A CMO I talked to last month spent $180,000 in Q1 on Meta ads. She ran the whole quarter on Advantage+ audiences and lookalikes because Meta’s own rep told her that was the way. Cost per lead climbed the entire quarter. By March she was paying more than double what she paid in January for the same intent quality. Her CFO wanted an explanation.
There isn’t a clever one. What happened to her is happening to every performance team that still leans on interest-based targeting: the signal has been priced in. Meta and Google spent a decade convincing marketers that behavior data plus machine learning would find buyers better than the marketer could. For a while, it did. Then everyone bought the same ad slot at the same time chasing the same “small business owners aged 25 to 54 who like productivity software.” Auctions inflate. Response rates flatten. CPMs move up and to the right.
The fix isn’t a smarter interest cluster. It’s a different kind of signal entirely.
Interest data is a guess. Search data is a decision.
When Meta tells you a user is “interested in” commercial real estate, that user probably watched a video about a warehouse two months ago and never came back. The interest label is inference. It’s a statistical guess dressed up as fact.
When someone types “warehouse for lease Dallas” into Google at 4:17 PM on a Tuesday, that is not a guess. That is a person actively looking for warehouse space right now. If you can put an ad in front of them within 24 hours, you’re not competing against every other advertiser guessing at the same interest cluster. You’re the first message they see after they searched.
This is the core of what Springbolt calls Custom Audiences. Real search-intent data, harvested in the last 24 hours, packaged into device IDs, and deployed straight into TikTok, Facebook, and Google as a custom audience upload. No lookalikes. No expansions. No modeled intent. The people in the list actually searched the exact keyword you specified. Yesterday.
What “24 hours” actually changes
Marketers rarely think in hours. They think in campaigns, weeks, funnels. But intent has a half-life measured in hours, not weeks.
Consider the data from research on B2B buying behavior: a prospect who searches for a solution today has a 3x higher probability of engaging with a related ad within 72 hours compared to the same prospect at 7 days out. At 14 days, the probability curve flattens toward baseline. The searcher has either bought, forgotten, or moved on.
Broad-interest targeting doesn’t know when someone entered the market. It only knows they fit a profile. So you’re paying to reach a mix of people who searched yesterday, people who might search next month, and people who will never search because Facebook’s interest algorithm was wrong.
A 24-hour intent list eliminates the noise. Every device in the file was actively researching yesterday. If you launch a campaign within the next day, you catch them at the exact moment their consideration set is forming.
The math that changes when you narrow the window
Broad interest campaigns on Meta typically run $18 to $35 CPM in B2B verticals. Click-through rates hover in the 0.9% to 1.4% range. Conversion from click to demo request runs 2% to 4% on a good day.
Intent-based custom audiences deployed to the same platform typically post CTRs between 2.1% and 3.8% and demo conversion in the 6% to 11% band. Not because the creative is better. Because the audience actually wants what the ad is selling.
Broad interest vs. 24-hour intent, typical B2B Meta campaign
|
Metric |
Broad Interest |
24-Hour Intent |
|
CPM |
$18 to $35 |
$20 to $32 |
|
Click-through rate |
0.9% to 1.4% |
2.1% to 3.8% |
|
Click-to-demo conversion |
2% to 4% |
6% to 11% |
|
Cost per opportunity |
$280 to $410 |
$110 to $180 |
The response rate lift Springbolt clients see on Custom Audiences vs standard broad targeting typically runs up to 50%. That number sounds like marketing spin until you sit with the underlying arithmetic: a list of 12,000 device IDs who searched “erp for manufacturers” in the last 24 hours will always outperform a modeled audience of 4 million lookalikes. The only question is which one your budget can afford to test.
Why the platforms don’t sell this natively
Meta, TikTok, and Google will never build the 24-hour intent audience natively. Their business model requires you to buy through their algorithm because their margin depends on it. If you could target only the people who searched yesterday, you’d spend a fraction of what you spend now. Their revenue would drop by an order of magnitude.
So the intent data lives outside the platforms, harvested from the open web, packaged by specialized data companies, and uploaded to the platforms as first-party audiences. That’s the workaround. It exists because performance marketers got tired of paying inflated CPMs for opaque targeting.
Where broad targeting still makes sense
To be fair to the machines: broad interest targeting still works for two things. Brand awareness at scale, where the point is impressions and not conversions. And retargeting, where the audience already came to your site and Facebook is just showing them your ad again.
For everything else, particularly demand generation, direct response, and account-based marketing, broad targeting is a tax you’re paying for convenience. The intent data is available. The upload takes minutes. The only reason not to switch is that switching requires unlearning a decade of “trust the algorithm” advice.
The uncomfortable question
Ask any performance team this: if the algorithm is so good at finding your buyers, why have your CPMs doubled in three years while your close rate hasn’t moved?
The honest answer is that everyone bought the same box. When the entire market pays the same platform for the same modeled audience, the audience stops being a differentiator. What you get is competitive parity, priced at auction. Search intent, harvested outside the platform and injected as a first-party list, is a way out of that parity.
Get started without blowing up existing campaigns
The lowest-risk move: pick one campaign that’s currently running on broad Meta interest targeting. Duplicate it. Swap the audience for a Custom Audiences upload with 10,000 to 20,000 device IDs from the last 24 hours matching your top three buyer intent keywords. Run both for two weeks with the same creative and budget. Compare CTR, CPL, and pipeline-qualified leads.
If the intent list doesn’t outperform, keep running broad. But budget for the test to work. Every Springbolt client that has run this A/B has moved budget after week two.
The three keyword archetypes that separate winners from losers
The audience is only as good as the keywords that built it. In our experience running intent campaigns across manufacturing, financial services, healthcare tech, and consumer subscription categories, the keywords that produce the strongest lift fall into three archetypes.
Comparison and evaluation phrases. “X vs Y,” “best X for Y,” “X alternatives,” “top X in 2026.” These are search queries from buyers actively narrowing a shortlist. Response rates on comparison-phrase intent audiences typically run 30 to 50% higher than baseline because the buyer is deep in evaluation, not exploration.
Problem-language phrases. “How to fix X,” “why is X happening,” “X keeps breaking.” These reach buyers earlier in the funnel, but the ones who convert convert well because the ad meets them at a specific pain point. Better for lead nurture campaigns than direct-response, and often the source of the best content-to-conversion journeys.
Vendor-substitute phrases. “Alternatives to X,” “replace X,” “cheaper than X.” These are gold if you’re a challenger in a category with an incumbent everyone knows. The audience is already unhappy with the current vendor. Your ad just has to show up.
The archetype to avoid at all costs: informational phrases like “what is X” or “how does X work.” Those are researchers, students, and casual readers. Almost none are in market. A campaign built on those keywords burns budget on people who will never buy.
When broad targeting still deserves budget
To be clear about where intent data doesn’t help: brand awareness campaigns at scale. If the goal is impressions and top-of-mind recall across a wide audience, broad interest and demographic targeting on Meta or YouTube still make sense because the point is coverage, not conversion. Intent data doesn’t have the volume to serve a $500K per month brand campaign efficiently.
The line to draw: any campaign measured by pipeline, opportunities, or revenue benefits from intent data. Any campaign measured by reach, frequency, or brand lift is fine on broad targeting. Most performance teams already know which of their campaigns fall on each side of that line. What they haven’t done is shift the pipeline-measured budget over to intent.
The competitive advantage window is closing
Right now, intent-based custom audiences are still a competitive edge because most performance teams haven’t adopted them. The teams running them are pulling ahead on CAC while their competitors keep paying inflated CPMs for modeled audiences.
That window closes within 18 to 24 months. As more advertisers move into intent-based buying, the auction dynamics on intent audiences will start to look more like the auction dynamics on broad targeting today. Pricing will normalize, edge will compress.
The teams that move now capture the arbitrage. The teams that wait until intent-based targeting is “the standard” will be adopting it at the price point where the edge has already been priced away. Not a reason to panic. A reason to run the test this quarter instead of next year.
Frequently asked questions
What is search intent data?
Search intent data is a record of what a specific device recently searched for on the open web. Unlike interest targeting, which infers what someone might care about based on profile and past behavior, search intent data captures an explicit query at a specific moment in time. Custom Audiences from Springbolt use 24-hour search intent data, meaning every device in the file searched a specific keyword within the last day.
How is search intent data different from Meta or Google audience targeting?
Meta and Google interest audiences are modeled. They estimate what someone might be interested in based on profile signals, page likes, and inferred behavior. Search intent data starts with an explicit search query on the open web. It is a direct signal of active consideration rather than a statistical guess.
Does search intent data work on Facebook and Instagram if it isn’t native to Meta?
Yes. Intent data is packaged as device IDs that Meta accepts as a Custom Audience upload. Once uploaded, the audience behaves like any other custom audience inside Meta Ads Manager, but the underlying targeting is based on real search behavior rather than Meta’s own interest categories.
How fresh does intent data need to be?
Intent is time-sensitive. Data older than 72 hours starts losing predictive value. Data older than two weeks is typically no better than baseline audience data. Custom Audiences from Springbolt use 24-hour intent for that reason.
Is search intent targeting compliant with privacy regulations?
Yes when sourced correctly. Reputable intent data providers work with consented, hashed identifiers and comply with GDPR, CCPA, and platform-specific policies. Springbolt only uses providers with documented compliance frameworks.
What kinds of businesses see the biggest lift from search intent targeting?
B2B companies with a defined ICP and a considered purchase cycle typically see the highest lift because their prospects actively research before buying. Software, financial services, healthcare, real estate, and professional services are strong fits. Impulse-purchase consumer categories see smaller lift because the buying window is shorter than the intent data can catch.