Here’s a scenario that plays out every week at agencies with real B2B clients.
Sales calls the marketing lead on a Thursday. They just closed three deals in the last two weeks from companies that all searched the same phrase before hitting the website: “vendor risk assessment software.” The sales team wants marketing to find more of those buyers, fast.
The old answer was: build a lookalike from the closed-won accounts, throw it into a Facebook campaign, hope for the best. Three weeks later you have some clicks, maybe two demos, and no way to know whether the lookalike had anything to do with buyers actually searching for what you sell.
The new answer is faster and more honest. Take that keyword, pull the list of every device that searched it in the last 24 hours, upload the list to Facebook as a custom audience, and run creative to them tomorrow.
That’s the play. This is how it actually works.
Step 1: Pick the keyword the way your buyers think
The biggest mistake performance marketers make with intent data is picking keywords the way an SEO would. SEO thinks about volume, competition, and difficulty. Intent-based ad targeting doesn’t care about any of that. It cares about specificity of buying signal.
For “vendor risk assessment software,” an SEO would look at monthly search volume (roughly 400 in the US), realize it’s low, and probably chase a broader head term like “vendor management” (12,000 monthly searches). That’s the wrong instinct for intent-based ads.
You want the specific phrase because the specific phrase indicates specific intent. Someone searching “vendor risk assessment software” is 6x more likely to be evaluating a purchase than someone searching “vendor management,” even though the second phrase has 30x the volume. Intent quality beats intent quantity when you’re buying device IDs by the thousand.
Springbolt’s Custom Audiences lets you type any keyword and see the exact count of people who searched it in the last 24 hours. If the count is 500, that’s 500 real device IDs of active buyers. Compare that to Meta serving your ad to 4 million modeled interest matches, most of whom aren’t in market. The 500 is the better buy.
Step 2: Layer in the right adjacent phrases
One keyword rarely captures a full buyer set. For vendor risk software, real buyers also search “third party risk management platform,” “vendor risk questionnaire tool,” “TPRM software,” “vendor cyber risk,” and probably five product-name comparisons like “OneTrust vs ProcessUnity.”
Build the audience with three to eight adjacent phrases, not one. Springbolt lets you stack keywords into a single audience so you’re not running eight separate campaigns for what is effectively the same buyer set with different vocabulary.
Where to draw the line: if the adjacent phrase is used more by researchers than buyers, cut it. “What is vendor risk management” is a research query. Someone typing that is not in market. Someone typing “TPRM software comparison” is.
Step 3: Upload to Facebook, TikTok, or Google in that order
The mechanics are the same across platforms. You get a file of hashed device IDs formatted for the destination. Upload as a custom audience. Wait 2 to 24 hours for the platform to match the file against its own users. Launch.
Typical device ID match rates by platform (US audiences)
Platform | Typical Match Rate | Notes |
Facebook / Instagram | 50 to 70% | Highest volume, fastest activation |
TikTok | 40 to 60% | Younger, mobile-first user base |
Google Customer Match | 30 to 45% | Higher signal confidence required |
DV360 / The Trade Desk | N/A (raw IDs) | Use for unmatched IDs across open web |
The unmatched device IDs aren’t wasted. Springbolt returns a separate export of raw IDs you can push into DV360, The Trade Desk, or any programmatic DSP to reach the same intent audience across the open web. Same buyer, different pipe.
Step 4: Match the creative to the search
If you’re targeting people who searched “vendor risk assessment software” yesterday, don’t run a general brand awareness ad. Run creative that speaks to that exact search.
Something like: “You searched vendor risk assessment yesterday. Here’s what most tools miss.” Or a specific-value hook: “The 3-question test to disqualify a TPRM vendor in one meeting.”
The mismatch most teams make is uploading a hyper-targeted intent list and then serving it generic brand ads. The audience quality is wasted because the creative is a broadcast, not a conversation.
Step 5: Measure by pipeline, not clicks
CTR and CPL will improve dramatically with intent targeting. The 50% response rate lift Springbolt sees is real. But the metric that matters is pipeline created and pipeline closed, not clicks.
Set up UTM parameters that tag every intent-audience click distinct from broad-audience click. Feed those tags into your CRM. At 30 days, compare demo booked, opportunity created, and closed-won across the two audience types.
We’ve watched clients cut broad Meta spend by 40% within 60 days of adopting intent audiences because the intent budget was producing pipeline at a fraction of the cost. That’s the point of the exercise.
The traps that kill first campaigns
Every team running intent audiences for the first time hits at least one of these. Get ahead of them and week one goes smoothly.
Trap 1: uploading a list and letting it sit. Match rates take 2 to 24 hours. If you upload a fresh 24-hour intent list on Monday and launch Wednesday, most of the intent value has already decayed. Set up a schedule: refresh the list at 8 AM, launch at 10 AM, kill the campaign by end of week and refresh again.
Trap 2: mixing intent audiences with broad audiences in the same ad set. Meta will optimize the delivery mix and blur your measurement. Run intent as its own ad set with its own budget and its own creative. Compare against your broad campaign at the campaign level, not inside a single ad set.
Trap 3: bidding for reach instead of conversion. Intent audiences are dense. Reach-based bidding wastes them. Set the objective to conversions and let Meta or Google optimize delivery toward the specific event you care about (demo booked, form fill, trial started).
Trap 4: expecting instant scale. A single keyword rarely produces enough intent volume to spend $10K per week. Build audience sets of 5 to 15 adjacent keywords, refresh weekly, and expand keyword sets as you validate what converts. Trying to scale a single keyword to enterprise-level spend usually means you’re spending against people who searched something adjacent, not the actual target term.
Trap 5: not investing in creative for the specific moment. The same generic ad you run to broad audiences will underperform against intent audiences because it doesn’t match the search that put the person in the list. Invest 20% of your intent budget in creative that speaks to the specific search or category. That’s what turns a good audience into great pipeline.
The compounding effect
The real payoff of intent-based targeting shows up at month three. By then, you’ve built a repeatable machine: keyword sets, audience uploads, creative templates, measurement pipelines. Your CAC is trending down. Your pipeline is denser. Sales stops asking marketing to explain the leads because the leads showed up with real intent behind them.
That’s when it stops being a test and becomes the way you buy paid media.
The reporting stack that keeps the machine honest
Intent audiences only compound if measurement is honest. Most first-time users report on top-line metrics (CTR, CPL) and miss the two numbers that actually matter: cost per opportunity created and pipeline-to-close ratio by audience source.
Set up the reporting stack once, before the first campaign launches, and the answers become obvious in month two. UTM tag every ad by audience type (intent vs broad), by keyword theme, and by creative version. Send the UTMs into your CRM as first-touch attribution fields. Build a weekly dashboard that segments opportunities and closed-won by first-touch source. Within 60 days the intent audiences will either be pulling their weight against broad or they won’t, and the data will tell you the story either way.
What the sales team notices first
Interesting side effect of intent-based targeting: sales stops complaining about lead quality after about week four. Because the leads arriving from intent audiences aren’t tire-kickers filling out forms because a $2 lead magnet caught them scrolling. They’re prospects who searched a specific evaluation query yesterday and clicked a message that spoke directly to that search. The demo-show rate typically climbs 15 to 25 percentage points. The average deal cycle shortens because the buyer is further along by the time they land in the pipeline. That’s the compounding no marketing report captures cleanly but every VP of Sales feels immediately.
The account-based marketing angle
Intent data isn’t just a broad B2B play. It’s the missing ingredient in most ABM programs.
Traditional ABM works by picking a target account list and running paid social plus outbound at everyone with the right title inside those accounts. The problem is that most people in those accounts aren’t in market at any given moment, so the ads and outbound both feel intrusive.
Layer intent data on top: filter your target account list to accounts where someone (any title, not just the buying committee) has searched a relevant intent keyword in the last 30 days. Now your ABM budget is focused on the 15 to 25% of your target list showing active buying signals right now, not the full list where 75% are quiet.
Springbolt clients running intent-informed ABM typically see meeting-set rates 3x to 5x higher than untriggered ABM at the same budget. The multiplier comes from timing, not creative.
The uncomfortable question about your current channel mix
If you’re managing a media budget and you haven’t tested intent audiences yet, ask yourself: what was the last channel change that meaningfully moved CAC in the last 24 months? For most teams, the answer is “nothing.” Broad Meta and Google spend has been optimized to death. Every quarterly performance review sounds the same.
That plateau is not a reflection of the team’s competence. It’s a reflection of every advertiser buying the same modeled audiences at the same auction. Intent-based custom audiences are the first meaningful channel change most performance teams can make in a year that isn’t just a creative iteration or a bidding tweak. It’s an actual different audience.
Test it once. If it doesn’t work in your category, at least you’ll know. If it does, you’ll wonder why you waited.
Frequently asked questions
How do I upload a list of device IDs to Facebook as a custom audience?
In Meta Ads Manager, go to Audiences > Create Audience > Custom Audience > Customer List. Upload the file of hashed device IDs (Springbolt provides this in Facebook’s required format). Meta processes the file and reports a match rate, typically 50 to 70% for US audiences. Once processed, the audience is available to any campaign as a targeting option.
What is a good match rate when uploading device IDs to an ad platform?
Match rates vary by platform. Facebook typically matches 50 to 70% of a US device ID file. TikTok matches 40 to 60%. Google Customer Match matches 30 to 45%. Unmatched IDs can still be activated through programmatic DSPs like DV360 or The Trade Desk.
How often should I refresh a custom audience built from search intent data?
Weekly at minimum. Search intent decays quickly. A device that searched a keyword 10 days ago is significantly less likely to convert than one that searched it yesterday. High-velocity purchase categories like software or financial services benefit from refresh cadences as short as every 3 days.
Can I use search intent audiences alongside my existing Meta Advantage+ campaigns?
Yes. The two approaches complement rather than compete. Advantage+ tends to scale reach efficiently for brand awareness while intent audiences drive higher-quality response for direct conversion. Most teams end up shifting 30 to 50% of broad budget into intent audiences within the first 60 days because the intent side produces better pipeline economics.
How do I know if my keyword is specific enough to build an intent audience?
The test is whether the phrase indicates purchase intent versus research intent. ‘TPRM software comparison’ indicates evaluation. ‘What is TPRM’ indicates research. Choose phrases people type when they are actively selecting a vendor, not when they are learning what a category is.
What kind of ad creative works best with intent audiences?
Creative that acknowledges the moment. If someone just searched ‘vendor risk assessment software,’ the ad should speak to that specific evaluation, not run a generic brand message. Reference the category, the decision, or the specific problem. Generic broadcast creative wastes the targeting quality.