A table below the breakdowns comparing your traffic sources on four signals at once, so you can see which channels send visitors who stay.
The channel quality table with several channels compared, one clearly stronger on engagement despite lower volume. Light theme, cropped to the card.
The four signals
| Signal | What it measures |
| Bounce rate | Share of visits that saw one page and left |
| Pages per visit | Average pages in a visit |
| Session time | Median visit length, ignoring visits under five seconds |
| Returning share | Share of visits from someone who had been before |
Returning share is derived from visits against distinct visitors: ten visits from seven people means three of them came back.
Why there is no single score
The card says it itself — the weighting would be arbitrary. There is no defensible answer to "how many bounce points is a returning visitor worth", so any 0–100 score would be one opinion presented as arithmetic.
It would also hide the case the card exists for. A channel can send few visitors who read everything, and another can send many who bounce. Those are different problems with different answers — one is a scale problem, the other a targeting problem — and a single rank tells you nothing about which one you have.
Read it against volume, not instead of it
A channel with five visits and a perfect bounce rate has told you nothing yet. The table shows visits alongside the signals for exactly this reason.
The useful comparison is usually between two channels that both send real volume, where one is clearly holding attention better. That is a reason to move budget; a tiny channel with flattering numbers is not.
Returning share needs the cookie mode
Under either cookieless identity mode a visitor cannot be recognised across days, so that column stays empty rather than showing a number that is wrong by construction.
See Cookieless mode for the trade-off, and How visitors are counted for what each mode can and cannot answer.
Related: When people visit.