Skip to main content
Every collection carries its own analytics, surfaced in three places: KPI tiles above the collections table, performance columns in the table itself, and a detailed analytics drawer per collection.

Limits and behavior

  • Metrics cover the last 30 days compared with the previous 30 days by default; the presets are 30, 7 and 1 days, plus a custom range. The drawer has its own period picker.
  • Add-to-cart facts appear only when add-to-cart events are tracked; they keep their product-performance basis.
  • Collection engagement uses Recorded collection visitors, Recorded clicking visitors and Visitor CTR for the selected markets and period. Revenue, Units and STR retain their existing definitions. There is no inventory column and no pinned-vs-auto column in the collections table.
  • Recorded collection visitors and clickers describe devices observed on that collection’s own page. Earlier product-based figures added activity on the collection’s current products, including activity elsewhere. The definitions are not directly comparable, and lower numbers alone do not show lost traffic.
  • Missing or withheld engagement shows —. A Tracking unreliable warning withholds Visitor CTR throughout the selected scope; otherwise eligible visitor and clicker counts remain. See Fix collection-page tracking.
  • Dead stock counts products with inventory but zero units sold in the selected period.
  • Average stock coverage is the number of days current inventory lasts at the current sales rate; it appears in the analytics drawer, not in the collections table.
  • Pinned vs Auto needs enough traffic on both your pins and the automatically ordered products before it gives a verdict.

KPI row

The tiles above the collections table retain their existing revenue and product-performance definitions:
  • Collection Revenue — revenue attributed to collection pages in the period.
  • Current product performance — current-product views and clicks, with add-to-cart and order facts shown separately from collection engagement.
These product facts are not the recorded collection visitors or clicking visitors shown for an individual collection. Per-collection device counts are not added together and presented as unique merchant-wide visitors. Every metric label has a tooltip explaining its basis. The permanent Tracking unreliable explanation and repair link are also accessible by keyboard and touch. Clicking a tile opens a trend modal with the metric charted over time (30d/7d/1d or a custom range). Next to the scope controls, a dead stock button counts the products at risk (“3 items at dead stock risk”, or “No dead stock risk”). It opens a searchable list of those products with CSV export. The Benchmark button is in the page header.

The analytics drawer

Open the drawer from a collection row’s View analytics button.
Collection analytics drawer showing RPM, sales velocity, inventory donut chart and per-product pinned versus auto CTR comparison with recommendations

The analytics drawer: performance KPIs, inventory breakdown and the Pinned vs Auto comparison.

Performance

Current product performance keeps revenue, Product CTR and sell-through on their existing product basis, with previous-period deltas and trend charts where available. Show more adds RPM (revenue per thousand product impressions), sales velocity, inventory and variant availability — in-stock variants as a share of all variants in the collection. These values are separate from Visitor CTR and recorded collection visitors. A product metric is never substituted for an unavailable visitor metric.

Collection engagement

Recorded collection visitors, Recorded clicking visitors and Visitor CTR describe recorded devices on this collection’s own page. Visitor CTR is clicking visitors divided by collection visitors, when the row and the selected scope are eligible. Tracking quality and availability are checked separately for the current and previous periods; an unavailable previous value does not hide an eligible current count. Product impressions, add-to-cart events and sales retain their own bases. They are not connected to recorded collection visitors with cross-basis conversion rates. The click cut-off describes how far down the listing shoppers reach before 95% of recorded clicks are accounted for.

Pinned vs Auto Lift

How your pinned products compare with auto-ranked ones — see Pinned vs Auto.

Product Health

The collection’s Top 3 Sellers and Bottom 3 Sellers. Show more opens the full ranking.

Stock Health

The Stock Health view lists the collection’s low-stock variants and dead-stock risk in one place. Open it from the variant availability tile in the drawer’s Performance section (under Show more); the tile also says how many products are low on stock or at dead stock risk.

Pinned vs Auto

For collections with pinned products, Depict compares the click-through rate of your pinned products with what automatic ordering would get, and turns the result into a recommendation: The drawer and the Pinned vs Auto modal break it down per product, with the lift between the two CTRs, per-product recommendations, one-click pin/unpin actions and CSV export. You can filter the list to Unpin advised or Keep pinned. An info tooltip explains how to read the numbers.

Inventory and dead stock

Dead stock counts products with stock but effectively no sales in the period — capital sitting still. Average stock coverage estimates how many days current inventory lasts at the current sales rate. The dead stock button above the collections table lists the specific products across all collections so you can act on them; the Stock Health view does the same for one collection.

Benchmarking

The Benchmark button opens a modal where you compare any set of collections side by side — against each other and against the “All collections” baseline — across revenue, engagement and inventory metrics. Recorded engagement uses the same selected markets, period and availability rules as the collection table. Product snapshots keep their product basis; they are not compared with live visitor values as though the definitions were interchangeable.
Benchmark modal comparing two collections against the average of all collections across revenue, views, clicks, CTR and more

Benchmarking two collections against the all-collections average.

Fix collection-page tracking

Depict may withhold Visitor CTR when recorded collection visitors and clicking visitors are inconsistent for the selected markets and period. This warning describes the recorded data. It does not identify a single integration fault or prove that every unflagged row has complete tracking.

What the collection numbers describe

Recorded collection visitors count devices associated with requests for that collection’s products. Recorded clicking visitors count devices whose product-card clicks were attributed to that listing. A product-detail page view, a card impression and a collection visitor describe different activity; one does not substitute for another. Visitor CTR is recorded clicking visitors divided by recorded collection visitors. When at least 10% of the returned collection rows in the selected markets and period have more recorded clickers than viewers, Depict marks that scope Tracking unreliable and withholds Visitor CTR on every collection in it. Otherwise eligible visitor and clicker counts remain available with the warning. They are recorded observations, not proof of complete tracking. The threshold is a warning rule, not an acceptable tracking-error rate. An inconsistent row has its visitor count and Visitor CTR withheld even when fewer than 10% of rows are inconsistent. Independently recorded clickers can remain available. Missing or withheld portal values show —; they are not invented zeros or rates capped at 100%. The Export Most Popular CSV uses the same definitions and rules, with empty cells for missing or withheld values. A different period or market scope can have a different warning and availability.

Check the listing request and shopper identity

Ask the team maintaining your storefront to inspect a listing-products request in the browser’s network tools. Confirm that it addresses the intended collection and market and carries the shopper’s session identity. In the Platform API, the products request’s session_id is required for tracking and attribution even though it is optional for serving products. A successful products response alone does not establish attributable visitor coverage. Use the same session identity in serving requests, DPC and custom events. Do not replace all shoppers with one shared dummy identity. After client-side navigation, confirm that the requested listing matches the collection displayed. Follow your existing consent requirements; this guide does not require sending events without consent. See Listings and Which ID goes where.

Check clicks against the display that was rendered

For custom product cards, use the product_listing_result_id from that exact display as the card’s data-product-listing-result-id. Keep listing, search and recommendation result IDs separate. Do not reuse an old display’s result ID for a newly returned card. Legacy Depict UI SDK cards already carry these attributes. Check the integration before adding a second sender. The Product Card Tracking guide covers the attributes and custom event sending. The ID guide explains the ten-minute result-ID lifetime and event validation. Send clicks promptly with the result ID from the clicked display. If clicks arrive but listing visits do not, check the serving request’s session identity and collection/market mapping alongside the click integration. Adding only product-card impression events does not repair missing listing-request visitor records.

Verify the repair in the right period

After an integration change, compare the same collection, markets and a period covering the new activity. Check both collection visitors and clicking visitors. A recent repair does not automatically fill earlier missing records, so an older period can remain unreliable. In the portal, Settings → Tracking shows live integration diagnostics for the last 72 hours, including event arrival and product mapping. Its Product page view events table concerns product-detail pages. It is not proof that the collection visitor denominator is complete for your selected period. Likewise, an accepted event batch does not prove that every result ID was attributed. If the collection numbers remain inconsistent, contact Depict with the collection/page, selected markets, date range and approximate test time. Depict can check the aggregate listing and click paths. Do not send shopper identifiers, tokens or raw customer-event exports.

Collection engagement and current product performance

Collection engagement describes recorded activity on the collection page. Current product performance describes activity on the collection’s current products, including activity elsewhere. Those definitions differ, so a fall in collection engagement alone does not show that site traffic fell. Tracking gaps are a separate possible cause. Sales, units and product reports retain their own definitions. Product CTR is a separate product metric and is not Visitor CTR. Explaining the definitions or withholding an inconsistent metric does not itself repair storefront instrumentation.