Start an investigation

An investigation is TraceX working one issue through to its root cause. Each one gets an id such as TX48211794554145, a page under Investigations that updates live, and a report at the end.

You can start one by hand from the dashboard, or let issues start investigations on their own from the places your team and customers already report them.

Issues reach TraceX where they are reported, and the answer goes back to the same place.

From the dashboard

Click New investigation (top right, or press ⌘K and search). Describe the problem the way it was reported to you.

What to include

Include	Why it helps	Example
What people see	Tells TraceX which symptom to explain	“Payment fails with card_declined”, “the dashboard shows 0 orders”
Who or what is affected	Gives it records to look up	an order number, a customer email, a request id, an account id
When it started	Narrows logs and suspect changes to the right window	“since 09:02 UTC”, “after yesterday’s release”
How widespread	Separates one bad record from a general outage	“only EU customers”, “every order over $500”

The more specific, the faster TraceX finds it. A good description reads like a message you’d send a colleague:

Since about 09:00 UTC customers can’t finish checkout. Orders stay in pending_capture and the payment page shows “Something went wrong”. Example: order 48213, customer [email protected].

You can also ask about your data in the same box, for example “Which campaign did [email protected] run last?”. Questions like that are answered straight away instead of running a full investigation.

From where issues arrive

Connect intake integrations and issues start investigations without anyone opening the dashboard. The result goes back to the place the issue came from.

Source	How an investigation starts	Result goes to
Slack	Mention @TraceX in a thread or message it	The thread
Gmail support inbox	A new thread to the connected address	A draft reply you approve
Jira	Assign the issue to the TraceX account	A comment on the issue
Linear	Add the TraceX label	A comment on the issue
Zendesk	Add the tracex tag	An internal note
Intercom	Tag the conversation tracex	A note for your team
PagerDuty	An incident is triggered	A note on the incident
Opsgenie	An alert at your chosen priority fires	A note on the alert
TraceX CLI and AI tools	start_investigation from your tool	The investigation link

Automatic intake from Jira, Linear, Zendesk, Intercom, PagerDuty and email is part of the Team plan (Plans).

What happens next

An investigation moves through three stages:

1. Triage: TraceX works out what kind of problem it is and where to look.
2. Diagnosis: it gathers evidence from your databases, code, logs, errors and recent changes, and tests explanations against it.
3. Fix: when the root cause is grounded, it drafts the change and how to check it.

The page updates live while TraceX works: which agents are running, the queries and searches it makes, and what each found. You get a notification in the dashboard when it finishes. See How TraceX works for the detail.

When an investigation can’t finish

Sometimes the evidence isn’t enough, for example when the relevant logs aren’t connected or the description has nothing to look up. The report then says what is missing instead of guessing. Add the source or the detail and start it again.

Duplicates

When the same problem is reported again while an investigation is open, for example by three customers in ten minutes, TraceX links the reports to the open investigation instead of starting from scratch. Everyone gets the same answer.

Limits

Each plan includes a number of investigations a month (Plans). Admins get an email at 80% and 100%, and a banner shows in the dashboard. See usage any time under Settings → Plan and limits.

Related

• Read a report
• Fixes and pull requests
• Conversations for questions that don’t need a full investigation