Analytics tabs

The Analytics screen splits your data into six tabs. This page walks through each tab and every number on it.

The screen is Analytics in the PersonalizerAI dashboard. The tabs are Revenue, Recommendations, Search, Products, Visitor Traits, and Session Replay. The last two need the Growth plan or higher. The numbers refresh once per night, at 02:00 UTC — today's activity appears tomorrow.

Pick a period

The Last selector offers 7d, 30d, and 90d. For any other range:

1. Click Custom.
2. Select a start day in the calendar.
3. Select an end day.
4. Click Apply.

Every table has a CSV button. It downloads that table for the selected period. On Session Replay, it downloads every attributed order in the period, not only the page on screen.

Revenue

Three cards sit at the top:

• Attributed Revenue — revenue from orders that PersonalizerAI helped cause.
• Attributed Orders — how many of those orders there were.
• Attributed AOV — the average value of those orders: revenue divided by orders.

Each card compares against the prior period of the same length.

The Attribution Funnel chart shows the path from view to order: Impressions → Clicks → Add to cart → Orders. Each step shows its count and its share of the first step, and the line below shows the rate from each step to the next. How attribution works, under the chart, explains which orders count — see How we measure revenue. The Daily Revenue chart shows one point per day. Point at a day to see its revenue and orders. When no order in the period has attributed revenue, the card says "No attributed revenue in this period." instead.

Below it, the What bundles add card shows the same results as the Results tab of Bundles. It is measured over its own window, so the period you pick does not change it. See What bundles add.

The Revenue tab: cards, the funnel, and the daily chart.

Recommendations

Two tables. The first is Performance by Model, sorted by revenue. It shows the same models as the dashboard, with Clicks and Orders in place of Type and Revenue per 1,000 impressions:

The second table is Performance by Placement: one row per placement and model, with Placement, Model, Impressions, Clicks, CTR, Add to cart, Orders and Revenue. Its revenue comes from widget clicks only. Same-visit revenue has no placement, so this table leaves it out. The checkout row shows orders and revenue only. Views and clicks of your checkout upsells are on the Checkout page.

To see one model on its own, pick it in the Model menu above the tables (All models shows them all). The tab then shows a card with that model's Impressions, Clicks, CTR, Add to cart, Orders and Revenue for the period — the same numbers as the dashboard and the model's card on Recommendations — and only its rows in the placement table. The models' add-to-carts can add up to slightly less than the Revenue tab's funnel: an add whose cart line did not say which model or placement it came from counts in the funnel but under no model. On Recommendations → Models, a model card's View detailed analytics link opens this view for that model.

Search

Four cards, from top to bottom:

• A summary row: Searches, Zero-result rate, Search conversion (searches that led to a purchase), and Avg position clicked (1 is the first result; lower is better).
• Searches per day — a chart.
• Top Queries — the most frequent searches, with Query, Searches, and Clicks. A link below it goes to Popular searches on the Search experience tab, where you choose which searches the empty search box suggests.
• Zero-Result Queries — searches that returned nothing, with Query and Zero-result searches (how many times the query found nothing). A red badge shows how many queries there are.

Zero-result queries are demand you miss. Each row has an Add synonym and an Add redirect button. They open the Search screen with that query already filled in. See synonyms & redirects.

When nobody searched in the period, the tab shows one message instead of the cards.

Products

The Top Products table shows up to 50 products with the most revenue from recommendations. Products with the same revenue are ordered by clicks, then impressions, from recommendations and search together.

The Underperforming Products card lists up to 20 products with one of these problems (it says so when there are none). Both rules count recommendation impressions only, and the card's columns are Rec Impr., Rec Clicks, Rec CTR and Issue:

• High impressions, no attributed sales — shown more than 500 times in recommendations, with no attributed revenue from a recommendation or search. The product may still sell in other ways.
• Low CTR — shown more than 200 times in recommendations, with a Rec CTR under 0.5%.

For these products, check the image, the title, and the price. Or use Create a rule on the row: it opens Merchandising with a new merchandising rule that already targets the product, so you only choose what the rule does (for example bury or exclude) and name it.

The Products tab: top products above, detected problems below.

Visitor Traits (Growth plan and up)

This tab groups your anonymous shoppers into behavior cohorts. It is aggregate only — there are no per-shopper profiles, no names, and no e-mail addresses.

A shopper's browser counts as qualified after 3 or more sessions in the last 90 days. Only qualified shoppers are scored. The tab discloses both counts: "Scored over N qualified visitors (3+ sessions) of M active in the last 90 days."

The Visitor Trait Cohorts table shows four traits, with Visitors and Share of qualified:

One shopper can hold several traits, so the shares do not sum to 100%. The Weekly trait share chart (titled with the period, for example "Weekly trait share, last 30 days") shows how each share moves week by week.

The table always uses a rolling 90-day window: it shows each shopper's latest score, whatever period you pick. The period only sets how far back the weekly chart goes.

Visitor Traits: cohort shares over qualified shoppers, and their weekly trend.

Session Replay (Growth plan and up)

Four cards summarize buyer behavior and click activity:

• Avg Touchpoints — the average number of widget and search clicks in a buying session.
• Avg Time to Purchase — the average time from the first event to the purchase, in minutes.
• Cross-Widget Attribution — the share of buying sessions that clicked more than one model.
• Most-clicked model — the model with the most clicks in the period.

The Attributed Orders table lists every attributed order:

The table shows 20 orders per page. Its CSV button downloads every attributed order in the period, not just the page you see — the line above the table says how many. One file holds up to 10,000 orders, the most recent first; for more, pick a shorter period. The file has no Touchpoints column, and the Sources column uses the internal model names.

To open a replay:

1. Open the Session Replay tab.
2. Click an order row in Attributed Orders.
3. Read the timeline in the session card above the table.

The timeline lists each event with its time — page visits, widget views, clicks, searches, cart adds, and the purchase. A zoom control spreads the timeline out. When session data is not available yet, the four cards show a dash and a message — the orders table still works.

Session Replay: the timeline of one buying session, event by event.