Funnel

cdp_funnel

Computes your brand's browse-to-purchase conversion funnel over a date range. Each stage's count is the number of distinct customers who performed that stage in the window, and conversion is the share of the previous stage's customers who also performed this stage — a cohort intersection, clamped to 0–100%, so the funnel is monotone-safe rather than a raw event-volume ratio. When the distinct-customer cardinality backend is unavailable it falls back to labeled per-stage event counts (basis: "events"). The stages, in order, are:

product_viewed → product_added_to_cart → checkout_started → checkout_completed / order_created

The final stage counts checkout_completed or order_created, whichever your store emits.

Read-only and scoped to your brand.

Parameters

Returns

• stages — an array in funnel order. Each stage has stage, eventTypes, count (the cumulative funnel-cohort figure: distinct customers on the unique_customers basis, or raw event volume on the events fallback), activityCount (the independent per-stage distinct-customer count via any entry path — equals count on the events basis), conversionFromPreviousPct, conversionFromStartPct, and dropOffFromPreviousPct.
• dateRange — { from, to }, the window you passed.
• scope — your brand's shop domain.
• basis — "unique_customers" when counts are distinct-customer cohort estimates, or "events" when the tool fell back to raw per-stage event counts.
• note — a human-readable explanation of the semantics for the chosen basis.

Example prompts

What's my browse-to-purchase funnel for last month?

Show me conversion from product view to checkout for June

Notes

• The brand scope is injected server-side, so counts only ever cover your own events.
• Percentages are computed per stage relative to both the previous stage and the funnel start; the first stage's ratios (and dropOffFromPreviousPct) are null. On the unique_customers basis they are cohort-intersection ratios (clamped to 0–100%); on the events fallback they are event-volume ratios and not directly comparable across stages.
• Aggregate queries are throttled — on a rate limit, the tool returns a clear rate-limit error.