Why multi-touch attribution undercounts your audio ads

A radio DJ plays a song you've never heard on your morning commute. It sticks with you, so that night you search for the band in your streaming app and add three of their songs to a playlist. Ask the app how you found that band, and it'll tell you search. It has no idea the DJ exists.

Audio advertising has the same blind spot. A listener hears a host read your ad in the middle of their favorite podcast, and a few days later they Google your brand, buy from your site, or grab your product off a shelf at Target. Measurement built around clicks sees the search, the site visit, or nothing at all, and the ad that started the whole thing gets little or no credit.

That gap gets more expensive every quarter you spend real money on audio. When your measurement can't see where audio's impact lands, the channel looks pricey and underwhelming on paper, so it's often the first thing cut when budgets get tight. Cut it, and the branded search volume and retail sales it was feeding tend to soften a few months later, long after anyone connects the two. Getting audio measurement right is how you avoid trimming the channel that's been filling the rest of your funnel.

Key takeaways

• Multi-touch attribution (MTA) gives credit to clicks and tracked touchpoints, and most audio ads don't produce either one.
• Audio workarounds like promo codes, vanity URLs, household IP matching, and post-purchase surveys capture a slice of audio's impact, but they consistently undercount it.
• Audio's effects can take weeks to build and months to fade, which often outlasts the attribution windows MTA relies on.
• Much of audio's value shows up as halo effects, like lifts in branded search, direct traffic, Amazon sales, and retail sales.
• Marketing mix modeling starts with the sales you can actually observe and estimates what drove them, so it doesn't need a trackable path from someone's ears to your checkout page.
• Promo code redemptions and survey answers are still useful as a floor and a gut check alongside a marketing mix model.
• Measuring audio accurately lets you scale the shows and placements that work instead of cutting the channel when platform numbers look thin.

How multi-touch attribution works, and why audio breaks it

To see why audio is such a tough fit, it helps to look at what MTA needs to function. Multi-touch attribution splits credit for a sale across the touchpoints a customer had along the way, like a social ad click, an email open, and a paid search click. To do that, it needs a user-level trail: a click, a cookie, a pixel firing, or a device ID that ties each touchpoint to the same person.

Audio rarely leaves that trail. People listen while they're driving, cooking, running, or folding laundry, so there's usually nothing to click and no screen in front of them. When they do act, it's often hours or days later, on a different device, through a different channel. MTA can only give credit to touchpoints it can see, and audio mostly happens where it can't see.

The workarounds audio measurement relies on

Since there's no click to follow, audio measurement has built a set of workarounds to connect listening to buying. Each one captures something real, but each one also misses a lot.

These are all helpful signals, and plenty of brands lean on them heavily. The catch is that every one of them is still trying to rebuild the path from listener to buyer, which keeps audio measurement stuck inside the same logic as MTA. If the path isn't trackable, the sale doesn't count.

Where multi-touch attribution falls short for audio

The workarounds above narrow the gap a little, but they can't fix the underlying problems with using MTA for audio. Four of them do most of the damage.

Audio's impact outlasts the attribution window

Audio tends to work like a slow burn. As we covered in our Laws of Marketing series, a podcast sponsorship might not move sales for weeks, but once it starts working, its effects can linger for months as listeners hear the same host recommend you episode after episode.

MTA tools typically look back over a fixed window, often a few days to a few weeks. Anything audio drives after that window closes either gets dropped or gets credited to whatever the customer touched more recently. That means the longer your audio keeps working, the less credit MTA tends to give it.

The last trackable touch takes audio's credit

Here's a common path: someone hears your ad on a podcast, searches your brand name a few days later, clicks a branded search ad, and buys. MTA sees the branded search click and maybe a retargeting ad that followed. It doesn't see the podcast at all.

The result is that channels that capture demand, like branded search, retargeting, and direct traffic, look more efficient than they really are, while the audio that created the demand looks like it's underperforming. If you shift budget based on that picture, you end up moving money away from the source of demand and toward the channels that were catching it.

The signal keeps shrinking

MTA's accuracy has been heading in one direction for years, and it isn't up. Privacy changes have chipped away at cookies, device IDs, and pixel tracking across the board, and audio's workarounds lean on some of the most fragile methods available. Household IP matching, for example, struggles with shared networks like office Wi-Fi and with privacy features that hide a user's IP address.

As those signals get weaker, the share of audio's impact that MTA can see gets smaller too. Even if your audio is working just as well as it did last year, your reports may say otherwise.

Retail and Amazon sales don't show up at all

This is the biggest blind spot for omnichannel brands. When a listener hears your ad and then buys at Walmart, Sephora, or on Amazon, they never touch your site pixel. There's no click, no code, and no IP match to record.

For brands with a real retail footprint, that can be a big chunk of what audio actually drives. MTA treats those sales as if they had nothing to do with your marketing, which can make audio look like a weak channel when it's actually moving product off shelves.

What audio is actually doing for your brand

If MTA isn't seeing audio's full impact, it's worth looking at where that impact goes. A lot of it shows up as halo effects, which is revenue a campaign drives indirectly through another channel. Halo effects can come from any campaign, but audio is a strong example because there's almost never a click to give it direct credit.

Here's where audio's halo effects commonly land:

• Branded search: Listeners remember your name and look you up later.
• Direct traffic: People type your URL straight into their browser, especially after hearing a host say it a few times.
• Organic traffic: Some listeners search for your product category and find you through organic results.
• Amazon: Plenty of shoppers go straight to Amazon to buy something they heard about, especially if they already have Prime.
• Retail stores: Listeners pick up your product the next time they're at Target, Ulta, or their grocery store.

None of those purchases look like audio sales in a click-based report. Measuring audio well means connecting your spend to revenue across all of those places, not just on your own site.

Multi-touch attribution vs. marketing mix modeling for audio

A marketing mix model takes a different approach to the same question. Instead of following individual customers, a marketing mix model looks at your spend and impressions across every channel alongside your actual revenue over time, then estimates how much each channel and campaign contributed. Here's how the two compare for audio specifically.

The key difference is where each method starts. MTA starts with the customer journey and tries to trace it forward, which only works when every step is trackable. A marketing mix model starts with what you can actually observe, like the sale on your storefront, your Amazon revenue, or your retail sell-through, and works backward to estimate what drove it. That's a much better fit for a channel that does most of its work where tracking can't follow.

How to get more out of the audio data you already have

Moving to a marketing mix model doesn't mean throwing out the audio data you're already collecting. Used the right way, it becomes a helpful check on what your model is telling you.

• Treat promo code revenue as a floor. Code redemptions tell you the minimum audio drove, never the full amount. If your marketing mix model shows audio driving more than your codes do, that's expected.
• Use surveys directionally. Post-purchase surveys are great for spotting which shows or hosts people remember, but self-reported answers shouldn't be your source of truth for budget decisions.
• Keep your show-level and delivery data flowing. Data from audio platforms and audio measurement tools, like impressions and spend by show, gives your marketing mix model better inputs to work with.
• Gut check the numbers against each other. If your model and your code data point in wildly different directions for a specific show, that's worth digging into before you renew or cut.

Incrementality tests come up a lot in audio, too. They can give you a useful read on a specific show during a specific window, but our view is that they're locally accurate and globally inaccurate. A test tells you about the conditions it ran in, not how that show will perform next quarter or at a different spend level. When a client runs a test, we show model accuracy with and without the test incorporated, and the client decides whether to include it.

Scaling audio without guessing

Once you can see what audio is actually driving, the next question is usually whether to do more of it. Should you renew a sponsorship, add new shows, or shift budget into audio from another channel? Those are the decisions where measurement pays for itself.

A few things to keep in mind as you plan:

• Give audio time before judging it. Since audio effects can take weeks to build, cutting a new sponsorship after two weeks of thin results often means cutting it right before it starts working.
• Don't assume audio is tapped out. Many marketing mix models bake in the assumption that every channel hits diminishing returns early. Prescient's own research on saturation assumptions found that those default assumptions often don't hold up for digital campaigns, which can lead brands to underspend on channels that still have room to grow.
• Forecast before you shift budget. Instead of guessing, model what happens if you scale audio spend or move budget into it from another channel while keeping your total budget steady.

Every campaign saturates differently, so the right move for one podcast sponsorship might be the wrong one for a streaming audio campaign. Looking at each campaign on its own keeps you from applying one rule to the whole channel.

Where Prescient comes in

Prescient's marketing mix model measures audio alongside every other channel you run, down to the campaign level, with daily updates. We integrate with audio platforms like Spotify Ads, iHeartMedia, and Acast, as well as audio measurement providers, so your podcast, streaming audio, and broadcast radio spend sits in the same model as your social, search, CTV, and retail media. Because our model starts with the sales you can see, including Amazon and retail revenue, it can show the halo effects audio drives into branded search, your storefront, Amazon, and the retail shelf.

When you're ready to scale, our Media Forecaster feature helps you see what's likely to happen when you increase audio spend or shift budget into it from other channels, using what our model has already learned about your seasonality. Want to see what your audio is really driving? Book a demo and we'll walk through how Prescient uncovers its full impact.

FAQs

Can multi-touch attribution track podcast ads?

Only partially. MTA needs a trackable touchpoint, and podcast ads usually don't produce a click. Podcast measurement fills the gap with promo codes, vanity URLs, and household IP matching, which can connect some listeners to site visits and purchases. Those methods miss listeners who search your brand later, buy on Amazon, or buy in a store, so MTA tends to undercount what podcast ads actually drive.

Why do promo codes undercount podcast ad performance?

Promo codes only capture listeners who remember the code and use it at checkout on your site. Plenty of people forget the code, find a different discount, or buy somewhere else entirely, like Amazon or a retail store. Codes also tend to leak onto coupon sites, which means some redemptions don't come from listeners at all. That's why code revenue works best as a minimum estimate rather than a full picture.

How long does it take for audio ads to show results?

It depends on the format and the show, but audio often works on a longer timeline than click-based channels. Host-read podcast ads in particular can take several weeks to build momentum as listeners hear your brand repeatedly, and their effects can keep showing up for months after the campaign ends. That's why it's worth giving a new audio campaign enough time before deciding whether it's working.

What's the best way to measure audio advertising?

The most complete approach is a marketing mix model that includes your audio spend alongside every other channel and all of your revenue sources, including Amazon and retail. Because it doesn't depend on tracking individual listeners, it can capture the delayed and indirect sales audio drives. Promo codes, surveys, and data from audio measurement tools are still useful as inputs and gut checks alongside it.

Should I stop using my audio attribution tool if I have a marketing mix model?

Not necessarily. Audio measurement tools are good at things like verifying that ads ran and giving you show-level delivery data, and that data can feed into your marketing mix model. The shift is in how you use their conversion numbers: treat them as a directional signal and a floor rather than the final word on how much revenue audio drove.

Does marketing mix modeling work for broadcast radio?

Yes, and broadcast radio is one of the channels where it's most useful. Traditional radio has even fewer tracking options than podcasts or streaming audio, since there's no download or stream to match to a household. A marketing mix model only needs your spend and delivery data alongside your revenue, so it can measure broadcast radio the same way it measures every other channel in your mix.