Overview

Overview is the first page an operator sees after signing in — it is the dashboard’s default route (/). It has no controls that change platform state; everything on it is a read view of your organization’s traffic over a selectable time range. A Products panel shows the status of each product and links to the Products page, where trials and billing are managed.

Time range

The picker in the header selects the look-back window for both the KPI cards and the recent-activity feed:

Range	Look-back	Series bucket
Last hour	1 hour	Hourly
Last 24 hours (default)	24 hours	Hourly
Last 7 days	7 days	Daily
Last 30 days	30 days	Daily

Each load anchors the window to the current time — reopening the page or switching ranges always measures back from “now”, not from a fixed point. Switching ranges also collapses any expanded activity row, since that row’s detail may not survive the refetch.

KPI cards

The three cards summarize the selected window for your active organization:

• Requests — the total count of metered traffic in the window: both structured-transaction REST calls and HSMaaS-over-TCP (payShield) commands. Visa/Mastercard ISO 8583 TCP traffic isn’t metered yet, so it isn’t counted here — see Recent activity below.
• Error rate — the share of those requests whose HTTP status code was 400 or higher, shown as a percentage with the raw error count underneath. This counts both client errors (4xx, for example a malformed request body) and server/delivery errors (5xx, for example a 502 when your responder wasn’t connected — see Delivery needs a live TCP connection). A 2xx/3xx response is never counted as an error.
• p95 latency — the 95th-percentile response time across the window’s requests. Because it’s a percentile rather than an average, one slow outlier barely moves it; a rising p95 means a real share of requests are getting slower, not just one.

All three go blank (—) rather than showing stale numbers while a new range is loading, and the whole metrics section is hidden — replaced by an error banner — if the load fails, so you never see cards from one range next to a feed from another.

Reading the pattern

• Error rate climbing while request volume stays flat usually points at something on your integration (a malformed request body) or your responder (not connected, timing out) rather than at Switchbench’s engines. An invalid, expired, or revoked API key is rejected before the request is metered at all, so it shows up as lower request volume, not a higher error rate — see Request volume dropping to zero below. Expand a failing row in the activity feed to see the actual request/response bodies, or check Audit Log for a recent config change (key revoked, certificate revoked, routing rule edited) that lines up with when the errors started.
• p95 latency spiking with error rate flat is more often a responder or network issue — your backend is answering slowly rather than failing outright. Compare against the Last hour view to see whether it’s an ongoing trend or a brief spike.
• Request volume dropping to zero usually means traffic stopped reaching Switchbench at all — check that your client certificate is still valid (Certificates) and that your API key hasn’t been revoked or rotated out from under your integration (API Keys).

Recent activity

The table below the KPI cards lists the newest requests in the selected window, newest first, capped to the 10 most recent. Each row shows the request time, method/transport, endpoint, status, and latency. A REST row’s method is a real HTTP verb (POST, and so on); an HSMaaS row’s method is the literal TCP — HSMaaS has no HTTP method of its own, but SMP still assigns it an HTTP-style status class, so the status column (color-coded — green for 2xx, amber for 4xx, red for 5xx) reads the same way for both.

Click Inspect on a row to expand it and load that event’s masked request/response bodies (JSON bodies are pretty-printed). Not every event has stored bodies — some are logged without a captured payload — in which case the expanded row says so instead. Only one row can be expanded at a time.

Recent activity is not the same as Audit Log. This feed is transaction traffic — every metered request, not just configuration changes. Audit Log (in the sidebar) is a separate, longer-lived record of configuration changes made through the dashboard or CPS API — API keys created or revoked, certificates issued or revoked, routing rules and engine config edited. When something on Overview looks wrong, check Audit Log for a change that might explain it, not for the failing transactions themselves.

The KPI cards and the feed cover every metered request, not only REST: the structured-transaction REST routes and HSMaaS-over-TCP (payShield) commands are both logged to the same usage record, so a spike here can come from either. Visa and Mastercard ISO 8583 traffic over TCP is not currently metered and won’t appear here. To understand exactly what counts as a “request” on each transport, see REST transactions and HSMaaS (raw payShield) over mTLS TCP.