Two sources, one question
A worked case for when you add a second source and want to know whether it is behaving like the first, without the two blurring into one set of numbers. The situation and the first two steps are from a real week; the last steps show how the saving and snapshot features are meant to be used.
The situation
An organization ran one production agent runtime, Cruz, as its only source. It then added a workflow tool, n8n, as a second emitter. The same day the Home page started showing an unattributed share where there had been none, and a new batch of candidate principals. Was the new source the cause, or was something happening on the first?
Step 1: separate them
Both emitters were named when they were registered, so they appeared by name in the header's Source filter. With All sources selected, Home shows the organization as a whole, so the two sources' records sit in one total. Choosing n8n alone narrowed Home, Investigate and Findings to that source. Its newest record was a couple of hours old, it had a few hundred records, and almost all of them showed unknown in the Agent column.
Step 2: compare, do not add
On the Compare page the two sources went side by side over 7 days:
| Cruz | n8n | |
|---|---|---|
| Actions (7 days) | about 11,500 | about 260 |
| Unattributed | almost none | nearly all |
Nothing was totalled. The table showed that the unattributed records were the new source's, so the Home spike was a consequence of adding it, not a sign the first source had degraded. The same page showed the first source's own problem separately: the Batches never arrived finding named that source, not the new one.
Step 3: act on each separately
- The new source's records were unattributed because nothing told the platform which agent they belonged to. See A new source's records have no agent.
- The first source's missing batches were a separate problem with their own cause. See Batch gaps and unsealed days.
- Look at each source's own findings with the Source filter, not the combined list.
Fixing one did not touch the other, because their data and chains were never mixed.
Step 4: keep the answer
Save the comparison under a name, and pin a snapshot with a note such as Before attributing the n8n source. After you map the source to an agent, pin a second one: the pair shows the unattributed share falling. Opening a snapshot later shows the numbers as they were, how far each source had got, and how many records have arrived since. It holds only aggregates, and it is labelled as analysis, not evidence.
Step 5: follow a run across both
Where both sources record the same trace ids, Runs seen by both lists the runs and how many records each source holds for them. A run where one source recorded noticeably fewer calls points at something the other source missed. Sources that do not share trace ids show nothing there, and the page says why.
What this case shows
Two sources are easier to understand apart than together. The tools here read each source on its own and put them side by side; they never produce a combined number. If you find yourself wanting one, ask what difference it would hide. See Sources, filters and comparing.