podcast ad testing: what's actually moving the needle?

chloeest

I've been running a series of controlled experiments on podcast ads for our B2B product, and the results have been... messy. Spent a good chunk of change across a few networks and directly with shows. Here's what I've found so far.

Networks vs direct: Tested SpotsNow and Acast programmatically vs reaching out to show hosts directly. Direct buys gave us better host-read quality, but the negotiation time is insane. Programmatic was faster and cheaper per spot, but the placements felt generic. Right now I'm leaning toward a hybrid: direct for 5-7 high-fit shows, programmatic for the rest.

Host-read vs produced: Hypothesis was that host-read would outperform produced spots for trust. Tested both on the same show (A/B within the same episode). Host-read generated 40% more unique promo code usage, but the CPM was 3x higher. The key? Host actually needed to use the product. We sent a trial account and brief, but one host just read our script verbatim and that spot bombed. Lesson: over-scripted reads are garbage.

Attribution is the nightmare everyone warns about. We pulled branded search lift, direct traffic to our demo form, and self-reported attribution on lead forms. Podcast didn't give us the clean signal we get from search ads, but the branded search volume spiked 25% during the flight. Sales calls also started including "heard you on [show]" without prompting. Best we've got.

Adjacent genres punch above their weight: Business and marketing podcasts are oversaturated - everyone's buying the same inventory. We had better cost-per-converted-visitor on finance/economy shows and a few comedy podcasts with professional audiences. The host-read spots on those felt more authentic too.

If I were doing this again, I'd run a pilot across ₁₀ shows with enough frequency to build familiarity instead of spraying budget across 50. The real question is whether you're optimising for brand recall or direct response - that changes the bidding strategy and attribution setup entirely.

Happy to share screenshots of our heatmap data on site behaviour post-campaign if anyone's interested.

ryanseo

Focusing on the attribution side of things - that's where most people get burned.

Whatever ad network data you're looking at, bin it. Seriously. The platforms engineer those numbers to keep you spending, not to give you a clear picture. If you want real attribution, you've got to split it into two buckets:

Direct attribution. Easy stuff. Coupon codes, QR scans, UTM-tagged links. But here's the thing - direct attribution usually makes up a fraction of the actual impact. Maybe 20-30 % of the real picture, sometimes less.
Indirect attribution. This is where you need incrementality testing or lift analysis. The cleanest way I've seen is geo-based splits. Serve the ads in specific regions, keep a control group in others, then compare the difference in behaviour between the two. No fancy tech required - just clean setup and patience.

Most people skip the indirect part because it's harder to measure. But that's where the real value lives. Happy to dig into the nitty-gritty if you're running a test.

nightowl

All that cold outreach from podcast networks is actually useful signal if you know how to interpret it. Gumball gives you a realistic view of CPMs in niche B2B spaces - treat it like a market temperature check. but the real leverage comes from host-read placements on three or four shows that map tightly to your ICP. That's spear fishing versus dragging a net. for pipeline building, it'll outperform any programmatic network buy, every time

tiktokguru

We tested podcast ads pretty heavily for B2B SaaS, and honestly the biggest surprise was that the "obvious" business and marketing shows were often the worst value.

Everyone's buying the same inventory now, so CPMs go through the roof and audiences are absolutely saturated with SaaS ads. The best-performing shows for us came from adjacent categories:

📊 Finance / economy
👨‍💼 Operator / founder podcasts
⏳ Productivity
😂 Even some comedy and culture pods with white-collar audiences

Host-read outperformed produced spots by a mile - but only when the host genuinely understood the product. Over-scripted reads felt dead from the first listen.

On attribution: I'd stop expecting paid-search-level clarity. The best signals we found were branded search lift, direct traffic lift, self-reported attribution on demo forms, and hearing "I keep hearing about you guys" on sales calls.

If I were you, I'd run a focused pilot across ₁₀ shows with enough frequency to build familiarity, instead of spraying budget across 50 podcasts once.

Curious though - are you optimising more for brand familiarity / category creation, or trying to make podcast directly measurable on pipeline from day one? That changes the strategy a lot.

sophiasaas

Totally agree - the podcast ad space is still the Wild West. i've been testing programmatic host-read placements to cut down the manual negotiation headache, and the CPMs land somewhere between full host reads and plain dynamic inserts. Feels like a solid middle ground if you've got the tracking infrastructure. Pricing transparency is still a nightmare across the board, everyone's playing with different variables. Just wish more platforms would open up real measurement beyond vanity metrics

davidsearch

nothing groundbreaking to add, but that Conan O'Brien LinkedIn spot is a perfect example of creative breaking through low intent. Podcast ads typically have terrible CTR and CPA if you just read a script, but when the host genuinely sells it, the conversion rates can rival search. you're right-thinking outside the box is the only way to justify the CPMs.

marketingmule

honestly, if I had that kind of budget to throw at podcast ads, I'd go completely the opposite direction to most people. Rather than jumping into a big network buy, I'd run a tight pilot with maybe 5-10 hyper-niche shows that align directly with SMB or marketing leaders. Direct host-read spots every time-audience fit beats download numbers hands down.

what really makes or breaks podcast ROI though is the attribution setup. Promo codes alone are garbage. you need post-purchase surveys, branded search lift tracking, and dedicated landing pages before you even think about scaling. The return is usually messier than paid search, but when it works it cuts deeper.

Anyone else found that podcast attribution just doesn't play nice with their regular measurement stack?

growthgeek

yeah, that's the standard advice everyone parrots - "focus on audience fit over category." Sounds sensible, but in practice it's often a trap. a small, super-engaged audience that's tightly aligned with your niche sounds great until you realise they're also too small to move the needle on ROAS. You need volume to test creative, iterate on hooks, and get statistically significant signal. if you're spending five figures, a handful of "perfect match" shows with 10K downloads each isn't going to cut it unless your product has a ridiculously high LTV.

i've run tests where a broad, semi-relevant show with 200K downloads out-performed the perfect fit show with 30K - simply because the sheer scale gave us enough data to optimise the ad and offer mid-flight. Audience fit matters, sure, but it's not the only variable. If you're not tracking UTM parameters and conversion pixels religiously, you're guessing anyway.

So my take: run a small test across shows that are adjacent to your category, not just a perfect match. Let the data tell you which audience actually converts. Don't get seduced by niche engagement metrics. Been burned by that too many times.

wordweaver

Oh, host-read sponsorships are the way to go if you want actual eyeballs (well, ears). Programmatic through Spotify or Megaphone? Sure, if your wallet's thicc enough. My hot tip: hunt down the tiny shows where the host genuinely loves whatever category you're flogging. Nothing beats that authentic nod.