Skip to main content

​
Preview with sort effect annotations

When you preview a sort order, the preview panel annotates each product with the sorting rules that affected its position. These annotations make it easy to understand why products appear where they do and to fine-tune your configuration. Each product can display:
  • Promoted / Demoted badges — color-coded indicators showing which priority rules matched the product and whether they pushed it up (green) or down (red).
  • Soft Boost badges — indicators showing which soft boost expressions matched, labeled with the concrete effect: positions moved (e.g. +3 positions) when the product’s rank changed, or the raw lift value when it didn’t. Hover a product to see a scoring card that breaks down the base value, position change, lift, and final score for each matched boost.
  • User Affinity badges — indicators showing when a product was lifted by a User Affinity expression because it matched the previewed shopper’s learned tastes, along with the computed lift.
  • Sequence badges — indicators showing when a product was placed by a product sequence. Hover for the tooltip Grouped by product sequence, which confirms the position came from a curated sequence group rather than a metric or rule.
  • Diversity badges — Diversity ↑ appears when a product was surfaced to broaden diversity coverage (tooltip: Surfaced for coverage). Diversity ↓ appears when a product was pushed down because its family already filled the diversity cap (tooltip: Capped by family limit).
  • Metric values — the actual value used for sorting when a metric or numeric attribute is part of the sort order. Date attributes are shown as human-readable dates, and values from segmented metrics are tagged Segmented (…). Hover to see the metric name.
  • Weighted score breakdown — when a weighted group drove the product’s score, the preview shows a stacked contribution bar and one row per feature. Each row lists the raw value, normalized score, weight share, and contribution (normalized × share = contribution), followed by a Weighted total. Each row’s color dot matches the feature’s slider tint in the editor. Layers normalizes scores against the current result set, so compare contributions within a preview rather than absolute values across sort orders. Features that are gated out or have no value for the product render as — with a 0 contribution.
Each badge includes the position number of the rule in your sort order (e.g., “Promoted by #1”), so you can trace every effect back to its source expression. Products that don’t match a given rule won’t show a badge for it. For step-by-step instructions, see Preview a sort order.

​
Preview as a shopper

When a sort order includes a User Affinity expression, results depend on the individual shopper viewing the collection. To preview this without waiting for live traffic, use the Preview profile shopper picker. Select or build a profile by simulating a shopper’s cart and purchase products. The preview then shows:
  • The collection reordered as that shopper would see it, with affinity matches lifted into place.
  • A plain-language summary of the shopper’s inferred preferences, so you can confirm the affinities make sense before publishing.
Switch between profiles to compare how the same collection looks for different kinds of shoppers, and verify your Affinity Weights and boost settings produce the experience you intend.

​
Preview by region or marketing segment

When a sort order uses segmented metrics or contextual conditions, the preview lets you pick a specific region (country, province, city) or marketing segment (channel, source, medium, campaign) to simulate as the visitor context. The collection then reorders as a visitor from that region or segment would see it. To keep previews meaningful, the region and marketing segment pickers only list values that have enough traffic to actually influence sorting. Long-tail values with only a handful of events are hidden — picking them wouldn’t change the sort order anyway, because segmented metrics fall back to global values when segment data is too sparse (see Smoothing factor). If a region or segment you expect to see isn’t in the list, it hasn’t yet reached the significance threshold. Wait for more traffic to accumulate, or preview a broader value (for example, the country rather than a specific city).

​
See also