Skip to main content

click_event — one row per click

click_event records every click a visitor makes, tied back to the page view it happened on. One row = one click = one (pv_id, moment, target) triple.

Use this material when the question is about interaction, not just traffic: click-through rates, which buttons actually get pressed, where people rage-click, what anchor text is working. If the question can be answered without looking at individual clicks, use allpv instead — click-event queries are more expensive.

Sample rows​

Hand-crafted sample. Only pv_id, session_id, page_id shown as IDs; the full schema in ai/materials.yaml carries more (e.g. event_sec, selector, element_text, element_id, element_class, page_x_pct, page_y_pct).

pv_idsession_idpage_idevent_secelement_textselectorto_urlis_external
9102348217104214"See our picks"a.cta-main/products/hario-slim/—
910244821710886"Add to cart"button#add-to-cart(no href)—
910244821710887"Add to cart"button#add-to-cart(no href)—
910244821710888"Add to cart"button#add-to-cart(no href)—
9102448217108811(image of product)img.product-hero(no href)—
91027482191201142"Buy on Amazon"a.affiliate-linkhttps://amazon.co.jp/...✓
910284822010884"Add to cart"button#add-to-cart(no href)—

Things the sample is trying to show at a glance:

  • pv_id 91024 has four rapid clicks on the same button over 5 seconds: three on button#add-to-cart, one dead-click on the product image. That is a rage-click pattern — the button did not respond and the visitor started stabbing at anything clickable. This is the kind of row that should surface in a "pages with frustrated visitors" query.
  • pv_id 91027 — one click, 142 seconds after page load, on an affiliate link. That is an earned click: the visitor actually read the content before converting. Completely different story from the rage-click session.
  • pv_id 91028 is another visitor hitting the same "Add to cart" button, also within 4 seconds. Paired with the irritation_click_count = 4 on that same pv_id in the allpv sample, it confirms the button has a real problem.

Columns the sample glosses over​

  • Event timing: event_sec is the seconds from page load to click — useful for "did they read before clicking" analysis.
  • Element position: page_x_pct, page_y_pct are the click coordinates as a percentage of the page, normalized across viewports. Useful for heatmap-style queries.
  • Element metadata: element_id, element_class, element_data carry the attributes of the clicked element for finer-grained rollups than selector alone.
  • Action classification: action_id is the internal action type for the click (navigation, form submit, outbound, …).

See ai/materials.yaml for the full list.

Common JOIN keys​

  • pv_id → allpv.pv_id (pull in the page-level context for a click — device, referrer, browse_sec, etc.)
  • session_id → allpv.session_id (session-scoped aggregations)
  • page_id → page_version.page_id (what version of the page was live when the click happened)

Good first questions for click_event​

  • "Which buttons on the product page have the highest click-through rate?"
  • "Where on the page do people click most — above or below the fold?"
  • "Which pages have the most rage clicks?"
  • "What is the average time-to-first-click on the hero CTA?"
  • "Which outbound links are actually driving traffic?"

Where to go next​