The challenge
Kaaya is a fast-growing D2C ayurvedic skincare brand selling across India and the GCC. Their Shopify storefront did the basics well but treated every visitor identically — the same hero, the same product grid, the same generic quiz. With paid CAC climbing past INR 940 and a 1.4% site-wide conversion rate, the founders knew the next leg of growth had to come from the experience itself, not more ad spend.
What we built
A headless rebuild on Next.js with Shopify Storefront API, fronted by an AI personalization layer:
- A skin-and-routine concierge agent powered by OpenAI GPT and Anthropic Claude (model-routed by query type) that conducts a 6-question diagnostic in conversational Hindi, English, or Hinglish and returns a tailored regimen with reasoning the customer can actually read.
- A vector-DB recommender (Weaviate) trained on 18 months of order history, product reviews, and ingredient-affinity data, so a visitor who selects "combination skin + monsoon humidity + fragrance-sensitive" sees a different bestseller list than the next visitor.
- Server-side personalized hero, social proof, and bundle blocks that render before first paint based on geolocation, referral source, and returning-visitor cohort — no client-side flicker.
- An n8n automation flow that nurtures abandoned diagnostics via WhatsApp Business API with a hand-off to the same agent context, so the conversation continues instead of restarting.
The whole stack is instrumented with PostHog so the team sees which agent decisions move revenue and which don't.
Results
Revenue per visitor climbed 3.2x within 90 days of launch. Add-to-cart rate more than doubled (+118%) and the 90-day repeat purchase rate rose 47% as customers built routines around the regimens the agent recommended. Average order value grew 38% on the strength of bundle suggestions. Paid CAC payback shortened from 4.1 months to 1.6.