Appearance
Reports
What the reports are, how the numbers are built, and the two things that make them look wrong when they are not.
The report families
Reporting is organized into families, each its own section under the Reporting menu: Account (signups, expiry, who created what, voucher and top-up code activation and availability, and voucher groups — how many groups and people you served with vouchers each month or year, see Counting group visits with vouchers; and voucher utilization — how many codes from each batch, profile or zone were used, are still valid or expired unused, with the data, time and face value they account for), Login (active sessions, auto-login, browser fingerprints, the login record, login and login-frequency summaries — overall, per profile, per device — login errors, session concurrency, dwell time — overall and per profile — top accounts, returning accounts, a login heatmap, session termination reasons, and a zone comparison), DPI (deep packet inspection traffic by application and client), Bandwidth (a summary, usage per account, top usage, the current top consumers, and bandwidth against login volume), Impression (portal/ad impression counts), Engagement (campaign impression, click and offline summaries, survey responses, and consent/marketing reporting), Delivery log (SMS, email and PMS transaction logs), Insight (device insight, registration insight, social network insight and data), Payment (transactions), AI (the AI reporting assistant, covered on its own page), and Scheduled reports (recurring exports run on a cron rather than clicked by hand). This page is about how reporting works in general; each report's own purpose is covered where it lives in the menu.
Hourly buckets and the current hour
Most report numbers don't write to the report tables as they happen — they accumulate as fast in-memory counters through the hour, keyed by tenant, zone and hour. A scheduled roll-up moves the previous completed hour's counters into the hourly report tables and then deletes the counters it just processed; it deliberately skips whatever hour is still in progress. That roll-up runs once an hour, five minutes past, so a finished hour lands in the reports a few minutes after it ends.
The practical effect: the current, still-running hour is never in the report tables yet. A report queried mid-hour will look short for "now" compared to a moment ago — that's not missing or lost data, it just hasn't been folded in. This is the same reason the AI reporting assistant flags a result covering today with a note that the current hour isn't included.
Time zones
Relative periods — "today", "this week", "last month" — resolve against the tenant's own configured timezone, not the server's or the browser's. The same convention applies wherever Kiwire reasons about dates on a tenant's behalf, including the AI reporting assistant. Because the hourly aggregation itself runs on server time, a report queried right at a day or month boundary can look a little short until the aggregation pass for that boundary hour has caught up — the same current-hour gap described above, just more noticeable at a boundary.
Exporting
Nearly every report screen has an Export button alongside its on-screen data view, producing a CSV of the same filtered result set rather than the raw hourly buckets. For reports an admin wants delivered regularly rather than pulled by hand, Scheduled Reports automates the export on a recurring cron instead.