Most writing about linear TV measurement is selling something, which means it is vague about the limits. This guide is organized around them. What linear measurement can tell you is genuinely useful and worth paying for. What it cannot tell you is worth knowing before you buy, because a vendor who will not say it out loud is a vendor whose numbers you cannot check.
What "linear" means here
Linear media is anything that airs at a time, in a place: broadcast, cable, satellite, local and national radio, satellite radio, live reads, long-form and infomercial, and live streaming TV.
The common thread is not the screen. It is that each airing has a timestamp and a geography. Those two facts are what make measurement possible at all, which is why the same method also works on things that are not advertising in the usual sense: a billboard with a known installation date, stadium signage with a known game time, a news story that ran at a known hour.
It also sets the boundary. On-demand streaming, podcasts and streaming simulcasts cannot be measured this way, because there is no single moment when the audience saw or heard the ad. No timestamp, no measurement. That is a limit of the method, not a gap that a better vendor closes.
Why linear is harder than digital
Digital advertising measures itself. A click carries a link, a tag and a parameter, and the path from ad to outcome is recorded by the same system that served the ad.
Linear carries none of that. There is no link to follow, no tag to fire, no identifier that travels from the screen to the browser. On top of that sits an analog buying world: broadcast calendars, date adjustments, affidavits, invoices, and a publisher landscape fragmented across hundreds of stations and markets, each sending its paperwork in a different shape.
So linear measurement is inference, done carefully. Nothing about it is a count.
What can actually be measured
For every airing, the measurable question is: how much response happened that would not have happened anyway?
That response can be:
- Website activity — incremental sessions and new users, the top-of-funnel signal that moves first and most visibly
- Lower-funnel events — add to cart, view item, registration, download, form submit, purchase
- Phone calls — inbound calls matched to the spots that drove them, where call data is available
- Orders — commerce data where it can be pulled in
The usual recommendation is to pick two: one top-of-funnel and one near the bottom, up to three in total. More than that and you are reading noise as signal.
How the measurement works
The mechanism is the same regardless of medium.
- Build a baseline. Months of your own history establish what traffic normally does at that minute, on that day of the week, at that hour — an intra-week, intra-minute baseline that already accounts for spots that ran in the past.
- Align the airings. Every spot from the postlog is converted to the clock the response is measured on. An hour of time-zone error puts every spot against the wrong minute of traffic.
- Measure the lift. Response after the airing, minus what the baseline says should have happened, is the incremental lift attributed to that spot.
- Close the window when the data says to. Response windows are dynamic — they end when volume returns to normal, rather than at a fixed cutoff. Spots that air close together share credit.
- Keep local local. National airings are measured against national traffic. A local spot is measured only against traffic from that market, so markets do not contaminate each other.
That last point is the one most often skipped, and it matters: measure a local spot against national traffic and you are reading other markets' noise as that station's performance.
What can't be measured, honestly
This is the section most vendor pages leave out.
A single airing proves nothing. One spot is an anecdote. The method becomes trustworthy in aggregate — across roughly 100 to 200 airings, patterns by station and daypart are stable and worth acting on. A verdict on one spot is not.
Small spots against thin traffic don't resolve. If a site sees a couple of visitors an hour, and four ads run inside a thirty-minute window, there is no spike to find. Measurement needs enough baseline traffic that a response stands out from normal variation — as a rough guide, a site under about a hundred visits a day will struggle — and spots large enough to produce one, roughly $100 and up per airing. Hundreds of very cheap airings against a quiet site is the case where the honest answer is that this will not tell you much.
Geography is coarser than people expect. Google Analytics 4 reports US state, not metro area. Market-level detail across roughly 250 US markets is recoverable, but it takes a BigQuery export rather than GA4 alone.
Results mature. A week measured immediately and the same week measured two months later will not look the same, particularly where the purchase is considered over days or weeks rather than minutes. Early numbers are real but incomplete.
It will not match your vanity tracking. Offer codes, custom 800 numbers, QR codes, dedicated landing pages and "how did you hear about us?" dropdowns are each their own attribution system, and most respondents ignore all of them. People who saw a spot and then searched the brand will say they came from Google, because as far as they know they did. Those tools undercount badly, and a measurement that matches them is a measurement that has inherited their blind spot.
None of it is a count. Every number here is a good-faith estimate built on a baseline. It is far better than no feedback loop at all, which is what linear advertising otherwise has. It is not the same kind of object as a row in a payment processor.
How to sanity-check any linear measurement
You do not need to take a vendor's word for this, including ours. There is a back-of-the-envelope check anyone can run.
Take total response for the weeks before the campaign started and the weeks after. The difference in the top line should be roughly the same size as the sum of all the per-spot incremental lifts in the report. If the report claims far more lift than the top line moved, the baseline is doing too little work.
The clearest warning sign is a report that credits television with nearly all of your sales when you had sales before any television ran. That is not a measurement; it is a sponge, soaking up everything that happened after each spot without asking what would have happened anyway. Setting the baseline properly is the hard part of this discipline, and it is exactly the part that is easiest to skip.
What you get back
The deliverable is your own postlog, returned with new columns beside every line: incremental sessions, new users, calls or purchases, and cost per response for that airing. From there it pivots by station, network, market, program, daypart, day of week, hour and creative.
That turns the weekly question from "is TV working?" into something answerable: which stations and dayparts delivered response at a cost you are happy with, which delivered almost nothing, and which creative works where.
The answer is often uncomfortable. Typically 20 to 30 percent of a schedule turns out to be wasted, and the inexpensive station nobody is proud of frequently beats the marquee spot on cost per response. Small, repeated upticks at a low cost per airing add up faster than one visible spike that cost thousands.
Common questions
How much history do you need before you can measure anything?
Generally at least three months of granular analytics data, so the baseline reflects a real pattern rather than a few weeks of noise.
Do I have to install anything?
No. Measurement runs from read-only access to your existing Google Analytics 4 property and your postlogs. There is no pixel and nothing for a developer to add.
Can you measure a campaign that already ran?
Often, yes, as far back as the analytics data stays granular and retained. Historical analysis carries more caveats than live measurement — retention settings, dimensions that thin out over time, and site changes nobody remembers.
Does this work for radio?
Yes, with the same method. Radio tends to mean more, smaller airings across more markets, which makes the aggregate view more important and the single-airing view less meaningful.
Is linear TV measurement the same as TV attribution?
In practice people use both phrases for the same thing. Measurement is the broader word; attribution is the part that assigns response to a specific airing.
Getting started
Quality Analytics has measured linear TV and radio response for advertisers and agencies since 2014, across hundreds of campaigns and hundreds of millions of dollars of media. Send us your postlogs and give read-only access to Google Analytics 4; you will be set up the same day and have your first report the next business day. We run up to a month of your airings before you pay anything, and plans start at $399 a month, month-to-month. Book a demo to see it on your own data.