How AI Marketing Lifted Deep Blue Health Revenue by 30%

Deep Blue Health (DBH) is a New Zealand health supplement brand with a solid product line and a loyal customer base.

When we started working together, they weren't struggling — but they'd plateau'd. Their marketing was functional. Google Ads were running. Some social content. A basic email list. The kind of setup that keeps the lights on but doesn't accelerate growth.

The brief was simple: rebuild the marketing operation for compounding performance. Not a campaign. A system.

Here's an honest account of what we built and what happened.

The Starting Point

Before touching anything, we ran a proper audit:

Google Ads: Active campaigns but low Quality Score across the board. Ad groups were too broad, match types were mixed, and conversion tracking was partially broken. Money was moving but we couldn't fully trust the attribution.

Meta Ads: A few boosted posts. No proper campaign structure, no retargeting audiences, no creative testing framework. Significant untapped potential.

Content: Inconsistent. No SEO strategy. Product pages were thin. The blog hadn't been updated in months.

Customer Support: The team was handling a high volume of repetitive queries via email — shipping timelines, product questions, returns. This was eating significant staff time.

Email: A basic welcome sequence. No post-purchase flows. No re-engagement campaigns. The list was an underutilised asset.

Five distinct areas. Each one improvable. The question was sequencing.

What We Built, In Order

1. Fix the Foundation (Month 1)

Before scaling spend, we fixed conversion tracking. This sounds boring. It's critical. You can't optimise what you can't measure.

Then we rebuilt the Google Ads structure from scratch. Tighter ad groups, exact and phrase match taking precedence, negative keyword lists rebuilt, Quality Scores targeted above 7. Ad copy variants created (AI-assisted, human-reviewed) to enable genuine A/B testing.

Result at 30 days: Same spend, better attribution clarity, early Quality Score improvements.

2. Build the Meta Engine (Month 1-2)

We built a proper campaign structure: prospecting campaigns with lookalike audiences built from customer data, retargeting campaigns segmented by behaviour (visited but didn't buy, added to cart, purchased once, purchased twice+).

For creative, we used AI to generate a high volume of variants — different hooks, different value propositions, different formats. Not spray and pray — structured testing with clear hypotheses.

The platform does the work of finding winners. Our job was giving it enough creative variety to test meaningfully.

3. SEO From Scratch (Month 2)

DBH had never done SEO properly. This was the longest-horizon investment.

We did keyword research, identified the product and informational terms worth targeting, and rebuilt the product page copy with SEO in mind. We launched a blog with AI-assisted articles targeting mid-funnel educational queries.

SEO is a long game. You don't see results in 30 days. But month 6, 12, 18 — that's where it compounds.

4. AI Customer Support Layer (Month 2)

We built an AI agent to handle tier-1 customer queries: shipping timelines, product information, returns process, general questions. The agent was trained on their specific products and policies.

Within 30 days, it was handling approximately 80% of incoming queries autonomously. The team's time went from reactive support to proactive customer success.

Hours saved per week: ~15

5. Email Automation (Month 2-3)

Built out the full email lifecycle:

Welcome series (5 emails, value-first, builds relationship before asking for purchase)
Post-purchase sequence (thank you, usage tips, review request, cross-sell)
Re-engagement campaign (targeting customers who hadn't purchased in 90 days)
Abandoned cart sequence

All AI-written, human-edited, brand-voice consistent.

November: Record Month

November is a significant month for retail in New Zealand — it includes Black Friday and pre-Christmas purchasing.

DBH's November sales beat their store record by over 30%.

That wasn't one lever. It was the compound effect of 2-3 months of system building: Google Ads running cleaner, Meta creative engine producing winners, email sequences firing at the right moments, SEO starting to contribute.

We can't attribute the exact lift to any single initiative because that's not how systems work. Everything was running better simultaneously.

What "AI Marketing" Actually Meant Here

I want to be specific about where AI featured and where humans were still essential:

AI did: Generate creative variants at scale, write email sequences, handle customer queries, assist with article drafts, optimise campaign performance in-platform.

Humans did: Strategy decisions, creative direction, brand voice editing, anything requiring business judgment, exception handling for complex customer situations.

The human-in-the-loop piece matters. The 20% of customer queries the AI escalated were often the most sensitive — unhappy customers, complex situations, edge cases. Those needed human care.

The system created capacity. Humans directed that capacity to higher-value work.

What This Looks Like as a Model

If you're a product business with an established customer base, a functional but underperforming marketing operation, and the willingness to invest 2-3 months in building properly — this model scales.

The investment is real. The setup phase involves genuine work and genuine cost. But the leverage on the other side is also real: a marketing system that runs, learns, and compounds.

That's what I build.

Apply to work together →

Tom Hall-Taylor is an AI-native marketing consultant based in Auckland, New Zealand. He builds integrated AI marketing systems for select businesses. Applications reviewed monthly.