Troubleshooting

Who this is for. Anyone with SmartResearch access: analysts, managers and finance leads. Access. Included for everyone at an organization whose subscription covers SmartResearch, whatever their role. No permission needs to be assigned. See the Roles and Access Reference.

Most SmartResearch problems are not faults. A number that looks wrong is usually a period difference, a currency with no activity in the window, or a metric definition that changed. Work through the situation that matches your symptom before you contact Support, because most of these resolve without a ticket.

Question returns an error

When SmartResearch cannot process a question, it returns an error message. Rephrase the question using different terms or a simpler structure. For example, replace a question that combines several dimensions and a long period with a narrower one covering a single dimension and a shorter period.

Question runs slowly or times out

Questions show their progress as they run, naming the current step with an elapsed timer. When a question takes longer than expected, SmartResearch notes that it is still fetching data. On timeout, it offers to try a smaller date range, or to email you the results when they are ready where that option is configured.

Narrow your filters to a shorter date range or fewer vendors.

Results run to many rows

Data tables show a row count and paginate, with a rows per page control. Use the pagination controls to move through results, or select Export all rows as CSV to download the full result set, or Export CSV to download only the page as shown.

A metric shows as Not validated

The definition has not been validated against an instance, or its formula changed since it last was.

To resolve this, open the metric with Validate & Edit, select an instance under Instances to validate against, select Revalidate, then save. Use the status filter in the catalog to find every metric currently unvalidated. For more information, see Create or Edit a Metric.

A number does not match your report

This is usually a definition difference rather than an error. Work through the ordered checks in Troubleshooting, which covers the instance, the connection, the period, the entity scope, the metric definition and who is asking.

Ratio metric returns an extreme value

Ratio metrics divide by recent activity. When little or nothing posted in the metric’s window, the denominator approaches zero and the result becomes very large.

This points at the input data rather than the formula. Check whether recent bill or invoice data has loaded. SmartResearch usually flags this in the answer and recommends confirming before you treat the figure as a live signal.

A currency is missing from a result

Sum-based metrics return one row per currency, but only for currencies with qualifying transactions in the period. A currency your organization transacts in can be absent because nothing posted in that window.

Ask a follow-up question for the underlying volume by currency to confirm whether the gap is real.

A metric you used has disappeared

Release 26.8.1 replaced the standard metric catalog with Core Metric Catalog v1. A metric you relied on was removed rather than renamed.

Check the catalog for an equivalent, or recreate the metric. For more information, see Create or Edit a Metric.

Data or features are missing

You may lack the necessary permissions, the field may not be connected, or your entity scope may exclude the data.

Check your entity setting in User Preferences first, then contact your Auditoria administrator. For more information, see What Data SmartResearch Can Reach.

Work through the checks in order

When no symptom above matches, or a figure simply looks wrong, work these six checks in order. Most surprising answers are scope rather than error. SmartResearch answers the question it was given, against the data it was pointed at, over the period it was told to use, and one of those three is usually different from what you assumed.

1. Check the instance

The SmartFlow Instance selector decides which system of record and mailbox the answer came from. A tenant can carry many instances with similar names, and the wrong one gives a correct answer about the wrong data.

If the instance shows No Connections in the instance table, it cannot answer from data at all. See SmartResearch Instances.

2. Check connection health

Select Connection sync status. If a connection has stopped syncing, the answer reflects data as of the last successful sync rather than now. If it reports that connections are not visible to your account, the scope is narrower than the instance suggests, which is an access question for a Super Admin.

3. Check the period

Every answer states its reporting window. Compare it with what you assumed. A single figure can combine two periods: a DPO can use a trailing 90-day billing denominator against a point-in-time payables balance.

Your Default time range applies to any question that does not state a period. Last 30 days and Last 12 months answer the same question differently, and neither is wrong.

4. Check the entity scope

Default entities narrows which entities are counted. A scope narrower than you expect returns fewer rows rather than an error, which is why it is easy to miss.

5. Check the metric definition

Where the answer cites a Metrics Studio definition, the formula is published in the catalog. When a figure disagrees with a report, comparing the two formulas usually settles it faster than tracing the data. Check the validation status too. It describes the definition, not the answer.

Two things catch people out here. For Days Payable Outstanding, Days Sales Outstanding and the Collection Effectiveness Index, confirm whether your own report uses a cost of goods sold basis: these metrics use a billed basis, so the two will not agree. And check Version history on the metric for a recent formula edit, which explains a figure that changed without your scope changing.

6. Check who is asking

Access differs by role, so two people can ask the same question and see different totals. That is the system working as intended, not a fault.

Implausible figures

SmartResearch flags results it cannot reconcile with normal ranges rather than presenting them flatly. For example, a DPO of 4,074.9 days is returned with an explicit caveat: that a healthy DPO is usually 30 to 60 days, that the result almost certainly reflects a data quality issue rather than operational reality, and that recent bill posting and open AP composition should be reviewed before the figure is used for cash planning.

Treat that caveat as the answer. The calculation is correct; what it is measuring is not what you assumed. The problem is upstream, in the source data.

Failed steps

Expand the reasoning trace. A step can fail and be retried automatically, and both attempts are listed with timings. A failure followed by a successful retry is not a reason to distrust the answer. A failure with no successful retry is.

Summary of checks

When a figure looks wrong, work through the checks; each row says what a mismatch means and who puts it right.

Related information

• For more information, see Frequently Asked Questions (FAQs).

• For more information, see Auditoria Support.

• For more information, see What SmartResearch Can and Cannot Do.