The Situation

Matroid provides advanced computer vision AI used in manufacturing environments to detect defects, improve safety, and automate quality inspection.

The company wanted to explore X (Twitter) Ads as a potential acquisition channel capable of reaching engineers, operators, and AI professionals working in industrial environments.

The challenge was that X offers far fewer targeting controls than platforms like LinkedIn or Meta. Before committing significant budget, the first question was whether qualified B2B leads could realistically be generated on the platform at all.

The Primary Challenge

Before optimization could matter, the channel had to clear a more basic bar: could X Ads reach qualified B2B buyers in a niche industrial vertical at all?

Limited audience segmentation compared to LinkedIn or Meta meant spending into the channel without that evidence carried real risk.

The Goal

Validate whether X Ads could produce qualified signups from technical audiences in manufacturing and AI, and identify the specific audience segments, creative formats, and messaging structures capable of driving conversions.

This was a channel validation exercise, not a scaling initiative.

Key Outcomes

• 7 confirmed lead signups from $4,650 in total spend
• 22 form intent events recorded
• 3.2 million impressions delivered
• Engagement rate maintained above 2%
• 100% of confirmed conversions came from narrow, manufacturing-focused targeting

Our Approach

We structured the engagement as a three-phase test, designed to validate, challenge, and confirm findings before drawing conclusions.

Each phase isolated a specific variable. Budget remained controlled throughout. The goal was signal clarity, not volume.

Execution Highlights

Phase 1: Initial Validation

The first campaign launched with a controlled $1,000 test budget targeting two audience groups: manufacturing-focused and AI-focused.

This early test generated four confirmed signups from the initial spend, indicating that X could produce conversions even in a narrow B2B vertical.

Two conversion metrics were monitored: confirmed signup submissions and form intent events, which captured users who began but did not complete the signup process.

The early results justified continuing the testing process.

Phase 2: Expanding Targeting

After Phase 1, campaign targeting was expanded based on platform recommendations. This included broader interests, wider audience pools, and increased reach across related technical topics.

The assumption was that a wider audience would allow the campaign to scale.

The opposite happened.

The two campaigns using expanded targeting produced zero confirmed signups and only generated form intent events. Even after reverting creative back to the original top performers, results remained poor.

This isolated the issue: targeting dilution. The broader audiences introduced users interested in AI topics but who were not qualified manufacturing buyers.

Phase 3: Returning to Precision

To confirm the findings, the original campaign structure was duplicated and relaunched with the same tight targeting parameters from Phase 1.

Confirmed signups returned immediately.

The conclusion was unambiguous: narrow targeting consistently outperformed expanded audiences across every phase of testing.

What Actually Converted

Audience

The manufacturing-focused audience produced all confirmed conversions. AI-interest audiences generated engagement and form intent events but failed to produce completed signups.

The actual buyer was operational, not theoretical. This distinction shaped every recommendation that followed.

Creative

The top-performing ad used extremely simple creative: a straightforward carousel format with clear examples of manufacturing defect detection, no decorative overlays, and direct technical messaging. This ad generated the majority of recorded conversions.

The second-best performing creative was a video version of the same message showing real manufacturing footage.

The common thread was clarity. Direct explanations of how the technology improves manufacturing processes outperformed broader AI positioning every time.

Device and Demographic Signals

Device data showed a strong skew toward mobile. iPhones accounted for 82% of conversion events, with Android and desktop each at roughly 9%.

The highest converting demographic was males aged 35 to 54, aligning closely with mid-level and senior engineering or operations roles inside manufacturing organizations.

Constraints We Navigated

• Limited platform targeting controls compared to LinkedIn or Meta, requiring tighter creative and audience discipline to compensate
• Small conversion volume in a niche vertical, requiring careful signal interpretation rather than statistical scale
• Platform recommendations that actively pushed toward broader targeting, which had to be tested and ultimately rejected based on data

The testing framework was built to withstand these constraints rather than be distorted by them.

Why This Worked

The leverage point was audience discipline.

X Ads can generate conversions for niche B2B companies, but only when targeting remains tightly aligned with the actual buyer. When targeting expanded, signal quality dropped and the algorithm could not find qualified users in a diluted pool. When targeting stayed narrow and aligned with operational buyers, conversions returned.

The three-phase structure made this visible. Without deliberately testing the expansion hypothesis and watching it fail, the instinct to broaden audiences would have persisted and quietly eroded performance.

Discipline in testing produced clarity. Clarity produced a validated channel.

Strategic Takeaway

Channel validation does not require large budgets. It requires structured testing and the willingness to let data overrule assumptions.

This engagement confirmed three things: X Ads can produce niche B2B leads when targeting remains disciplined, operational audiences outperform broader interest-based audiences in technical verticals, and clear technical messaging converts better than high-level positioning.

The result was not just leads. It was a validated acquisition channel with a clear playbook for what works, what does not, and why.