The 2026 CPAP Data Report

Finding one

Reading your own CPAP data works about six times out of seven

Of the 4,997 visits where someone started an import, 86.2% ended with their data on screen. The rest hit one of a short list of problems, almost all of them about which folder got picked rather than which machine they own.

The result that surprised us: 17.8% of those visits were on a phone or tablet, and they succeeded just as often as computers did. 86.4% on a phone against 86.2% on a computer. On this site, at least, a phone is not the harder way to read your own CPAP card.

Measured per visit, not per attempt. Someone who picks the wrong folder twice and the right one on the third try counts once, as a success, because they ended up with their data. Counting raw attempts instead would report that person as two thirds a failure and would put the success rate near 66.4%, which describes our funnel rather than anyone’s experience.

Finding two

What goes wrong is the folder, not the machine

55.7% of all recorded import errors are a single thing: nothing the site could recognize in the selected folder. Add the next few rows and the picture is consistent. Most failed imports are someone pointing at the wrong folder, exporting instead of copying, or handing over a settings subfolder without the therapy data beside it.

Shares divide by the 1,618 errors that recorded a reason, and that is fewer than the number of attempts that did not succeed. Roughly 1,510 unsuccessful attempts recorded no reason at all, so this taxonomy describes the failures we can name rather than every failure there was. We do not know what happened in those, and this report does not guess. Two of our own counters disagree by about one and a half percent here, and we divide by the one that classifies the errors rather than the one that merely counts them. Classes are grouped from the error text a reader was shown, so two wordings of the same problem count together. If you are stuck on one of these, the fix is almost always to copy the whole SD card rather than a folder inside it.

Finding three

Which machines turn up

ResMed accounts for 88.0% of every import where the brand could be identified, and a single model, the AirSense 11 AutoSet, is 54.6% of all identified machines. The long tail is real but thin.

Of the 7,757 attempts that got far enough to report a brand at all (the rest stopped before that point, which is why this is fewer than the 9,320 above), a further 1,411 (18.2%) held nothing identifiable at all and are excluded from the shares above rather than folded into them. Most of those are the wrong-folder problem from finding two.

By model

A card only names its model if its settings file came along, so a further 2,170 imports read their therapy data perfectly while telling us nothing about the machine. They are excluded from the shares rather than guessed at.

Finding four

When each brand became readable

The obvious chart to draw here would be brand share month by month, and it would be wrong. A brand appears in this data the month the site learned to read its card, so any apparent growth is our parser catching up, not anybody switching machines. What the dates below actually record is coverage.

First month an import of each brand was recorded. Brands the report withholds are not listed here either, because naming one would identify it as a withheld row and give away the count.

Finding five

At least one import in twelve brings an oximeter too

Of 4,519 visits that completed a CPAP import, at least 8.5% also brought a pulse oximeter and at least 7.3% brought a watch or ring. Both are floors rather than estimates.

A visit is a single browsing session, so somebody who imports their CPAP card tonight and their ring tomorrow counts as two visits and is invisible in these numbers. There is no account and no server-side identity here to join them with, and building one to improve this figure is not a trade this site makes. The true rate is higher than 8.5%; we cannot say how much higher, so we publish the floor.

Finding six

Where readers are

CPAP data help is overwhelmingly an English-language, North American and Australian conversation online. The page views below are a little broader than that, reaching 120 countries.

Counted in page views, not visits or people, because that is what the underlying figure actually is: one count per page loaded. It is the only section drawn from everyone who reads the site rather than from the smaller group who imported a card, so it is not comparable with the numbers above. Country is never crossed with machine brand or model. A handful of imports of an uncommon machine in a small country would describe a person rather than a population, and no finding here is worth that.

Method

How these numbers exist without health data

Reuse

Cite this report

The figures here are free to quote, republish and chart, with a link back. Published under CC BY 4.0.

CPAP Clarity. "The 2026 CPAP Data Report." Published September 2, 2026. https://cpapclarity.com/report/2026

Every figure on this page is also published as JSON and CSV, so you can chart it without retyping it. Both are the exact file this page renders from, including the withheld remainders.

Figures this report does not cover, including which wearables people pair with a CPAP and what joined nights actually show, are in the companion piece on what CPAP users import.

Questions about the method, or want a figure we did not publish? Ask on the about page. If you own a machine this report does not list, that is usually because the site cannot read it yet, and hearing from you is how that changes.