Analytics

Insights → Analytics is a dashboard builder over two kinds of data: the recommendations NSL served, and the events your users generated. Use it to see how each placement performs, which items and events drive engagement, and how people move through your product.

​

Dashboards

A dashboard is a saved arrangement of widgets with its own filters.

• New from template creates a dashboard from a starting point that you then own and can rearrange: Overview, Events, Items, Users, Funnels, User journeys, Comparisons or Experiments. Or start empty.
• Save as saves the current view under a new name.
• In Dashboard settings, rename a dashboard or Set as team default, so it loads first for everyone in the workspace.
• Share copies a link to the dashboard with its current filters.
• Refresh reloads every widget.

​

Widgets

Select Add widget, choose a type, and pick what it measures. Drag a widget’s corner to resize it.

​

Metrics

Served requests, served items, served users, events, weighted event value (events multiplied by their weights), event users, event items, sessions and engagement rate, plus a count and a rate for each of your event types.

​

Dimensions

Event type, context, item, user, session, request, placement, model version and pipeline version. Rare values are grouped as Other.

​

Filters

• Time range: 24 hours, 7, 30 or 90 days, 6 or 12 months, year to date, this or last week, month or year, or a custom range.
• Context: one placement, or all contexts.
• Data: Auto shows sample data until real data exists, then real data. Choose Real data or Sample data to force one.
• Drill-down: narrow from a user, to one of their sessions, to a single request. Any combination applies at once, so you can isolate one journey.
• Event IDs: limit event metrics to particular event types.

​

Reading the results

• Many served requests, few clicks: the placement is serving but not resonating. Review its recipe and rules, and check a few results in Explainability.
• A sudden change in one event type: compare with releases or campaigns from the same day, and use the session filter to rule out test traffic.
• A drop after a model change: break served requests and events down by model version to compare the old and new versions directly.
• Comparing variants: the Experiments template breaks results down by pipeline and model version. For proper A/B tests with lift, use Experiments.

​

Retention

How far back analytics can query depends on your plan and retention add-on (90, 180 or 365 days, or custom on Enterprise). A label beside the page title shows the retention that applies.