From Opaque Match Lists to Scalable Lookalikes — How Verkada 3.9× Pipeline Efficiency

Primer lets us shift away from brittle CSV match lists to dynamic seed audiences purpose-built for the offer – increasing pipe-to-spend efficiency 3.9x for MOFU campaigns.”

Senior Growth Marketing Manager

Patrick McMahon

3.9× higher

pipe-to-spend efficiency vs. previous match lists

4× more

audience experiments in hours, not days

Lookalike audiences

scaled from 1% to 4% without rapid fatigue

$20 qualified

CPL on niche campaigns

Verkada replaced low-match rate audiences and “best guess” targeting with transparent audiences and scalable lookalikes. The result was a Meta program that could expand reach without sacrificing persona precision.

The Overarching Challenge

Verkada’s product is persona-sensitive.

If the ad reaches the wrong role — marketing, HR, or unrelated teams — the message fails instantly. The challenge wasn't reach, it was precision at scale.

Before Primer, the team relied on Meta's native targeting: job titles, industries, and interest layers, sometimes paired with their own CRM data uploaded as a custom audience. The problem is that native ad platform targeting was built for consumer marketing, not B2B. The filters Verkada actually needs to define its ICP — accurate job titles, seniority, company headcount, industry, function — either don't exist on Meta or are too coarse and self-reported to be reliable.

"You can target 'IT' on Meta, but you can't tell it you want IT Directors and above at companies with 500+ employees in industries that buy physical security," Patrick said. "That's the audience. And the platform just doesn't give you those controls."

CRM-based custom audiences helped on precision but match rates rarely cleared 25%, and the seeds were too small to scale without fatigue. The team needed a way to define audiences using real B2B firmographic and demographic logic, see exactly who was in them before launch, and use those seeds to scale through lookalikes without losing persona integrity.

As Patrick puts it:

"Primer’s database has consistently proven to be accurate and up to date, especially for niche roles in the physical security space. That allows us to stay focused on our products, promotional campaigns, and engaged leads who are interested in our business, rather than spending time managing a ton of cold contacts who may never convert.”

Verkada’s Audience Scaling Framework

Verkada replaced their list-driven workflow with a four-step framework built around precision seeds, algorithmic scale, and faster experimentation.

Step 1 — Build a High-Integrity Seed Audience

Instead of importing large datasets, Verkada began defining audiences directly from ICP logic.

Using Primer’s demographic and firmographic filters, they created audiences based on:

job title contains logic
Industry and headcount filters
explicit role exclusions

Using Primer’s demographic selector, Verkada could define audiences using job titles and industries and preview who those audiences would include.

The team could see the composition of their targeting before launching campaigns — making the process far more transparent than traditional data vendors.

Patrick summarizes the strategy simply:

“Less is more. It’s better to have a small audience of really good titles — even if it’s 50K — than a big audience of mixed results.”

A smaller but higher-integrity seed produced stronger match lists and more reliable lookalike performance.

Audience design rules

Keep seed audiences ~50k–100k users
Avoid mixed roles (e.g., generic “Manager” titles)
Exclude operational roles that inflate audience size

Step 2 — Scale Through Lookalike Expansion

Once the seed audience performs, Verkada scales reach through a structured lookalike ladder. Typical expansion sequence:

Start advertising to the custom audience.
After fatigue sets in, launch 1% lookalike (highest similarity)
Once performance flattens again, add 2% lookalike
Expand to 3% and 4% as frequency rises
When larger audiences fatigue, rotate back to 1%

This produces audiences of 1–2M users that maintain persona relevance. “That audience of two million doesn’t go stale fast, and by the time you're on the 4% lookalike, you can go back to the 1% again.”

This solved a core scaling challenge in B2B paid social: audience staleness. Instead of small audiences exhausting quickly, the lookalike system created millions of qualified prospects. It also counteracts a common failure mode in the Andromeda algorithm—over-concentrating spend on a narrow pocket of users—by continuously refreshing the eligible audience pool.

Because the seed audience was tightly aligned with Verkada’s ICP, the lookalikes consistently outperformed other audience foundations.

Step 3 — Feed Meta Higher-Value Signals

Verkada didn’t rely on targeting alone. They also improved the signals Meta’s algorithm learned from.

Instead of optimizing purely for form submissions, the team sends lead score values from Salesforce back to Meta, allowing the platform to prioritize higher-quality prospects.

“We’re always fine-tuning the lead score and the value we send back.”

For example, personas that historically drive stronger pipeline — such as IT Directors and physical security leaders — receive higher value weights.

This allows Meta to optimize not just for conversions, but for pipeline quality and expected deal value.

Over time, this feedback loop helps the algorithm find more buyers that resemble Verkada’s best customers.

Step 4 — Let Creative Do Its Job

With a large enough audience, Verkada could finally let Meta’s algorithm operate properly. Without lookalikes, audiences are often too small to support algorithmic learning.

“You can’t really make use of Meta’s algorithm when your audience is only five figures.”

Now campaigns run with a controlled but diverse creative set:

3–5 creatives per ad set
different formats (static, video, GIF) - creative diversity helps
regular creative swaps to avoid fatigue

Launching 20+ creatives at once proved unnecessary. “The algorithm will just find the best three or four anyway.” This balance allows Meta to learn while still keeping creative experimentation structured.

Verkada also uses a creative performance as a diagnostic signal. Patrick’s rule:

Strong creative performs across multiple audiences
Weak creative fails everywhere

This makes it easier to isolate whether performance changes are caused by targeting or creative. If a creative performs well across several audiences, the message is likely strong. If it fails everywhere, the problem is usually the creative itself — not the targeting

Step 5 — Accelerate Mid-Funnel Experimentation

The biggest operational change was speed. Before Primer, testing a niche audience required iterating on layered native targeting, launching campaigns blind, and waiting days to see if the composition was rightThat process made experimentation slow and expensive. With ICP targeting directly inside Primer, Verkada could launch audiences immediately.

This unlocked a new MOFU testing workflow:

Launch a content offer to a niche audience
Let Meta optimize engagement
Validate which segments respond
Retarget those audiences with demo offers

One example was a webinar campaign. The audience responded extremely well.

“The target audience just ate it up.”

The campaign generated $20 CPL leads, which were later retargeted with demo offers to drive pipeline. This workflow allowed Verkada to test 4x times faster than before.

Key Principles from Verkada’s Playbook

Theme	Takeaway
Precision first	Smaller, cleaner seed audiences outperform large mixed lists
Lookalikes create scale	Expanding 1% → 4% prevents fatigue
Signal quality matters	Lead scoring improves algorithm learning
Creative diversity > volume	3–5 strong creatives outperform large batches
Logic beats lists	Transparent targeting replaces opaque data purchases

The Outcome

Switching from purchased match lists to ICP-driven audiences fundamentally changed Verkada’s paid-social performance. Results across campaigns showed: