# SHANZAY: 4X to 10X ROAS in three months

> An integrated performance, social, AI-assisted creative, and enquiry-automation program for a jewelry brand, run as one loop from media to enquiry handling.

Source: https://www.truenorthmarketing.ae/en/case-studies/shanzay
Client: SHANZAY
Published: 2024-01-25
Disciplines: Performance Marketing, Social Media Marketing, AI-Assisted Creative, Enquiry Automation
Publisher: True North Marketing (truenorthmarketing.ae)

## Challenge

SHANZAY entered at roughly 4X ROAS with a credible product and site, but budget allocation, creative, and audiences had stopped moving the account forward. Jewelry runs on trust, gifting, and detail. The softlines playbook was capping it, and repetitive enquiries were consuming the team.

## Approach

Own the marketing system, not just the ad account. SKU economics and measurement first, then AI-assisted creative testing under human review, gifting and self-purchase intent separated, audiences refined on clean conversion signals, and scaling held to margin and fulfilment, with enquiry handling automated inside clear guardrails.

## Outcome

Reported ROAS moved from roughly 4X to about 10X in three months. CAC fell 42%, CTR rose 47%, audience conversion improved 3.2X, monthly revenue tripled, and AI chatbots absorbed about 85% of repetitive enquiries.

## What We Did

Performance media across Meta and Google, social content system, AI-assisted creative production with human approval on claims and pricing, audience refinement and scaling discipline, enquiry chatbot automation, and the commercial reporting underneath it.

## Full article

![SHANZAY jewelry campaign creative](../images/case-shanzay-cover.jpg)

SHANZAY entered the case-study window at roughly 4X ROAS and reached roughly 10X three months later. The approved record also shows a 42% reduction in CAC, a 47% lift in CTR, a 3.2X improvement in audience conversion, tripled monthly revenue, and an AI chatbot handling about 85% of repetitive enquiries.

Those outcomes came from an integrated marketing program spanning performance media, social media marketing, AI-assisted creative testing under human review, audience refinement, and enquiry automation. Social content supplied the hooks, objections, and proof that paid campaigns could test and return to the wider brand system. The evidence does not isolate one tactic as the cause, so I treat the figures as one account's historical outcome, not a reusable promise.

  
  
  

## The baseline and the scope

The baseline was approximately 4X ROAS. The work covered media, creative iteration, audience conversion, CAC, revenue, and automated handling of repetitive enquiries. That is the scope supported by the approved case record; this case study does not claim a SKU-allocation diagnosis or a precise week-by-week sequence that the source does not document.

## Why jewelry is not fashion

| Factor | Fashion (apparel) | Jewelry / accessories |
| --- | --- | --- |
| Consideration | Often trend- and fit-driven | Trust, authenticity, and detail |
| AOV strategy | Bundles, multi-SKU carts | Gifting, engraving, premium packaging |
| Creative risk | Lifestyle and UGC scale more easily | Macro detail, reflection, and careful claims |
| Seasonality | Collections and drops | Occasions, holidays, cultural calendars |

Jewelry sells on trust and detail, not trend and fit, so the creative and landing experience have to reduce doubt before you scale, not after. Applying a softlines playbook to a high-AOV, gifting-heavy category is the quiet reason many jewelry accounts plateau. The [fashion program we ran](/en/case-studies/joey-and-pooh) shared the discipline but not the tactics.

## The operating system behind the result

This was marketing ownership, not ad-account maintenance. The team connected the social narrative, creative production, paid distribution, audience feedback, enquiry handling, and commercial reporting so each workstream improved the next one.

### Measurement before expansion

Before expanding spend, the account needs one agreed reading of purchase events, CAC, and revenue. The public case record shows the change in these outcomes, while the detailed event map and platform exports remain private.

Do not expand a campaign while the platform and order system disagree. Reconcile the conversion and revenue records first; otherwise a higher reported ROAS may be a measurement change rather than a commercial one.

### Creative and offer iteration

We increased variant testing across hooks, gifting, self-purchase, materials, and occasion windows, and tightened landing parity with the product pages. AI-assisted drafting accelerated the variants; humans approved claims, policy, and brand fit, which matters acutely in jewelry. We do not publish tactic-level lift tables here, because those figures are not comparable across accounts without full margin and audience disclosure.

### Audience and scaling discipline

- Refined lookalike and intent stacks once the clean conversion events were reliable.
- Paused clusters that produced clicks without purchases, once statistical confidence allowed.
- Scaled only where return and fulfillment could absorb the additional volume.

### Automation within guardrails

Automation helped with budget shifts inside defined rules, creative rotation, and reporting. We avoided blind automation on claims, pricing, and customer-facing messaging without review.

## Results

Reported ROAS moved from roughly 4X to roughly 10X over the three-month case-study window.

### Approved public proof points

These proof points are public-safe for this case study, but they still belong to one account with its own margin, fulfillment, and demand pattern. The point is the operating method, not a reusable promise.

| Signal | Change |
| --- | --- |
| Reported ROAS | Roughly 4X to about 10X |
| Monthly revenue | Tripled during the program window |
| CAC | Down 42% |
| Creative testing | CTR up 47% after AI-assisted variant testing |
| Audience conversion | 3.2X improvement on the lookalike conversion signal |
| Enquiry handling | AI chatbots handled about 85% of repetitive inquiries |

## Why this worked beyond "using AI"

AI was an accelerator, not the strategy. The driver was operational discipline: clear hypotheses by funnel stage, aggressive creative iteration with fast pruning, audience refinement from conversion-quality feedback, and controlled scaling only after a consistent weekly signal. The repeatable version:

- Audit SKU-level performance before you increase spend.
- Build separate hooks for gifting, self-purchase, and occasion intent.
- Use dynamic retargeting with product-level relevance and frequency caps.
- Track payback and margin, not blended ROAS alone.

## What must be watched after the program

A 10X result is useful only while the underlying measurement and economics hold. Creative performance, event integrity, CAC, fulfilment, and repeat behaviour all need continued review; the headline multiple cannot answer those questions by itself.

Creative can fatigue and tracking can drift as products, offers, and campaigns change. The responsible response is to use the next batch and the reconciled revenue record to test whether the result is holding, rather than assuming the historical multiple is permanent.

A strong first-purchase ROAS can mask weak repeat behavior in a gifting-heavy catalog. The durable scoreboard was repeat purchase rate, payback period, and contribution margin by campaign, reviewed weekly with returns netted out. That unglamorous part is what decides whether the number survives the next quarter.

## The number belongs to the account

The 10X belongs to SHANZAY's account, offer, market, and measurement window. What another brand can reuse is the practice of connecting creative iteration, audience quality, enquiry handling, and commercial reporting so that no team optimizes its own dashboard in isolation.

The SKU-economics-first, intent-specific, margin-led process transfers to watches, furniture, and other considered purchases, but the discipline to defend the number is the part most teams underestimate. For a scoped review, [contact True North](/en/contact) with your catalog, margin bands, and current creative batches. The same care that protects [material truth in AI product photography](/en/blog/ai-product-photography) protects trust here, where the render has to match the piece in the customer's hand.