Health overview
Reading the pipeline dashboard, and what each panel is good for noticing.
The Overview view is the health dashboard for the pipeline as a whole. It is built for noticing problems; the explorer is for diagnosing them.
The headline metrics
Four numbers sit across the top: total received, replies accepted, events created, and errors.
Read them together rather than individually. Total received falling while errors stay flat means fewer people are reporting — a field problem. Errors rising while volume holds steady means the system is mishandling messages it is still receiving — a configuration or infrastructure problem. The two call for entirely different responses.
Gateway
The gateway panel reports what share of replies DHIS2 accepted, split into accepted, rejected and unverifiable. Its verdict is healthy, degraded, down, or unknown when there were no replies in the window — no traffic is not the same as health.
Accepted is not received
Accepted means DHIS2 took the reply. Nothing on this deployment confirms that a phone received it. A reply whose fate could never be established is counted as unverifiable and counts against the rate, never for it.
This is the only panel about the outbound leg, and it fails independently of everything else: messages can be parsed correctly and written to DHIS2 perfectly while every reply silently fails to reach anyone. A falling accepted rate while events are still being created points at DHIS2's SMS set-up or the SMS provider, not at eCEBS. The delivery page follows each message further.
Outcome breakdown
The distribution of outcomes across the selected period. Its value is in the shape rather than the numbers — a single outcome growing as a share of the whole is the signal, and it names the stage to investigate.
Throughput
Volume over time, by hour or by day, split into successful and failed. The hourly view shows the daily rhythm of reporting; the daily view shows trends across weeks.
Look for cliffs, not bumps
A step change is almost always a system event — a flow published, a credential expired, DHIS2 becoming unavailable. Gradual drift is usually a field-level change in reporting behaviour. The two rarely have the same cause, and the shape of the line distinguishes them.
By signal code
Which codes are actually arriving, with their EBS family and count.
This is the most direct feedback on whether a configuration matches reality. A code that never arrives may mean reporters were never told about it. A code arriving far more than expected is worth understanding before it is treated as a surveillance finding.
Top unparsed messages
The most common messages the system could not read. Blank messages are shown as such.
This panel is where configuration improvements come from. A message shape appearing repeatedly is reporters consistently doing something the grammar does not allow — often a spelling variant or a different separator. Each entry is either a grammar that should be widened or a group of reporters who need to be told the correct format.
A weekly routine
Check the headline metrics and the throughput shape, then read the top unparsed messages and the code breakdown. The first two tell you whether anything broke; the second two tell you what to improve. Anything anomalous is followed into the explorer, filtered to the period concerned.