How to build and scale a performance marketing function

Executive overview

Most companies underinvest in paid because they treat it as optional. It isn't. Organic reach is collapsing — Google buries listings below paid results, Meta is near pay-to-play. At minimum, every company should run paid search.

The real leverage is knowing where users already find you, running small "signs of life" tests on those platforms, and only scaling once the signal is clear. The mistake agencies make — and the reason in-house teams consistently outperform them — is failing to go deep enough into the data.

The job of a performance marketer is to find the signal and ignore the noise.

Who should invest in paid, and when

• Paid search is user-driven (someone typed a relevant keyword) — it applies to almost every business.
• Display and social ads are disruptive media; they make less sense until you have the right creative and a clear audience.
• If you're not sure where to start: Google search first, then Meta (Facebook/Instagram), then YouTube if you have video.
• LinkedIn is 3x more expensive than other channels but uniquely powerful for high-LTV B2B — only add it after validating Google and Meta.
• Emerging channels worth attention: connected TV, podcast ads. VR and AI advertising are too early to bet on.
• Do not invest heavily before product-market fit. Low conversion rates waste budget and create lasting negative brand impressions.

Signs of life tests

• Start with your own customer data — upload your list and build lookalike audiences (1%, 2–4%, 5–7%, 8%+).
• Begin with the 1% lookalike when budget is constrained; it's the most correlated to your existing customers.
• Budget, not statistical significance, usually determines test length — build an opsize from early results to justify more spend.
• To diagnose poor performance: check targeting first (are the right users seeing the ad?), then engagement (click-through rate), then creative.
• Use focus groups or controlled A/B tests (same copy, different creative vs. same creative, different copy) to isolate the variable.
• Give yourself permission to fail — either you're winning or you're learning.

Platform strategy

• Each platform has different user behavior; creative that works on Meta rarely transfers unchanged to TikTok.
• Video is outperforming other formats when measured correctly — build a creative flywheel (steady supply of fresh video content).
• Start YouTube ads with emotion (comedy, warmth) — users who feel something remember the ad and take action later.
• LinkedIn targeting by job title, company, and geography is uniquely powerful for enterprise deals (example: geo-fencing a prospect's office and targeting their decision-makers with specific objection-handling ads).
• Podcast and connected TV ads are performing well for companies that can measure them correctly.

Agencies and early hiring

• Agencies apply playbooks across clients; they rarely go deep on any single account.
• Use an agency or consultant to get started, but set a clear milestone (e.g., $50K/month spend) to trigger an in-house hire conversation.
• Have that transition conversation with the agency upfront — a good one will welcome it; reluctance is a red flag.
• Your first in-house hire: a growth marketing specialist/manager — generalist who can run platforms and interpret data.
• Hire for data thinking, not platform knowledge. Interview question: throw a wall of metrics at candidates and ask what they'd optimize. The right answer starts with "What's the campaign goal?"
• Second hire: a creative (graphic design/branding) — in-house creatives match tone faster, iterate quicker, and aren't capped by agency hours.
• Third hire: a dedicated data scientist — builds incrementality tests, custom reports, and analyses that generalists can't produce.

Operating the team

• Run an ops cadence: a spreadsheet mapping each activity (keyword reviews, search query reports, ad copy tests, negative lists, finance checks) to its frequency (weekly, bi-weekly, monthly).
• This creates accountability, enables spot-checks, and gives instant answers to cross-functional questions.
• Use a capacity calculator each quarter: estimate days needed vs. days available. If you're in the red two quarters running, build a headcount case. If you're in the red once, cut low-value meetings first.
• Onboard new hires fast: assign real ownership in the first weeks, show your workflow over screen share, and set explicit expectations. Aim for impact at 30–45 days, not 90.

Metrics and reporting

• Work with finance to set CAC guardrails before running campaigns — don't try to derive benchmarks from raw account data alone.
• Key Google Ads signals to prioritize: ad strength (expected CTR, landing page experience, ad relevance), quality score, and click share (not just impression share).
• Ad relevance is the most commonly neglected lever — improving it raises quality score, which Google rewards with more impressions (roughly 12% lift from below-average to above-average).
• Pause poor-performing ads immediately — a single poor-quality ad drags down account-level quality score.
• Impression share measures how often you show up; click share measures whether the right users are responding. Optimise for click share.
• For competitor analysis, use the "true competition metric" from PPC Hero (Jacob Brown) — it uses auction insights to identify genuine threats vs. close-variant accidents or ego bidding.
• Don't chase generic industry benchmarks for CPC or CVR — ask your platform partner (Google, Meta, LinkedIn) to anonymise and benchmark you against your specific competitors.

Attribution and incrementality

• Multi-touch attribution with time decay is the recommended default — users forget where they first encountered a brand.
• Attribution alone is biased: it cannot tell you whether the user would have converted without the ad.
• Run geo-based holdout tests (GeoX) or conversion lift tests on each platform to measure true incrementality. All major platforms will partner on this.
• Encode results as an incrementality-adjusted factor (IAF) per channel, ideally by region.
• Don't run incrementality tests until you're spending at least $50K/month on a platform — below that threshold, the signal is too thin to be meaningful.
• Smart bidding, recommendations, and ad copy suggestions are all already AI — this isn't new. Generative AI's biggest near-term impact is on creative production and content.