News and Retail Dive and labeled as secondary.

Every figure on this site is sourced and dated precisely so that it can be checked. If you find one that is wrong — a misread specification, a figure that has changed, a discontinued machine still listed — tell us and include the manufacturer link if you have it.

We correct the page, update its “last verified” date, and note the change where it is material. A corrections policy nobody can act on is decoration; the contact form goes to a real address that a person reads.

Our editorial policy sets out how often each type of page is re-verified — prices daily, subscription figures quarterly, cost pages quarterly, spec tables on model revision.

I'm an enthusiast, not a coach and not a lab. I read the manuals, compile the published specs, and do the arithmetic. I haven't rowed on most of these machines, and I won't pretend I have. What I can do is put every published number in one place, show where each one came from, and work out what these things actually cost to own.

More about the site and why it exists.