Summary
Our client runs a U.S.-based high-end online jewelry store. He came to us with a focused challenge and a sharp idea: help gift shoppers feel confident enough to complete their purchase.
We built and launched an AI Gift-Finder that fits seamlessly into his Magento website. Within 90 days of going live, conversion rates nearly doubled. The solution is now on track to pay for itself in under eight months.
Cooperation Period
Dec 2023 – Feb 2025
Location
Industry
Business Challenge
10 000-SKU catalog, gift conversions stuck at 0.64 % versus the 1.18–1.73 % luxury benchmark. Shoppers hesitated on style, metal and stone size for four-figure gifts. Any solution had to slot into SFCC cartridges, meet CCPA for photo data, and keep first-contentful-paint < 2 s for paid-traffic SLAs.
Solutions
1
One-Screen Quiz & Secure Photo Intake
- Captures occasion, budget and style cues in < 60 s.
- AES-256 photo encryption; auto-purge after seven days; consent logged in Salesforce CDP.
2
Real-Time Style Engine
- Vision + NLP model fine-tuned on 42 k product images & 15 k style labels.
- p95 inference 1.1 s; monthly retrains absorb new SKUs and trend data.
3
SFCC Delivery Layer
- React cartridge injects on-image try-ons, “add to bag,” save & share — zero page reloads.
- Session events stream to GA4 and CDP for closed-loop optimization.
4
MLOps & Cost Control
- Kubernetes autoscaling proven to 1 200 parallel inferences; cloud cost ≈ $0.003 / session.
- Drift, latency and error SLAs monitored via Prometheus → Grafana dashboards.
Project Results
The project was completed resulting in a dramatic improvement in the online jewelry store’s performance:
Conversion Rate
Gift-session conversion rate increased from 0.64% (297 purchases out of 46,440 sessions) to 1.05% (513 purchases out of 48,890 sessions), representing a 64% lift, statistically significant (p < 0.01).
AOV
Average order value rose from $1,250 to $1,425, a 14% increase.
Return Rate
Return rate dropped from 8.2% to 6.8%, showing a 17% reduction.
