Know Widgets Reference

This page has one row per widget on the Know dashboard, in the order they appear on screen, so you can scroll the dashboard and this page side by side. Every widget’s date range comes from the one picker in the Know header — with a single exception, Broken links, which is called out below.

The widget table

Fourteen cards, thirteen slots: Top growth and Top decline are drawn by one component into one grid row, so they share position 7.

Widgets 1 to 8 and 13 sit on /know alone. Widgets 9 to 12 are the only ones that go through the general metric API, which is why they are also the only ones that can return a plan gate.

Top insights

The insights strip runs two heuristics against the selected range and emits at most one card each. Both are computed server-side from real form fills and real bounces; neither is a model output, despite the “Insights powered by Leed” badge.

OPPORTUNITY finds the page whose conversion rate most beats your site average. Its headline reads “… converts 3.2x the average” and the stat in the corner is that multiple.

RISK finds the highest-traffic page that bounces hard and never converts. Its headline reads “… bounces 74% with zero leads” and the stat is the bounce percentage.

The Refresh control re-runs both against the same range. When neither clears its thresholds, the whole strip collapses to a single centered card carrying the badge Insights powered by Leed and the line “No insights to surface yet. As Leed learns from your content’s real performance, opportunities, risks, and ideas will appear here.”

One wart worth knowing before you go looking for a bug: an insight headline names the page by its page id, not its title. The card is otherwise correct, and the id is the one in the page’s editor URL.

The third insight card you will never see

The design for this strip had a third card type — unanswered reader search queries — and you will never see it. Leed keeps no log of searches that returned nothing, so the endpoint deliberately emits nothing rather than inventing a card. It is not a permissions problem, a plan gate or a data-volume problem, and no amount of traffic will produce one.

Journey funnel

Four steps, each labeled and drawn as a bar whose width is its share of the first step:

The funnel is computed for one content type only — whichever entry content type had the most sessions in the range. That is why the subtitle changes from the generic “Step-by-step path to conversion” to “Sessions entering on Blog” (or whatever your busiest entry type is) once there is data. There is no control to switch content types on this card.

The percentage beside each bar is its share of the first step, not of the step above it, so the numbers only ever fall. When there are not enough closed sessions to build a funnel at all, the card reads “Not enough session data for a funnel yet”.

Top growth and Top decline

Both cards come from the same ranking: every page’s conversion rate in the selected range against its conversion rate in the prior equal-length window, ranked by the size of the movement, top eight. The list is then split — pages that moved up go to Top growth, pages that moved down to Top decline — so the two cards rarely hold four rows each.

A movement of less than half a percentage point in either direction is treated as flat and appears in neither card.

Search gaps — permanently empty, and why

The Search gaps card renders its heading, its magnifying-glass icon, and the text “No search gaps detected”. It will do that on every plan, at every traffic level, for every date range, because the endpoint behind it returns an empty list and marks itself unavailable. There is no zero-result search log in Leed for it to read.

This is not the same thing as saying reader searches are not recorded. When live documentation search is enabled, every query your readers type is logged, including the ones that found nothing — this widget just does not read that log yet. It is listed among the known limitations.

Sessions by channel

The card fetches up to 50 attribution rows for the range, adds up their counts per utm_source, and draws the top five as labeled bars. A row with no source is bucketed as other. That client-side roll-up is why the card shows five channels while the underlying metric has six columns — the list metrics reference gives the full column set, and journeys, funnels and attribution explains what does and does not qualify as an attributed arrival.

Below Growth this card behaves the way a gated surface should: it stays on the page, the heading and the inbound attribution eyebrow still render, and the body is replaced by an upgrade banner reading “Full-journey attribution is available on the Growth plan”. The query is never sent, so nothing 402s in the background.

For what an upgrade banner means and who on your team is allowed to act on it, see when a feature is gated.

Top referrers

Referring hosts, ranked by sessions, top eight, each with a share bar. A session that arrived with no referrer — typed, bookmarked, or from an app that strips the header — is grouped as (direct).

This card is never plan-gated. Note that its heading is drawn as a small uppercase eyebrow rather than the larger card titles used elsewhere on the page; it is a card, not a section of the one above it.

There is a naming trap here that costs people time. The card headed Top referrers is backed by the catalog metric whose own name is “Referrers” — referrerHostMetric. There is a different metric in the catalog actually named “Top Referrers”, and this card does not use it. The pair is disentangled under the two referrer metrics.

Most viewed videos

Four columns — Video, Viewers, Watch time, Avg / viewer — ranked by unique viewers, top eight. The asset id is resolved to the video’s name, and clicking a row opens that asset. Watch time is deduplicated: a section a viewer rewatched counts once, so this figure is deliberately lower than raw seconds played.

Most viewed videos is one summarized view of your media. Per-video retention, pause and replay curves — including for audio, which has no Know widget at all — are covered in media engagement analytics.

Sessions by geography

Two choropleths side by side in one card: a US state map and a world country map, from sessionsByStateMetric and sessionsByCountryMetric. Both request up to 100 rows and both follow the header’s date range.

Regions with no sessions are not blank or missing — they are normalized to zero and painted as neutral land, so an empty map and a map with one visitor in one state look different. Location is approximated from network-level request data; Leed never asks a visitor’s browser for their position.

Broken links

The only widget on Know that ignores the date range. It shows the current state of your site — the results of the most recent external-link crawl — not a window of history, and its subtitle says so. Changing the picker to “Today” or “Previous 365 days” changes nothing in this table.

Four columns: Page, URL, Status, Message. The page cell links into the editor; the count of rows found is shown at the top right. When the last crawl found nothing, it reads “No broken links detected.”

It is also the only widget on the dashboard that does not use the analytics privilege — it reads company data instead. How the crawl runs, what statuses it reports and how to clear an entry are on link health.

The endpoints behind the page