“The podcast ads are running. Are they working?”
So you open your attribution report. And the channel looks like it did almost nothing. A handful of orders, maybe. A lot less than you paid for.
That’s not a verdict on the channel. It’s how the report works.
Why can’t you measure CTV, TV and podcast ads by clicks?
These channels create demand that shows up later. It lands in search, in another channel, or as someone typing your URL. TV and radio never show a click at all. And CTV and podcast ads are rarely the last thing anyone clicks before they buy.
So last click undervalues them by design.
Last click isn’t nothing. It’s just the wrong place to look for proof on a channel that doesn’t produce clicks. Here’s why click-path attribution can’t see this media.
This matters most right before a budget cut. An upper-funnel video can look weak on last-click ROAS and still be adding sales. Before anyone pauses it, a credible test result should be part of the conversation.
What can you measure without clicks?
More than you’d think. Here’s the short map:
- Total sales and new customers over time, lined up against when the channel ran.
- Differences between regions that got the media and regions that didn’t.
- Surveys (“How did you hear about us?”), as a directional read only. People forget, and they misremember the path.
- Platform-reported views or impressions, as proof of delivery. Not proof of impact.
Only the first two can show what the channel added. And they’re strongest when you create the change on purpose.
How do you test CTV with a geo test?
Where the buy can be targeted by region, a geo test is usually the cleanest read. If you’re new to it, start with what geo testing is.
There are two basic designs. In a holdout, you turn the channel off in some regions and keep it on in the rest. In a scale-up, you add spend in some regions. Either way, you compare sales in your own order record, not the platform’s report. And agree on what you’ll do with the result before it starts: “If this works, we will do it.”
One note on labels. In a holdout, the regions that go dark are the holdout. The regions that stay on are what I call the treatment. Providers label this differently, so check which is which.
Google’s researchers describe geo experiments the same basic way: regions assigned to test or control, with ads targeted by location.
The result is an estimate with a range, not an exact number. So size the test before you buy it. Run three checks on your own sales history, and work out the minimum detectable effect (MDE) as a range.
Can you see CTV’s effect on other channels? Sometimes. If you can measure search or direct sales by region, compare them in the test and control regions too. Total sales stay the bottom line.
For a worked version with paid search, see a geo holdout.
How do you measure national TV when you can’t buy it by region?
National TV and national radio can’t be contained in a geographic cell. So a clean geo holdout is off the table.
My answer: both of these, where possible.
- An MMM that uses flighting and spend changes over time. An MMM (marketing mix model) estimates each channel’s contribution from how sales moved as spend moved. Here’s how I approach marketing mix modeling.
- Geo tests that use regional spot buys as a stand-in for national.
But the brand’s own buying strategy may limit either one. If you don’t have a big, stark regional add-on, or you don’t want major spend changes, both get harder.
Regional buys come with limits, too. Regional CPMs can run higher, so they may not reflect the cost of a national buy. Regional buying may also leave out national and broadcast inventory that a national campaign would use. So translate a regional result to national economics openly, and treat it as directional.
Check the regional group with baseline analysis first: relative scale, correlation and MAPE (mean absolute percentage error). These limits are risks to weigh. They aren’t reasons to demand perfect certainty before you decide.
And look past revenue. Check orders and new customers too. If only some of them move the way you expected, the risk goes up that the result won’t hold nationally.
If you have both, the test result is what you check the MMM against. Calibrating the model to that result is optional. Here’s what a test can calibrate in an MMM.
If you can’t buy the channel by region and aren’t sure what your buying strategy allows, talk through how to measure it.
Can you use promo codes and vanity URLs for podcasts and creators?
Yes, as a clue. With a strong caveat.
Two things can be true about promo codes. They can undercount, because listeners don’t always use them as instructed. And people who never heard the ad can find them.
Coupon browser extensions look for codes at checkout. Honey says it will “test codes at checkout” to help shoppers find a deal. Capital One Shopping says it “automatically applies coupon codes at checkout”.
Example (hypothetical): Say a podcast code shows up on a lot of orders. Some of those buyers heard the ad. Some found the code at checkout. And some listeners bought without using it at all. The code count alone can’t tell you which is which.
So a code can be used directionally. It’s just as possible that it leads you astray. Vanity URLs are a clue too, not proof.
How long should the test run?
Longer than you’d like.
The consideration cycle sets the floor. For upper-funnel media, a brand may need two to three full customer cycles, plus a cooldown. So a three-to-four-week path to purchase can mean three to four months of testing, or longer. A four-week test may miss the effect entirely.
My rule of thumb is a 4-week minimum, and no-click media usually needs more. Then check MDE and power over that length. Shorter tests give only a directional read, and a higher chance of a null result. Use the geo test duration planner, and see why lift tests end too early.
How do you choose a measurement partner or platform?
A quick disclosure first: Brinker Analytics sells measurement services. So read this with that in mind.
You’ll usually see three approaches. A platform lift study is the platform’s own read. A geo test reads your own order record. An MMM reads sales and spend over time. Each answers a different question.
Whoever you talk to, ask:
- What outcome do you read: my orders, or the platform’s conversions?
- How do you build the comparison?
- Will you show the result as a range?
- What do you do when a result is inconclusive?
- Who owns the data and the model?
- Can I rerun the analysis myself?
- What’s the full cost after a good year?
No brand names here, on purpose. The answers matter more than the logo.
FAQs
Does CTV have last-click attribution?
Not in a way you should lean on. CTV ads are rarely the last thing someone clicks. People see the ad and buy later, through search, direct or another channel. So last click undervalues CTV by design. Measure it by what total sales did when the media changed, with a geo test where you can buy by region.
Can you measure podcast ads with promo codes?
Only as a directional clue. Codes can undercount when listeners don’t use them. And coupon browser extensions can surface codes to people who never heard the show. So a code count can point the right way, or it can lead you astray. A regional test or an MMM gives you stronger evidence.
How do you measure TV if you can’t geo-target it?
Use both, where you can: an MMM that reads flighting and spend changes over time, and geo tests that use regional spot buys as a stand-in for national. Your buying strategy may limit either one. Translate regional results to national economics openly, and treat them as directional.
When should you talk to someone?
When how you can buy the channel limits how you can measure it. That’s a conversation, not a formula. Start with the duration planner or the MMM or incrementality quiz. Or talk through a measurement plan with me.
Big picture
No-click media isn’t unmeasurable. It’s just unmeasurable by clicks.
So stop asking it for clicks. Change something on purpose, by region or over time, and watch what your total sales do. No model owns the truth. But a well-run test gets you a lot closer!