TV attribution is the practice of measuring what each television or radio airing actually did for your business — how many extra website visits, sign-ups, calls or orders it produced — so you can keep buying what works and stop buying what does not.
For digital advertising that question answers itself. Every click carries a tracking code, and the path from ad to purchase is recorded as it happens. Television has no such link. Nobody clicks a TV commercial. Someone sees your spot at 8:14 pm, picks up their phone a minute later and types your name into Google, and nothing in that chain says "this came from the 8:14 airing on channel 7."
Attribution rebuilds that missing link statistically. This page explains how, what it can and cannot tell you, and what to expect when you start.
Why linear TV is harder to measure than digital
Three things make linear television and radio harder than any online channel.
There is no direct link. No pixel fires when someone watches a commercial, and no tracking parameter follows them to your website. The only evidence a spot worked is what happens to your own traffic, calls and sales in the minutes after it airs.
The buying world is analog. Television still runs on broadcast calendars, broadcast-day conventions that start at 5 am rather than midnight, affidavits and invoices. The record of what actually aired — the postlog — arrives as a PDF, a spreadsheet or a station-portal export, in whatever format that station or network happens to use, with times sometimes in the local market's zone and sometimes not.
The publisher landscape is fragmented. A single campaign can run across dozens of individual stations, cable networks and radio groups, each with its own paperwork. Before any measurement can happen, all of it has to be collected, cleaned, and put on the same clock.
That combination is why TV attribution is not a feature you can switch on in an analytics tool. It is a data problem first and a statistics problem second.
How linear TV attribution works
The method rests on one idea: compare what happened after a spot aired with what would have happened anyway.
1. Build the baseline
Your website, phone line or store already has a rhythm. Tuesday at 8:15 pm looks different from Saturday at 8:15 am, and both look different from Tuesday at 3 am. A proper baseline is built from months of your own data at that level of detail — the same day of the week, the same hour, the same minute — so that for any moment in time there is a clear expectation of normal activity. It also accounts for the spots that ran in the past, so previous advertising does not inflate what "normal" looks like.
2. Find the spike
When an airing produces a response, it shows up as a rise above that baseline shortly after the spot runs. The attributed lift for that airing is the total response in the window minus what the baseline expected. That subtraction is the whole point. A method that credits a spot with every visit in the fifteen minutes after it aired — without subtracting the traffic that would have arrived regardless — will always over-credit television.
3. Size the response window to the data
How long a response lasts depends on the spot, the audience and the hour. A fixed window of fifteen or thirty minutes is simple to implement and routinely wrong. A better approach lets the window end when your activity returns to its baseline, so a strong national spot and a small local radio read are each measured over the time they actually had an effect.
4. Share credit when airings overlap
Two spots ten minutes apart on different stations produce one blended rise. Crediting both with the full response counts the same people twice; crediting only the first ignores the second. When airings land close together, the expected response of each is modeled and the credit is shared between them.
5. Measure local ads against local traffic
A spot on a Phoenix station should be judged by what happened in Phoenix, not across the whole country. For local television and radio, response is measured against traffic from that market — the relevant DMA or state — only. National airings are measured against national activity. Without this, a busy evening in one market gets credited to an airing in another, and station rankings turn into noise.
What you need to get started
Two things, in most cases:
- Your postlogs — the record of what aired, when, on which station, with which creative.
- Read-only access to your Google Analytics 4 property. No pixel to install and nothing for your developer to do. Because GA4 has been collecting data since the day it was set up, campaigns that already ran can be measured too.
Response data from elsewhere can be used the same way when it carries a timestamp and a location: call-tracking systems, Shopify or Amazon order data, and similar sources.
What can be measured
The test is simple: did the ad air at a known time in a known place? If so, it can be attributed. That covers:
- Broadcast, cable and satellite television, national and local
- Local and national radio, including satellite radio and live reads
- Long-form and infomercial programming
- Live streaming TV that airs on a schedule
- Out-of-home and sports signage with a known schedule
- PR moments and news coverage with a known air time
What cannot be measured this way is anything without a fixed air time — on-demand streaming and podcasts, where each viewer or listener chooses when to watch. There is no single moment to measure from.
Which results to track
Most advertisers get the clearest picture from two or three measures at different depths of the funnel:
- Top of funnel: sessions or new users — how many people came to look.
- Middle: actions that show intent, such as add-to-cart, product views, registrations or downloads.
- Bottom of funnel: purchases, qualified leads or booked calls.
Each is attributed independently. There is no assumption that a website visit becomes a sale; each measure is read from its own data.
What the results look like
The output is your postlog — every airing — with the incremental response and cost per response added alongside each spot. From there it can be cut any way a buyer needs: by station, network, program, daypart, day of week, hour, creative and market.
That turns a question like "is late-night working?" from an opinion into a table. The comparison that matters is not cost per thousand impressions but cost per incremental response — what you paid for each additional visit, lead or sale a spot actually produced.
What advertisers usually learn in the first month
Every schedule is different, but a few patterns come up again and again:
- A meaningful share of the schedule is not producing measurable response — often in the range of 20 to 30 percent of airings. That is budget that can be moved.
- Something is outperforming. Almost every schedule has stations, dayparts or creatives doing noticeably better than the rest, and they are the obvious place to put that budget.
- Price and performance are not the same thing. Inexpensive airings on smaller stations often deliver a better cost per response than premium spots. A high-profile placement can produce a visible spike and still cost far more per response than a steady run of smaller ones.
Where attribution is strong — and where it is not
Attribution works best when there is enough signal to see. Two conditions help:
- Enough baseline activity. A site with a steady flow of visitors throughout the day gives the baseline something to work with.
- Spots large enough to move the needle. Airings of roughly $100 and up tend to produce responses that are clear and consistent.
It is weakest at the other extreme: a few visitors an hour, and hundreds of very small airings competing for them. When one visitor an hour becomes two, that is within normal variation, not a measurable response.
It is also worth being clear about what attribution is. It is a well-calibrated, good-faith estimate — not an audit. It is most valuable as a feedback loop: a weekly read on what is working, combined with a media buyer's experience and the rest of the marketing picture.
How to sanity-check any attribution result
Before trusting any vendor's numbers, including ours, run this check: compare the weeks before television started with the weeks after. The overall lift in your business across that period should roughly match the sum of the lift attributed to individual airings. If the spots add up to far more than your business actually grew, the baseline is too low. If they add up to far less, response is being missed.
And be wary of any report that credits television with nearly all of your activity when you were selling before television started. For a fuller checklist, see how to evaluate a TV attribution vendor.
Why self-reported sources never match
Vanity URLs, dedicated phone numbers, QR codes, offer codes and "how did you hear about us" questions are attribution systems too — but they depend on customers using them, and most don't. People search your brand name instead of typing the URL, call the number on your website, or pick the first option in a dropdown. Those sources under-count, and never line up exactly with statistical attribution. The gap is expected, not a sign that either method is broken. For phone response specifically, see how to attribute phone calls to TV and radio ads.
Next steps
Quality Analytics has measured linear TV and radio for advertisers and agencies since 2014 — hundreds of campaigns and hundreds of millions of dollars of media. We work from read-only GA4 access and your postlogs, with same-day onboarding, and we will run your first month of airings before you pay anything.