Which profiles are online right now. A live count within the online window, a map and feed of the ones just arrived, and the peaks recorded every minute since the project started.
What "online" means
A profile is online when its last activity reached the server within the Online window. Every request the SDK sends, a session update, an event, a view, stamps the profile with the time of that request, and the page counts the profiles whose stamp is recent enough.
That makes "online" a snapshot, not a session. Nothing signs in or out. A profile that closes the app stays online until the window passes, and a profile that keeps the app open but does nothing that reaches the server drops out after the same window, even though the session is still running. The window is the same one the server uses when it records its own counts, so the headline number here and the peaks in the history describe the same thing.
New profiles are the online profiles on their very first session: their session count is 1. They are the bright pulses on the map and the "First-time" rows in the arrivals feed. Everyone else is returning.
Live view and recorded history
The page reads two sources that refresh on their own clocks.
The counts at the top are the server's own view: its latest sample of profiles active within the online window, and those among them on their first session, re-read every 5 seconds. The map and the Just arrived feed are read live from the profiles within the same online window, refreshed every 5 seconds as well, so arrivals show up while their visit is still going.
The recorded history comes from a job on the server that runs once a minute, counts the online profiles at that moment, and keeps the highest value it has seen for each minute, each half hour, each day, and overall. The last hour, the last 24 hours, the 30-day chart with its highs and lows, and the all-time peak all come from that record. The page re-reads it every 10 seconds, so a new minute shows within seconds of being sampled.
Two consequences follow. The history only exists from the moment the job started running for the project, so a new project shows an empty 30-day chart until it has a day of samples. And a peak is the highest one-minute sample, so a spike that lasted less than a minute may fall between two samples and never appear.
Use Cases
Watch a launch or a campaign land
See, as it happens, whether a release, a push, or a campaign brings profiles in.
- Open Online Profiles before the moment and note Online now and Peak in last hour. That is the baseline.
- Watch Just arrived as the moment passes. Rows appear newest first with a First-time badge for profiles on their first session; a campaign that acquires shows badges, a push to existing profiles shows returning ones.
- Watch the map. Bright pulses are first-timers, soft pulses are returning profiles, and the map zooms to the countries with activity, so a campaign aimed at one region should light up there.
- After a few minutes, read Last hour by minute. The minute the traffic arrived and how long it held are both visible; the new-profile series shows how much of it was acquisition.
- The next day, check the 30-day highs. A day that made the list is a day worth explaining.
Know your normal load
Know what a normal amount of concurrent activity looks like, so an unusual one is recognizable.
- Read the 30-day highs and lows beside the history chart. The highs are the days with the largest one-minute peak, the lows the days with the smallest. The distance between them is the range your product lives in.
- Read Last 24 hours by 30 min for the daily rhythm: when the day starts, when it peaks, when it goes quiet. Compare it with the time zone your audience lives in.
- Note the all-time peak. If it dates from a launch or a load test that no longer describes normal use, reset it from Projects so the next real peak stands out.
- Come back after a change in infrastructure or in the SDK. The rhythm should hold; a flatter curve after an SDK upgrade can mean fewer requests reaching the server, not fewer profiles.
See where the audience is right now
Know where the profiles online right now are.
- Look at the map. Every dot is a location with at least one profile active within the online window, placed by the country and city the server resolved from its last request. The map zooms to the countries that have activity.
- Read the count above the map, the size of the live set. It describes the map's own window and cap, so on a busy project it can be less than Online now.
- Read Just arrived for names, country and platform, and open a row to see that profile.
- For the full geography of sessions over a period rather than this instant, use Sessions, Segments, Countries.
FAQs
Why is Online now higher than the sessions I can see starting?
Online now counts profiles active within the online window, whatever they did. Events, views and session updates all count, and a profile stays online until the window passes after its last request.
The app was closed. Why is the profile still online?
The last request stamped the profile, and the stamp stays valid for the online window. There is no sign-out; the profile drops off when the window passes.
Why does the map show fewer profiles than the KPI?
Both measure the same online window, but the KPI is the server's own count while the map holds a live sample, at most 1,000 profiles, that reaches the page with up to a minute of delay. On busy projects the map undercounts.
Why is the 30-day chart empty?
The server job that records the daily peak has not run for a full day on this project yet. Values appear once it has.
How do I reset the all-time peak?
On the Projects page, in the project's Online Profiles section, use Reset maximum counts. The next sample becomes the new maximum.
What is a bright pulse on the map?
A location where at least one profile is on its very first session, the same profiles the New profiles count and the First-time badge in the feed describe. Soft pulses are returning profiles.
Do these numbers refresh on their own?
Yes, while the tab is visible: the counts and the map every 5 seconds, the history every 10 seconds. The server itself records a new sample once a minute.