Who this is for

• Shopify and WooCommerce operators adding assistant flows for sales and support.
• Founders and PMs who need channel‑level ROI and payback windows.
• Marketing and RevOps teams instrumenting new surfaces without a full data team.

The outcomes you’ll get

• A clean event taxonomy for assistant journeys (voice, chat, in‑car).
• UTM and link recipes that survive deep links, short links, and phone calls.
• Dashboards that show revenue, AOV, CPA, and assisted conversions by assistant channel.

The KPI scoreboard to track from Day 1

Instrument these seven KPIs first. They’re comparable across channels and simple to explain to finance.

1. Revenue by channel: Alexa+, WhatsApp, In‑Car (and sub‑intents like reorder, track, book).
2. Conversion rate: Transactions / assistant sessions.
3. Average order value (AOV): Revenue / transactions by channel.
4. Cost per acquisition (CPA): Media + assistant ops cost / transactions.
5. Response rate: Replies / assistant messages or prompts sent.
6. Handoff rate: Human‑agent transfers / assistant sessions (aim to reduce without hurting CSAT).
7. Payback period: CAC / gross margin contribution from assistant‑sourced orders.

Event taxonomy: the 12 events you actually need

Keep the schema stable across surfaces so your dashboards align. Prefix events with the channel to make queries effortless.

{
  "alexa_session_start": {"channel":"alexa","intent":"browse|reorder|book|track_order"},
  "alexa_cta_click": {"destination":"/collections/winter","variant":"hero|recommendation"},
  "alexa_add_to_cart": {"sku":"ABC123","qty":1,"price":49.00},
  "alexa_checkout_start": {"cart_value":98.00,"items":2},
  "alexa_purchase": {"order_id":"A1001","revenue":98.00,"aov":49.00},

  "wa_session_start": {"channel":"whatsapp","flow":"support|reorder|browse"},
  "wa_handoff_to_human": {"reason":"refund|complex_issue|payment"},
  "wa_purchase": {"order_id":"W2001","revenue":64.00},

  "icar_session_start": {"channel":"in_car","context":"commute|errand|trip"},
  "icar_add_to_cart": {"sku":"XYZ999","qty":1},
  "icar_purchase": {"order_id":"C3001","revenue":120.00},

  "assistant_refund": {"order_id":"*","amount":*}
}

Implement on your storefront (client) and your backend (server) so you capture both pre‑ and post‑purchase events reliably.

UTM and link recipes that won’t break

Use a consistent scheme so every click, deep link, and callback carries context into analytics, CRM, and billing.

utm_source={alexa|whatsapp|in_car}
utm_medium={voice|chat|auto}
utm_campaign={intent}-{q1_2026}
utm_content={prompt|slot|reply}-{variant}

• Alexa+: Link to canonical PDPs/collections with UTMs. Use short links if your assistant requires spoken URLs, then 302 to the canonical URL with UTMs preserved.
• WhatsApp: For “Send catalog” or “Track order” replies, append UTMs to the links your bot returns. Also include a wa_session_id param so server events can join the journey.
• In‑Car: Prefer app deep links if your app is installed; otherwise send users to mobile‑optimized PDPs or quick‑checkout links with preserved UTMs.

Shopify and WooCommerce: quick implementation patterns

Shopify

• Add UTM‑aware cart.js or Cart Transform to store utm_* and assistant_channel into cart attributes and then order attributes.
• Fire purchase server‑side from your backend using the order webhook, enriching with assistant_channel, session_id, and intent.
• Create a “Voice & Chat” collection to use as a single landing hub for assistant traffic.

WooCommerce

• Capture UTMs into order meta at checkout; mirror to GA4 via server‑side Measurement Protocol.
• Use query‑string controlled PDP blocks (e.g., ?assistant=reorder) to show smaller, faster layouts for in‑car sessions.

Server‑side events: the glue for reliable attribution

Client‑only tracking drops when users switch apps or networks. Always emit server‑side events on:

• Order created / paid (source of truth for revenue).
• Refunds (so ROAS and payback are honest).
• Assistant milestones (e.g., WhatsApp reply, Alexa prompt → link sent) joined via session_id.

Keep PII out of analytics payloads; use stable, hashed IDs to join across systems.

Phone calls and handoffs: don’t lose attribution

• Wrap tel: links with UTMs (e.g., tel:+1-212-555-0100?src=alexa&intent=book) and capture as an event before dialing.
• If you use dynamic numbers for call tracking, carry the assistant_channel into the call session so closed‑won orders attribute correctly.
• Log handoff reasons to reduce future friction (payment, inventory, policy, etc.).

Privacy and compliance guardrails

• Only store what you need for attribution; keep content of conversations out of analytics.
• Provide opt‑out paths for marketing measurement and respect user locale/country settings.
• Encrypt assistant session IDs at rest; rotate keys on a set schedule.

Your 7‑Day instrumentation sprint

1. Day 1: Define your UTM schema and event names. Create a “Voice & Chat” collection or landing page.
2. Day 2: Capture UTMs into cart/order attributes (Shopify) or order meta (WooCommerce). Add server‑side purchase events.
3. Day 3: Implement assistant milestones (*_session_start, *_cta_click, *_purchase). Add handoff_to_human.
4. Day 4: Build a GA4 Exploration (funnel + pathing) and a Looker Studio overview (channel card, revenue, AOV, CPA).
5. Day 5: Create quick‑checkout links for top 10 SKUs with UTMs and test via Alexa+/WhatsApp flows.
6. Day 6: Validate refunds and cancellations post back to analytics. Sanity‑check payback math.
7. Day 7: Set alerts: CR drop > 30%, handoff rate > 15%, AOV delta > 20% vs web baseline.

Dashboards that answer revenue questions fast

Build two views and share them with finance and support:

1) Executive Overview

• Revenue, Orders, AOV by channel (Alexa+, WhatsApp, In‑Car) for the last 7/30 days.
• CPA and Payback by channel.
• Top intents by revenue (reorder, book, track → upsell).

2) Journey Explorer

• Paths: assistant → PDP → cart → checkout → purchase.
• Drop‑offs by device context (voice, chat, auto) with error logs.
• Handoff reasons and their impact on conversion and CSAT.

Common pitfalls (and easy fixes)

• Broken UTMs via short links: Always 302 to a canonical URL with UTMs preserved. Test with and without trailing slashes.
• Assistant traffic mixed with owned email/SMS: Reserve distinct utm_medium values like voice, chat, and auto.
• No refund events: Your ROAS is fantasy without assistant_refund. Instrument it.
• Only client‑side tracking: Add server‑side orders and refunds; they’re your source of truth.

Related playbooks to ship next

• Become voice‑bookable on Alexa+ in 72 hours
• Assistant SEO for 2026
• Build a compliant WhatsApp support bot
• CES 2026: What changed and your 7‑day plan

Wrap‑up

Assistant channels move fast, but measurement doesn’t have to lag. With a stable taxonomy, resilient UTMs, and server‑side events, you can trust your numbers and invest where revenue is real.

Try HireNinja: Ship assistant analytics the easy way