How Kaiway measures

What Kaiway records, what it doesn't, what you agree to, what you can take back, and what we publish.

Last updated 23 September 2026. When this page changes, the date changes and the change is listed at the bottom.

The short version

Kaiway is free and will stay free. It pays for itself by selling aggregate measurement: what households watch overall, never who watched what.
Your household is only included if you say yes, and only viewing from after you say yes. Section 5 covers what saying no means.
You can stop at any time. When you do, your household drops out of all measurement, including past periods.
Children's viewing is never published, never sold and never broken out, in any form. That's why TV viewing isn't included at all for households with a child's profile.
Nothing has been published yet. The panel is small, invitation-only and not representative of the country. We'll say so on anything we publish.

1. Where things stand today

Kaiway is a desktop app for Windows and for Macs with Apple silicon. Intel Macs aren't supported yet; we've only built for Apple silicon. It keeps one list of shows across the streaming services you already pay for, and opens each one where it plays.

The group of households whose viewing is measured (the "panel") is small and invitation-only. Everyone in it asked to join, so it isn't a random sample: people who go looking for a new desktop TV app aren't typical viewers. Invitations control only the order people come in, which we balance by state and household size. We describe who's in the panel with the four questions below.

We have not published any figures yet. We'll publish once the panel is large enough for the numbers to mean something. The rules on this page apply from the first figure onwards.

2. What Kaiway records

Kaiway records these things for every household, whether or not you agree to measurement, because the app needs them to work. They are how it remembers where you got to, what's next, and your children's time limits.

What was opened: the service, the show or film, and the season and episode when Kaiway knows it. For live channels, the channel only, not the programme.
When: when it was opened in Kaiway, when it finished loading, and when it was closed.
How it ended: for example, you finished it, you closed it early, the service showed an error, the service asked you to sign in, or the app was quit.
Which profile: which person in the household it was for, and whether that is a child's profile.
Which version of Kaiway recorded it, so a change in how the app works can't be mistaken for a change in viewing.

Your measurement choice is stored with your household, not on each of these records. That's why changing it takes effect for your past viewing too (see section 6).

Consent decides whether any of this is used in measurement, not whether the app keeps it. Most people would assume the opposite, which is why we say it here.

3. What Kaiway does not record, and cannot see

Whether anyone is actually watching. Kaiway knows how long a show was open on screen. It can't tell whether anyone was in the room, and it can't see where in the show you are. Hours we report are hours on screen, not hours watched, so they are an upper bound.
Viewing outside Kaiway. If you watch a service in its own app, on your phone, or on your TV, Kaiway doesn't see it. The one exception is the optional Roku observation described below, which is off unless you turn it on.
What you watch on your TV. Even with Roku observation on, Kaiway sees which app is open on the Roku and for how long, never which show. Roku doesn't provide that, and we don't try to get it another way.
Who you are, beyond four questions. We don't ask for your gender, income, date of birth or ZIP code. See the next section for why.

4. The four household questions

After you install, Kaiway asks four questions. Each one has a "prefer not to say" option, and skipping any of them doesn't keep your household out of the panel.

We ask	Why this and not something finer
Age band	A range like 35–44, never a date of birth. A range is enough to describe the panel and doesn't identify anyone.
Household size	How many people live with you.
State	Not ZIP code. ZIP plus age plus household size can point to a single home. State can't, and it's all the panel needs.
Number of TVs	Kaiway sees at most one screen in your home. Knowing how many TVs there are lets a figure say "seen on 1 of 3 TVs" rather than quietly assuming 1 of 1.

What we deliberately don't ask. We don't ask gender, because we would rather measure what a household watches than guess from who they are. We don't ask income, and we don't ask which services you subscribe to, because Kaiway can see which services you've connected.

The waitlist asks for your state and household size too, so invitations can be balanced. Those waitlist answers are kept separately and are never used as your household's answers. The app asks again after you join.

There are three separate decisions. Each one is asked on its own, and none is bundled with another:

Measurement. Whether your household's viewing is included in aggregate measurement. Asked when you first set up the app. "Include my household" and "No thanks" are the same size, side by side, and neither is pre-selected.
The four questions above, each answered or skipped on its own.
Roku observation. Whether Kaiway may check which app is open on your Roku, and for how long, while Kaiway is running. This is only offered in the Kaiway phone app and only after you've set up a Roku. The desktop app never observes your TV. If your household has a child's profile, your TV viewing isn't included in measurement even if you say yes. See section 7.

Only viewing from after you say yes is included. Viewing from before that, including any time when you had said no, is never part of measurement, even though the app keeps it for its own use.

Today, a household that says no keeps the whole app, including its history and Up Next. That's one of the terms we're still settling: if you're invited, you'll see the final terms before you install, and if you already use Kaiway, we'll tell you before anything changes for your household.

Until you answer the measurement question, your household is treated as not included. Never asked is not the same as yes.

6. What you can withdraw, and what happens when you do

Measurement

Turn it off at any time in Profile ("Stop including my household"). From that moment:

Your household is left out of every future figure, including figures about periods when you had said yes. Withdrawing isn't only "from now on". It takes your past viewing out of measurement too.
Figures already published before you withdrew can't be recalled from whoever received them. They won't be repeated with your household in them.
Today, the app keeps working the same way and still remembers where you got to. Section 5 covers what may change.

You can turn it back on later, and that is recorded as a new decision. As with a first yes, only viewing from after that new yes is included. Viewing from the earlier period stays out.

Roku observation

Turn it off at any time in Services in the phone app. Kaiway stops checking your Roku, and the TV viewing it already recorded leaves measurement too, including past periods. It works the same way as measurement, and your in-app viewing isn't affected. If you turn it back on, only TV viewing from after that is included.

Your answers to the four questions

Email us to change or remove any answer. Removing one means your household is no longer described on that dimension, and the figures that use it get a smaller base.

Everything

Email us to have your household deleted. Today this is done by hand, not by a button in the app. What's left afterwards names nobody:

a record that one invitation was used, so the list of who we invited stays accurate; and
the household count on figures that were already published while you were included ("based on N households").

The waitlist

Email us to be removed from the waitlist. Waitlist details are used only to choose who to invite and to send the invitation. They aren't part of measurement, and they aren't sold or shared.

7. Children

Children's viewing is never published, never sold and never broken out. Not "anonymised", and not "carefully aggregated": it is left out of every external figure, in any form.

Kaiway does record what's watched on a child's profile, because that's how it keeps their place and applies their time limits and bedtime. Every record is marked, at the moment it's written, as coming from a child's profile or not. That mark can't be changed afterwards, and records marked as a child's are left out of anything we publish or sell.

TV viewing in households with children

If your household has a child's profile, none of your TV viewing is included in measurement. Roku observation can see that an app was open on the TV, but not who was watching. In a home with children, some of that TV time is children's viewing, and there's no way to separate it out. Including it would break the rule above while looking as if it didn't.

This applies if your household has ever had a child's profile in Kaiway, even one that's since been removed. It also applies to TV viewing recorded before the child's profile was added. Your in-app viewing on adult profiles is still included if you said yes.

8. What we publish, and what the numbers mean

Anything we publish comes only from households that said yes to measurement, only from adult profiles, and never includes our own testing of the app, which is flagged and left out.

We publish totals and proportions across households, never one household's record. Every figure comes with these alongside it:

Who the panel is: invited desktop households on Windows and Apple silicon Macs (Intel Macs aren't supported yet), not the general population, and the number of households behind the figure.
Hours are hours on screen, not hours watched.
Three measures, always together: how often something was started, how often it stayed on screen long enough to count as more than a sample, and how often it was finished. None of them is published alone, because each one on its own is misleading.
Within-panel figures, not national audiences. The figures we can stand behind look like "among Kaiway households that watched show A, a third also watched show B". "Eight million people watched show A" is a projection to the whole country, and we won't present one as a measurement.
Household descriptions rest on the households that answered. A figure broken out by, say, state uses only households that gave a state, and says how many that was.
TV viewing is by app, never by show, and only for households that use Roku observation and have no child's profile. For every other household it's unknown and reported as unknown, never as zero. So TV figures describe homes without children in Kaiway, and we'll say so alongside them.
Small groups. They're how aggregate numbers end up pointing at individuals. There's no minimum group size yet; before we publish anything, we'll set one, build it into how figures are produced, and state it here.

When we find a problem in the data, such as a bug that recorded something that didn't happen, we mark the affected records and leave them out. We don't delete them, so the correction can be checked.

9. If Kaiway changes hands

If Kaiway is ever sold or merged, the measurement data would be part of what changes hands. We'll update this page and tell households before that happens.

10. Contact and changes

Questions, corrections, removal or deletion: [email protected]

Changes to this page

23 September 2026. First version.