Best Prompts to Segment Shoppers by Intent Using Event Data Only (2025)

If you want fast, reliable, privacy-safe intent segments, you can build them using only first-party event (clickstream) data—no PII, no third-party enrichment. This guide gives you copy-paste prompt templates you can run or adapt in GA4 Audiences, BigQuery (GA4 export), Amplitude, Mixpanel, and Segment (Twilio) Engage.

• Time: 60–120 minutes (to build all core cohorts across tools)
• Difficulty: Intermediate (analytics and basic SQL familiarity)
• Output: Runnable prompt templates + a repeatable QA and troubleshooting routine

What “event-only” means here: we rely exclusively on behavioral events and parameters like product views, add_to_cart, begin_checkout, purchase, and item-level attributes—aligned to the GA4 ecommerce schema and item-scoped parameters documented in Google’s official references (see the GA4 ecommerce events list and parameters in Google’s Analytics Help and Developers docs).

According to Google’s current GA4 guidance, standard ecommerce events include view_item, add_to_cart, begin_checkout, and purchase, with critical item-level fields like item_id, item_name, and price captured in the items[] array, which can be registered as item-scoped custom dimensions for analysis and audience building when needed, as covered in the Google Developers’ pages on item-scoped ecommerce and custom definitions (2024–2025) (GA4 ecommerce events list; item‑scoped ecommerce parameters; create item-scoped custom dimensions).

Event-only signals cheat sheet (GA4-aligned)

Use these standard events and parameters as your raw materials. They are defined in GA4’s recommended events and parameters documentation (2024–2025) (recommended events reference; event parameters overview).

• Page/product exploration
  • view_item, view_item_list, select_item; item_id, item_name, item_category, item_brand
• Cart progression
  • add_to_cart, remove_from_cart, view_cart
• Checkout progression
  • begin_checkout, add_shipping_info, add_payment_info
• Purchase
  • purchase with transaction_id, value, currency, items[] (price, quantity, etc.)
• Price sensitivity (if tracked)
  • coupon at event level or item.coupon in items[]; price filters as custom events/properties
• Recency/frequency
  • Count of events in last X days; last purchase age

Implementation and QA: Validate GA4 ecommerce events and item arrays in DebugView/Realtime, ensuring required parameters like transaction_id (for purchase) and currency when value is set, per Google’s validation guidance (2024–2025) (validate ecommerce; verify implementation).

Intent tiers and decision rules (vendor-neutral)

Tune thresholds to your traffic, but start with these definitions:

• Browsers (low intent)
  • ≥2 view_item in last 7 days; exclude add_to_cart and purchase in last 7 days
• Considerers (medium intent)
  • view_item then view_cart OR add_to_wishlist within same session (or within 24h if session constraint not available)
• High intent
  • add_to_cart ≥1 and begin_checkout within the time window; exclude purchase
• Cart abandoners
  • add_to_cart OR begin_checkout; no purchase within the time window
• Price-sensitive
  • coupon used or price sort/filter interaction ≥2; no purchase yet (requires custom events/properties for price filters)
• Repeat buyers
  • purchases ≥2 in last 180 days; most recent purchase in last 60 days
• Churn risk (buyers)
  • last purchase ≥180 days ago; zero sessions/visits in last 30 days

Note on platform limits: GA4 Audiences use day-based windows; minute-level or “same session” sequences are not explicitly documented there, so use BigQuery or tools that support sequence windows if you need higher precision. Google’s Audience metrics/time windows Help article (2024–2025) explains state-based and time-windowed evaluation (GA4 audiences/time‑windowed metrics).

Prompt templates by tool

Use the templates as-is or swap placeholders like {time_window} and {min_events}.

A) GA4 Audience Builder (natural-language rule blocks)

These map directly to Audience conditions. Exact UI labels can vary; consult Google’s Audience and Ads-linked overview for current terminology (GA4 audience basics).

1. Cart Abandoners (7 days)

• Include: Users who performed add_to_cart OR begin_checkout in the last 7 days
• Exclude: Users who performed purchase in the last 7 days
• Membership duration: 7–14 days (pick based on activation need)

2. High Intent (24 hours)

• Include: Users who performed add_to_cart AND begin_checkout in the last 1 day
• Exclude: Users who performed purchase in the last 1 day
• Membership duration: 1–3 days

3. Browsers (7 days)

• Include: Users with view_item count ≥ 2 in the last 7 days
• Exclude: Users who performed add_to_cart OR purchase in the last 7 days

Tip: If your price sensitivity logic relies on coupons, GA4 supports coupon as an event parameter and in items[]; see Google’s parameter reference (2024–2025) (event parameters overview).

B) BigQuery SQL (GA4 export)

Use this when you need precise sequences or minute-level windows. Google’s GA4 BigQuery export schema and query patterns are documented in the Developers site (2024–2025) (BigQuery basic queries for GA4).

1. High Intent within 30 minutes (add_to_cart → begin_checkout)

-- Replace project.dataset and dates as needed
WITH atc AS (
  SELECT user_pseudo_id, event_timestamp AS ts
  FROM `project.dataset.events_*`
  WHERE event_name = 'add_to_cart'
    AND _TABLE_SUFFIX BETWEEN FORMAT_DATE('%Y%m%d', DATE_SUB(CURRENT_DATE(), INTERVAL 7 DAY))
                          AND FORMAT_DATE('%Y%m%d', CURRENT_DATE())
), bc AS (
  SELECT user_pseudo_id, event_timestamp AS ts
  FROM `project.dataset.events_*`
  WHERE event_name = 'begin_checkout'
    AND _TABLE_SUFFIX BETWEEN FORMAT_DATE('%Y%m%d', DATE_SUB(CURRENT_DATE(), INTERVAL 7 DAY))
                          AND FORMAT_DATE('%Y%m%d', CURRENT_DATE())
)
SELECT DISTINCT a.user_pseudo_id
FROM atc a
JOIN bc b USING (user_pseudo_id)
WHERE b.ts BETWEEN a.ts AND a.ts + 30*60*1000000; -- microseconds

2. Cart Abandoners (7 days)

WITH atc AS (
  SELECT DISTINCT user_pseudo_id
  FROM `project.dataset.events_*`
  WHERE event_name IN ('add_to_cart','begin_checkout')
    AND _TABLE_SUFFIX BETWEEN FORMAT_DATE('%Y%m%d', DATE_SUB(CURRENT_DATE(), INTERVAL 7 DAY))
                          AND FORMAT_DATE('%Y%m%d', CURRENT_DATE())
), pur AS (
  SELECT DISTINCT user_pseudo_id
  FROM `project.dataset.events_*`
  WHERE event_name = 'purchase'
    AND _TABLE_SUFFIX BETWEEN FORMAT_DATE('%Y%m%d', DATE_SUB(CURRENT_DATE(), INTERVAL 7 DAY))
                          AND FORMAT_DATE('%Y%m%d', CURRENT_DATE())
)
SELECT user_pseudo_id
FROM atc
WHERE user_pseudo_id NOT IN (SELECT user_pseudo_id FROM pur);

Identity note: GA4 exports use user_pseudo_id (device/browser-scoped) and optional user_id if you set it. Without a stitching strategy, cohorts are device-scoped. See Google’s User-ID guidance (2024–2025) (GA4 User‑ID).

C) Amplitude behavioral cohorts (event-only)

Define cohorts by behavior and properties; keep session/minute constraints general to avoid overfitting to changing UI terms. For behavioral segmentation patterns and examples from Amplitude’s team, see their product blog (2023–2025) (behavioral segmentation examples).

1. Cart Abandoners (7 days)

• Performed event: Add to Cart at least 1 time within last 7 days
• Exclude: Performed event Purchase within last 7 days
• Optional: Sequence — Add to Cart then Begin Checkout within 1 day (if your plan supports sequence windows)

2. Price‑Sensitive Browsers (14 days)

• Performed event: Product Viewed at least 2 times within last 14 days
• And event/property: Price Sort = ascending OR Applied Coupon = true (use your actual property names)
• Exclude: Purchase within last 14 days

Identity note (conceptual): Anonymous users are tracked by device-scoped IDs and can later merge with authenticated IDs after login. Treat pre-login segments as device-scoped unless you implement ID merge.

D) Mixpanel cohorts (sequences and windows)

Mixpanel supports event sequences, time windows, and exclusions; related capabilities are covered across Mixpanel’s official docs for funnels/time windows and identity management (2024–2025) (Funnels advanced: conversion windows and filters; Identity management overview).

1. High Intent (24h, sequence)

• Did event: Add to Cart
• Then did event: Begin Checkout within 24 hours
• Exclude: Did event Purchase within 24 hours
• Cohort type: Rolling, last 1 day

2. Repeat Buyers Likely to Churn

• Did event: Purchase at least 2 times in last 180 days
• And did NOT do event: Session Start (or Site Visit) in last 30 days

Identity note: Mixpanel’s simplified ID merge unifies $user_id and $device_id into a single distinct_id, enabling cross-device continuity (2024–2025) (ID management and simplified merge).

E) Segment (Twilio) Engage audiences

Segment’s Audiences and Journeys let you define event-only lists and activate them to destinations with continuous sync, as described in their official documentation (2024–2025) (Audiences overview; Journeys logic).

1. Cart Abandoners (7 days)

• Who performed event: Add to Cart within last 7 days
• And NOT who performed event: Purchase within last 7 days
• Activate: Facebook Custom Audiences; enable ongoing sync (Facebook Custom Audiences destination)

2. Considerers (7 days)

• Who performed: View Item then View Cart within last 7 days
• If the simple Audience UI lacks sequence constraints, implement as a Journey with ordered steps and a 7‑day window

Build and validate: a quick playbook

Preflight (10–20 minutes)

• Confirm ecommerce events exist and are spelled per GA4 recommended names; custom naming (e.g., add_to_basket) won’t populate standard ecommerce reports without mapping. See Google’s recommended events page (2024–2025) (GA4 recommended events).
• Check that purchase sends transaction_id and items[] with item_id, item_name, price, quantity, and currency when value is set; verify in DebugView/Realtime (validate ecommerce).
• Ensure price filters or coupon usage exist as events/properties if you plan a price-sensitive cohort.

During build (20–40 minutes)

• Match windows to tool limits: GA4 Audiences are day-based; for minute-level sequences, use BigQuery, Mixpanel, or an Amplitude plan that supports sequence windows. Google’s audience time-window behavior is described in their 2024–2025 Help article (GA4 audiences/time‑windowed metrics).
• Apply exclusions carefully: remove recent purchasers from abandoners/high intent; keep cohorts mutually exclusive where it matters.
• Identity scope: communicate that event-only segments without cross-device stitching are device/anonymous scoped in GA4 exports; Mixpanel/Amplitude can merge IDs after login per their identity docs (2024–2025) (GA4 User‑ID; Mixpanel identity overview).

Post-build QA (20–30 minutes)

• Sanity check sizes: Abandoners should be a subset of add_to_cart users—if it’s near 100%, your purchase exclusion is misconfigured.
• Funnel consistency: Abandoners should over-index on prior add_to_cart relative to site average; validate with a quick funnel report (Mixpanel’s funnels docs illustrate conversion windows and inline filters, 2024–2025) (Funnels advanced).
• Back-test outcomes: Compare 7‑day conversion for “High Intent” vs site baseline; you should see a lift even without personalization.
• Overlap analysis: Ensure cohorts align with strategy (e.g., High Intent and Abandoners should be mutually exclusive).
• Monitor drift: Track weekly counts and conversion rates; investigate sudden level shifts (often instrumentation or destination sync issues).

Troubleshooting guide (fast fixes)

• Missing or misnamed events

  • Fix naming to match GA4 recommended events so your analytics and audiences behave consistently (2024–2025) (GA4 recommended events).

• Purchase records ignored or deduped

  • Ensure transaction_id is unique and items[] present; set currency with value per GA4 rules. Validate in DebugView and test with varied transaction_id to avoid deduplication (validate ecommerce).

• Identity fragmentation

  • Without user_id (GA4) or ID merge (Mixpanel/Amplitude), you’re device-scoped. Align expectations and consider earlier login prompts or post-login activations (GA4 User‑ID; Mixpanel ID management).

• Sessionization and sequence precision

  • GA4 Audiences don’t document same-session or minute-level sequence constraints; for precise sequences, prefer BigQuery SQL or tools with sequence windows (see GA4 BigQuery reference 2024–2025) (BigQuery basic queries for GA4).

• Event inflation/duplicates

  • Debounce client events; consider server-side purchase confirmation; watch for repeated add_to_cart event fires on page reloads.

Toolbox: Platforms that support event-only intent segmentation

Use any of these to implement, analyze, and activate event-only cohorts:

• Google Analytics 4 + BigQuery (native export for precise SQL) — see Google’s BigQuery export docs (2024–2025) (BigQuery basic queries for GA4).
• Amplitude (behavioral cohorts and activation; see examples and guidance on their blog, 2023–2025) (behavioral segmentation examples).
• Mixpanel (cohorts, sequences, funnels windows, identity merge) — covered in Mixpanel’s official docs (2024–2025) (Funnels advanced; ID management).
• Segment (Twilio) Engage (audiences and journeys with destination sync) — see Segment’s Audiences and Journeys documentation (2024–2025) (Audiences overview; Journeys logic).
• WarpDriven — AI-first ERP and event-driven decisioning platform suitable for operationalizing event-only segments across commerce workflows. Disclosure: WarpDriven is our product.

We don’t rank these; choose based on your stack, needed precision (sequence windows), and activation endpoints.

Quick-reference: copy-and-adapt prompts

You can paste these into your tool builders (or use them as checklists):

• GA4 Audience — “Cart Abandoners (7d)”

  • Include: add_to_cart OR begin_checkout in last 7 days
  • Exclude: purchase in last 7 days

• GA4 Audience — “High Intent (24h)”

  • Include: add_to_cart AND begin_checkout in last 1 day
  • Exclude: purchase in last 1 day

• BigQuery SQL — “High Intent (30m sequence)”

  • add_to_cart then begin_checkout within 30 minutes; exclude purchasers (add a NOT EXISTS purchase clause if you need a pure pre‑purchase list)

• Amplitude Cohort — “Price‑Sensitive Browsers (14d)”

  • Viewed Product ≥2 times; Price Sort = asc OR Applied Coupon = true; exclude Purchase (all within last 14 days)

• Mixpanel Cohort — “Repeat Buyers Likely to Churn”

  • Purchase ≥2 in last 180 days AND no Session Start in last 30 days

• Segment Engage Audience — “Considerers (7d)”

  • View Item then View Cart within last 7 days (use Journeys for ordered steps if needed)

What good looks like (success criteria)

• Segment correctness: Random user spot checks show events match definitions.
• Predictive signal: “High Intent” cohort converts materially higher in 7 days than baseline.
• Operational health: Audience sizes and destination syncs update daily; alerts fire on failures.
• Maintenance: Weekly drift review; re-validate after instrumentation changes.

For deeper references on GA4 ecommerce events/parameters and audience/time-window behavior, use Google’s current 2024–2025 documentation: the ecommerce events list and item-scoped parameters, plus audience metrics/time windows (GA4 ecommerce events; item‑scoped ecommerce; audiences/time windows). For behavioral cohorting and identity stitching concepts in product analytics, see Mixpanel’s and Amplitude’s official materials (2023–2025) (Mixpanel identity overview; Amplitude behavioral segmentation examples).

Next steps

• Start with High Intent and Abandoners cohorts; validate conversions in 7 days.
• Add Price‑Sensitive and Considerers; test personalized messages or offers.
• Productionize in your CDP/ESP with rolling updates and guardrails.

Want a unified way to operationalize these segments across product, marketing, and supply chain workflows? See how WarpDriven can help translate event-only intent cohorts into automated actions across your stack: WarpDriven.