The DML analysis can only report what a user wrote. For a sales manager who reads dashboards all day, or a support agent who mostly looks things up, that comes out as almost nothing — and a licence review based on it alone will quietly propose removing people who use the org constantly.
Visits reconstructs the other half: what people looked at. You will find it as a sub-tab of Licensing, next to DML.
For the same period the DML analysis covered:
Screens — which pages were opened, and how often
Paths — the sequences of screens people move through
Sessions — how long a visit lasts, and how often people come back
Per user — sessions, views, distinct screens, time on screen, and the screens each person spends their time on
Everything is filterable by the DML category each user already falls into, so you can ask the question that matters: of the people who write nothing, what are they actually doing?
Salesforce records UI activity in several different event types, and which ones your org produces depends on how it is built. A Classic or portal-heavy org emits one set; a Lightning-native org emits another.
Rather than assume, Visits measures. It samples what your org emits, works out how much of the population each event type can actually account for, and then streams only the types that cover a real share of your users. The coverage it measured is shown on every view, so you always know how much of the picture you are looking at.
The practical effect is that there is nothing to configure, and nothing to get wrong.
Coverage is the first thing to check. If Visits can only see a third of a population, treat it as a sample, not a census. The banner says so rather than leaving you to infer it.
"Cap reached" is about the list, not the data. Path lists are capped — the number of distinct screen sequences in a busy org grows enormously, and the display keeps the most travelled ones. Screen, session and user totals are unaffected.
The per-user screen count tops out. Each user's screen list keeps their heaviest 50, so a very broad user reports 50 rather than their true total. Their top screens are still their top screens.
Visits stores aggregates only. No record ids, no URLs and no IP addresses are written to your workspace at any point — the raw logs are read a day at a time and deleted as they go.
Visits is refreshed by the Visits (ELF) step of a full analysis, over a period you choose. It defaults to the same window as the DML analysis, so the two views line up.