Most people searching for a TV attribution vendor already have one. They are looking because something went wrong — and after twelve years of onboarding advertisers and agencies who arrived that way, we can tell you it is almost always one of two things: the math, or the logistics.
The math complaint sounds like "the numbers went haywire" or "it took credit for everything." The logistics complaint sounds like "we spent six weeks getting set up" or "they couldn't read our station's log." Both are avoidable, and both are visible before you sign if you ask the right questions. Here are the nine we would ask any vendor — including us.
1. How do you set the baseline?
This is the question that exposes the most common failure in TV attribution, which we call the sponge method: the vendor looks at the fifteen minutes after your spot aired and credits the spot with every visit, call or order in that window.
That is not attribution. It is a timer. Your site was getting traffic before the spot aired and would have kept getting it if the spot never ran. A vendor who cannot tell you what should have happened in those fifteen minutes cannot tell you what the spot added.
A real baseline is built from months of your own data — the same day of the week, the same hour, the same minute — and it accounts for spots that ran in the past. The attributed lift is the spike minus that expectation. Ask the vendor to explain their baseline in one paragraph. If the paragraph does not contain the words "expected" or "would have happened anyway," keep looking.
2. How do you handle spots that overlap?
Two airings ten minutes apart on different stations produce one blended spike. A vendor that credits both spots with the full spike is counting the same visitors twice; a vendor that credits only the first is throwing away the second station's contribution.
Ask how credit is shared when airings land close together, and whether the response window is fixed or dynamic. A fixed window (always fifteen minutes, always thirty) is easy to build and wrong half the time. A dynamic window ends when your traffic returns to baseline — which is a different length for a national spot at 8 pm and a local radio read at 6 am.
3. What do you need to install on our website?
There are two answers, and they lead to very different first months.
The pixel answer. Most vendors ask you to place their tag on your site so they can rebuild your web analytics inside their own system. Your developer installs it, you wait for data to accumulate, and from that day forward the vendor's numbers and your Google Analytics numbers will never quite agree — because they are two different measurement systems counting the same visitors with different rules. There is also no history: measurement starts when the pixel goes live, so last quarter's campaign is invisible.
The read-only answer. The alternative is to give the vendor read-only access to the Google Analytics 4 property you already have. Nothing to install, nothing for your developer to review, and one source of truth — the same numbers your team already reports. It also means the vendor can measure backwards: GA4 has been collecting data since the day it was set up, so a campaign that ran three months ago can be attributed today.
We take the read-only route, and there is a second reason to prefer it that vendors rarely say out loud. Rebuilding Google Analytics is not a small project. Google owns the search engine, the ad platform, the video platform and the email service that feed your traffic; a third party recreating session and source data from a pixel is starting from behind on every one of those, and any error in that reconstruction flows straight into the attribution math. Ask any vendor which system their numbers are built on. If the answer is "ours," ask how it reconciles with yours.
4. What happens when our postlog arrives as a PDF?
A postlog is the receipt for your airings: what ran, when, on which station, which creative. It is also the least standardised document in advertising. They arrive as Excel files, CSVs, PDFs and station-portal exports. Times are sometimes in the market's local zone and sometimes not. Some are offset by a broadcast-day convention that starts at 5 am. Most arrive on Tuesday, Wednesday or Thursday for the previous week, which leaves very little time before next week's buy has to be placed.
Large vendors are built for autopilot: your log fits the template or it does not load. Ask what happens when it does not. Ask who reconciles the time zones. Ask whether a human looks at your first log or a parser does. We onboard the same day we get access, and we spend the first few days going through every one of those quirks by hand, once, so that every week afterwards is automatic.
5. How long until the first report — and do we have to sign first?
The typical vendor path is contract, payment, pixel installation, data accumulation, then a first report weeks or months later. By the time you find out whether the numbers make sense, you are committed.
Ask whether you can see a report on your own data before you pay. Few vendors offer it. We do: you give us read-only GA4 access and a postlog, we return the first report the next business day, and we will run up to a month of your airings before you decide anything. There is no contract to sign first — at most an NDA if you want one.
A month is the right length for a trial. It is enough airings to see whether the numbers hold up, and short enough that you are evaluating the vendor rather than getting free analysis.
6. How do you measure local ads?
A spot on a Phoenix station should be measured against Phoenix traffic, not national traffic. That sounds obvious, and most vendors do not do it — because it means pulling by-minute data for every market separately instead of one national series, and that is a great deal more data to move.
The cost of skipping it is that local markets contaminate each other. A Denver spike gets credited to a Phoenix airing that happened to be nearby in time, and your station rankings become noise. Ask whether local airings are evaluated against the traffic from that market only. If the vendor measures radio, ask twice: local radio is where this matters most.
7. What can you measure — and what can't you?
The honest answer to "what can you measure" is the same for every serious linear vendor: anything with a timestamp and a geography. Broadcast, cable, satellite, local and national radio, satellite radio, live reads, long-form and infomercial, live streaming TV such as YouTube TV, out-of-home with a schedule — all of it produces a moment in time and a place, and all of it can be attributed the same way.
The honest answer to "what can't you measure" is anything without a timestamp: on-demand streaming, podcasts, anything where the audience chooses when to watch. A vendor who claims to attribute those is doing something other than the method described in this article. Ask which it is.
8. Will your numbers match our vanity URLs and 800 numbers?
No — and be cautious with any vendor who promises they will.
Vanity URLs, dedicated phone numbers, QR codes, offer codes and "how did you hear about us" dropdowns are attribution systems too — ones that depend on the customer cooperating. Most do not. They type the brand into Google instead of the URL they saw. They pick the first option in the dropdown. They call the number on the website, not the one on screen. Self-reported attribution under-counts, sometimes badly, and it under-counts inconsistently.
Statistical attribution does not depend on the customer doing anything. The two will never agree, and the gap is not a bug in either one. Ask the vendor how they explain that gap to a client on the first call. If they cannot, the first call will not go well.
9. How would we check your math?
You should never have to take a vendor's numbers on faith, including ours. Here is the check we tell every client to run.
Take the weeks before TV started and the weeks after. The top-line lift in sessions or orders over that period should roughly equal the sum of the incremental lift the vendor attributed to individual spots. If the vendor's spots add up to far more than the business actually grew, the baseline is wrong — usually the sponge method from question one. If they add up to far less, something is being missed.
And distrust any report that credits television with nearly all of your sales when you were selling things before television. That is not a strong result. It is a broken one.
Who should not buy any of this
It is easier to trust a vendor who tells you when not to hire them, so: TV attribution is a poor fit for very small setups. If you are running a couple of thousand dollars a month across ten-dollar airings, or your website sees fewer than a hundred visits a day, there is not enough signal for any method to find. One visitor an hour becoming two visitors an hour is within normal variation, not a measurable spike. The method works when there is enough baseline traffic to see a pattern and the spots are large enough — roughly $500 and up — that the response is unmistakable.
If that describes you, attribution turns a channel with no feedback loop into one with a weekly one. If it does not, spend the money on media until it does.
What switching actually involves
Less than you expect. There is no data to migrate, because the vendor's data was never yours to begin with; your own analytics and your own postlogs are the inputs, and you keep both. What you gain is a second set of eyes on the same airings, which is worth having even if the two vendors agree — and instructive when they do not.
A note on building it yourself
The most common objection we hear in 2026 is not about another vendor. It is "couldn't we build this ourselves with AI?" You could. We started in a spreadsheet. But the code was never the hard part; the hard part is the twelve years and the hundreds of millions of dollars of media it took to learn where the baseline breaks, how overlapping spots share credit, what a broadcast-day offset does to a postlog, and how each station's logs differ. That knowledge is what you are buying, and it is what you should be asking every vendor about.
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, from a few thousand dollars a month to several million, with our longest client relationship now past seven years. We work from read-only GA4 access and your postlogs, no pixel, with a first report the next business day. Plans start at $399 a month, month-to-month. If you would like to see it on your own data before deciding, book a demo.