AXO for retail & ecommerce.

Post-click + inbound personalization

Every visitor lands on a page tuned to how they arrived and what they're doing in the first few seconds.

A Comparison Shopper arriving from Google CPC sees a price-anchored hero with reviews above the fold; a Surfer from social sees lifestyle imagery and a low-friction "browse the collection" CTA — same page, different shopper.

Why it matters: The brand spent money to get them here. Default "same homepage for everyone" wastes that spend.

• Anonymous behavioral signals · Data collection
  Click depth, scroll velocity + depth, dwell, hover, pause, tab-switch, cadence — collected without requiring identity.
• UTM + referrer + device context · Data collection
  Every event carries the channel, referrer, viewport, and device class for downstream targeting.
• Per-vertical default segment packs · Audience segmentation
  6+ segments seeded per industry — Deliberator / Comparison Shopper / Gifter for retail; Donor Ready / Lapsed for fundraising; etc.
• Zone-based authoring · Site personalization
  Any DOM element with [data-vt-zone] becomes an addressable surface. Visual + crawler-driven discovery.
• 9 variant types · Site personalization
  copy_swap, hero_swap, overlay, lightbox, banner, badge, product_reco, lead_form, move_object, plus donation_form for fundraising.
• Flicker-safe rendering · Site personalization
  Per-zone pre-hide preserves LCP — no body-blank flash, no fight with the CDN.
• Self-healing zone resolution · Site personalization
  Selector fallback ladder + MutationObserver for SPAs — variants survive ordinary HTML churn.

Behavioral audience discovery

Anonymous clickstream surfaces real shopper cohorts without a CDP team or an audience-builder UI.

Day one of collection: the dashboard surfaces Deliberators, Surfers, Converters, Comparison Shoppers, Gifters, and Engaged Paid Arrivals — named with sample sessions you can click into.

Why it matters: Most retail brands either lack a CDP or have one their marketing team can't operate. AXO replaces "audience building" as a job.

• Per-vertical default segment packs · Audience segmentation
  6+ segments seeded per industry — Deliberator / Comparison Shopper / Gifter for retail; Donor Ready / Lapsed for fundraising; etc.
• No-code custom segment builder · Audience segmentation
  Composable predicates: event_occurred, page_viewed, utm_matched, propensity, feature_anomaly, lifecycle, affinity.
• Hidden-persona ML clustering · Audience segmentation
  Surfaces audience cohorts the operator didn't pre-define, with stability + lift validation.
• Audience lens in the variant editor · Audience segmentation
  Author variants directly against named segments + see provenance badges (which signal sourced each match).
• CRM-attribute-driven segments · Audience segmentation
  When a CRM is connected, segments can key off lifetime value, last gift, plan tier, etc.

Catalog-aware recommendations

Every recommendation slate comes from your live catalog, in stock and on brand, tuned per segment.

PDP "you may also like" surfaces 4 items: 2 cooccurrence-driven from your catalog, 1 bestseller from the same category, 1 new-arrival — all in stock, all in the brand voice you uploaded.

Why it matters: Recommendations are the moment of truth in any ecom session. Generic carousels miss; segment-aware in-stock recs convert.

• Catalog auto-import · Recommendations
  Shopify native; custom feed via importer. Reclassification + brand-guideline extraction at ingest.
• 3 reco strategies · Recommendations
  Curated (operator-pinned), rec_engine (cooccurrence + embeddings), mixed (pin some, fill the rest).
• 4 ranking modes · Recommendations
  session_behavior, bestsellers, new_arrivals, similar — chosen per zone + per segment.
• In-stock + on-brand filtering · Recommendations
  Recommender respects inventory and the brand-guideline registry automatically.
• Shopify source · CRM + commerce
  Native Shopify app: tag install, catalog ingest, order webhook → AXO profile.
• BYO DAM (Cloudinary, S3-compatible) · AI + agentic
  Bind brand-team-curated assets to variants without duplication. Push-through architecture, 7-day retry TTL.

Journey + message orchestration

Cart-abandonment, browse-abandonment, win-back, lifecycle messages — fire through the ESP / SMS / ad-platform you already use.

Cart abandoned → 1-hour-delay Klaviyo email scoped to Comparison Shoppers (with a value prop) vs. Deliberators (with social proof); if no purchase in 24h, fire a Meta retargeting audience push via CAPI.

Why it matters: Abandonment recovery is the highest-ROI lifecycle motion in retail. AXO routes the logic through your existing Klaviyo, not a competing platform.

• Trigger engine · Journey + message orchestration
  Composable eligibility (segment + UTM + behavior, all_of / any_of) → render an action or fire to a destination.
• Abandonment templates pre-baked · Journey + message orchestration
  Cart, browse, form, donation, demo-request templates ready to clone per destination.
• Email/SMS destinations · Email + SMS destinations
  Klaviyo, Braze, Customer.io, Resend, SendGrid, Postmark, Mailchimp, Twilio (SMS), Intercom, SFMC, Emarsys.
• Ad-platform server-side CAPI · Ad platforms
  Meta, Google Ads, GA4 MP, TikTok, Pinterest, Reddit, LinkedIn — server-side conversion API for iOS-14-safe attribution.
• Composable eligibility predicates · Journey + message orchestration
  segment_member + utm_matched + event_occurred + visitor_lifecycle + propensity, AND/OR composable.
• Multi-channel dispatch · Journey + message orchestration
  One trigger can fire to email + SMS + ad-CAPI + webhook simultaneously, each with its own template.

Ad-platform retargeting + CAPI measurement

Server-side conversion events for iOS-14-safe attribution. Build retargeting lists from behavior. Suppress converters from spend.

Purchase event fires server-side to Meta + Google + TikTok via CAPI — survives iOS 14.5+ tracking limits. Cart abandoners flow into a Meta Custom Audience automatically. Converters get suppressed from acquisition campaigns within minutes.

Why it matters: iOS attribution loss has hammered ecom ROAS. Server-side CAPI + behavior-built audiences restore measurement and cut wasted impressions.

• Ad-platform server-side CAPI · Ad platforms
  Meta, Google Ads, GA4 MP, TikTok, Pinterest, Reddit, LinkedIn — server-side conversion API for iOS-14-safe attribution.
• Ad-source ↔ campaign auto-link · Ad platforms
  UTM source detected on-site → linked to the matching ad-platform campaign for closed-loop measurement.
• Behavior → ad-platform audience export · Ad platforms
  Push behaviorally-built segments to Meta Custom Audiences, Google Customer Match, LinkedIn audiences.
• Trigger engine · Journey + message orchestration
  Composable eligibility (segment + UTM + behavior, all_of / any_of) → render an action or fire to a destination.
• Composable eligibility predicates · Journey + message orchestration
  segment_member + utm_matched + event_occurred + visitor_lifecycle + propensity, AND/OR composable.

Holdout-proven incremental lift

A 10% control group runs against every change, always on. Causal attribution, not correlation.

Every variant ships to 90% of matching shoppers; 10% stay in control. The Scorecard reads "Comparison Shopper hero variant: +4.2% conversion lift over holdout, p<0.05, +$18.3k incremental revenue last 30 days."

Why it matters: Every personalization vendor claims uplift. Only the ones holding out a control group can prove it. Critical for CMOs reporting to a board.

• 10% always-on holdout · Measurement
  Live control group on every change. Real incremental lift, not asserted lift.
• Lift attribution per variant + per segment · Measurement
  Causal attribution scoped to the specific variant + the specific cohort it served — not site-wide proxies.
• AI Scorecard · Measurement
  Plain-language "is the AI working?" verdict for a non-technical reader. Holdout vs. personalized side-by-side.
• Bandit optimizer · Measurement
  Auto-allocates traffic to winning variants once statistical significance crosses threshold.
• Conversion funnel by segment · Measurement
  Per-segment funnel view — where each cohort drops off, where they don't.
• Per-variant render flicker telemetry · Measurement
  Real-user-monitoring metrics per zone — pre-hide duration, decode wait, reveal source.

Agentic creative + ops

Drive AXO from natural language — in-app via the AXO Assistant or externally via Claude / Cursor / ChatGPT / VS Code through MCP.

"Draft three Comparison Shopper hero variants for Pawnest's PDP — price-anchored, reviews-led, and shipping-led." Agent reads your brand voice + active campaign brief, generates with provenance, queues for human review.

Why it matters: Marketing teams don't have engineers. The agentic surface is what gets non-technical operators from "could do this" to "did this today."

• AXO Assistant (in-app) · AI + agentic
  Chat-based agent that drives AXO from natural language. Reads your data, takes actions, asks for confirmation on writes.
• MCP server · AI + agentic
  Connect Claude, Cursor, ChatGPT, VS Code, or any MCP-compatible tool. Same toolset as the in-app assistant.
• AI-drafted variant copy · AI + agentic
  Brand-voice guardrails + active-campaign context baked in. Drafts go through human review before publishing.
• AI provenance audit trail · AI + agentic
  Every AI generation logs the prompt, brand voice, active brief, model, and DAM provenance. Compliance-grade.
• BYOM text generation · AI + agentic
  vLLM, Ollama, Azure OpenAI, LM Studio — OpenAI-compatible. Route variant authoring through your own model.
• BYOK image generation · AI + agentic
  Gemini, OpenAI, Adobe Firefly, Replicate — your keys, your costs, your model choice.
• BYO DAM (Cloudinary, S3-compatible) · AI + agentic
  Bind brand-team-curated assets to variants without duplication. Push-through architecture, 7-day retry TTL.

First-party data activation

Profile + event data flows to the CDP, warehouse and CRM you already run — Segment, Snowflake, BigQuery, Salesforce, HubSpot. AXO feeds your stack, doesn't replace it.

Behavioral cohort assignments stream to Segment + Snowflake for downstream BI; purchase events sync to Klaviyo + Salesforce CRM in real time; cart-value distributions land in BigQuery for finance.

Why it matters: Brands invested in a data stack don't want a new island. AXO sits on top, feeds the stack, doesn't compete with it.

• Warehouse destinations · Data infrastructure
  Snowflake, BigQuery, Amazon S3 — stream profile + event copies to your data plane.
• CDP destinations · Data infrastructure
  Segment, RudderStack, mParticle, Adobe AEP — works alongside existing CDPs without replacing them.
• CRM destinations · CRM + commerce
  Salesforce, HubSpot, Intercom — push profile updates + activity, read CRM attributes back.
• Shopify source · CRM + commerce
  Native Shopify app: tag install, catalog ingest, order webhook → AXO profile.
• Signed custom webhook · Data infrastructure
  Generic HMAC-signed POST to any URL you own — escape hatch for anything not in the catalog.