Who this is for. Implementation teams and administrators handing a tenant over. See Roles and Access Reference.
A short set of questions worth running before a tenant is handed to its users. Each one exercises a different part of the stack, so a failure tells you where the problem is rather than simply that something is wrong.
Run them in order, against the instance the users will actually select, and signed in as a role a real user will hold rather than as a Super Admin. A Super Admin sees every row, so a permissions problem that will affect every analyst is invisible from that account.
Run the checks
Ask each question in turn. A pass proves one part of the stack works, and a failure tells you which part does not.
Table 1. Go-live checks
Ask |
What it proves |
What a failure means |
|---|---|---|
Current DPO |
A governed metric computes end to end and cites Metrics Studio. |
The metric definition or the data behind it is not ready. |
Top vendors by spend |
Vendor and spend data is readable and rankable. |
Vendor master or bill data has not synced. |
Overdue AP aging |
Ageing buckets compute against today. |
Date coverage or open-payables data is incomplete. |
What is our total outstanding payables balance right now? |
A point-in-time balance resolves. |
Open AP composition is wrong or missing. |
How much was billed by vendors in the last 90 days? |
A trailing-window question resolves. |
Recent posting activity is missing, which also distorts DPO. |
The first three are the workspace’s own starter questions, so they can be run with a single click each. The last two must be typed.
Expected results
The answer names its SOURCES.
The reasoning trace shows no failed step, or a failure followed by a successful retry, which is normal.
The figure is plausible, or SmartResearch says why it is not.
The data table returns rows, pages, and exports to CSV.
A governed metric is cited by name where one applies.
If all five hold for all five questions, the connection, the sync, the metric layer and the export path are all working.
Interpret a failure
Table 2. Diagnosing by pattern
Pattern |
Most likely cause |
Where to look next |
|---|---|---|
Every question returns nothing |
No instance selected, or the connection is unhealthy. |
The header, then the connection sync status. |
Metric questions fail, plain questions work |
The metric definition is missing or not validated. |
|
Totals look implausibly small |
Entity or time-range defaults are narrowing the scope. |
|
One record type is missing throughout |
That object has not synced. |
Your ERP connection and its sync history. |
Answers differ between two testers |
Role-based row filtering, or different personal defaults. |
Compare both accounts’ roles and preferences. |
Do not sign off on a tenant tested only as a Super Admin. Repeat at least the first three checks as an ordinary analyst role. Row filtering means the two accounts can return different results, and users will see the analyst’s view.