Research an insight for a root cause
Before you begin
You need Act on FleetHQ Insights, and FleetHQ must be enabled for your organization. You need a finding that shows detector-only analysis — one marked Not researched — that is not already being researched.
This scenario spends AI credits from your organization's monthly balance. The confirmation dialog shows the balance before you commit, and that dialog is your last chance to stop. If you'd rather not spend, follow the steps to that point and read the rest.
What you will accomplish
You'll ask FleetHQ to investigate one finding that only reports measurements, and read the root cause it comes back with.
Steps
In Insights and recommendations, select a finding marked Not researched.
Read the evidence it already reports, and the next check it names. This is what a detector gives you: measured facts, with no claimed root cause.
Select Research this one.
In Research this insight?, read the remaining monthly credit balance and the explanation of what the investigation will add.
Select Research this one again to confirm the spend.
Note that the card is now marked Researching. Leave the page if you want — the investigation continues in the background, and the state survives a reload.
Come back when the Researching badge has cleared, and select View analysis.
Read the root cause, the evidence, the ruled-out alternatives, the confidence level, and the recommended plan that now sits beside the finding.
What you just did
You crossed the line between what FleetHQ measures and what it concludes. A detector finding is deliberately modest — it reports what moved and when, and it does not invent an explanation, which is why so many findings arrive without a root cause. Manual research is the paid escape hatch for one finding you care about, and what came back is the same shape of analysis you read in the second tutorial: a root cause, the evidence behind it, the alternatives that were tested and rejected, and a confidence level. Two things to remember. If the result says Deeper dive needed, the investigation found something but not a complete answer — read it as a lead, and FleetHQ will retry it automatically. And if a request fails or seems stuck, do not resubmit it: FleetHQ blocks duplicate in-progress research, and the card or dialog will tell you whether the obstacle is credits, eligibility, or an unfinished run.
Next steps
Research an insight — the full task, including every reason a request cannot start.
How FleetHQ finds and researches insights — detectors, the evidence boundary, and where paid research fits.
Review an insight's analysis and evidence — read the result the way you would any other analysis.