Journey Findings

Journey Findings looks for patterns that repeat across whole user journeys: the events and AI traces of one visit, one order, or one user’s session, in the order they happened. Findings reads one AI trace at a time; Journey Findings reads the whole journey, so it can connect what the AI did with what the user did next, such as “the AI quoted a wrong shipping price, then the user left”. Journey Findings works on the journeys your linking rule forms, across all your projects. Open it from Journey Findings in the sidebar.

​

Discover patterns

On the Discovery tab, choose New discovery and describe what to look for, or pick a ready-made question such as Where did it go wrong for the user?. You can also set:

• Quality: Cheapest, Balanced, or High quality. Higher quality costs more.
• Journeys last active: the last 7, 30 or 90 days, or all time.
• Events the journey must contain, and At least this many events.
• Share of journeys to analyze: analyze a sample to keep the cost down.
• Only journeys with AI traces (on by default).

Choose Discover. DataFramer drafts a few questions from yours, asks them about every journey, and groups the journeys into findings. Each finding has a name, a description, and the journeys in it; a finding that describes something going wrong is marked Failure. Open a finding to see its journeys, each with why it belongs there, and open a journey to see its full timeline. Refresh analyzes the journeys the session has not analyzed yet. If you edit the question or the filters, the current findings stay until the next refresh replaces them.

​

When journeys are analyzed

A journey is analyzed once it has had no activity for 24 hours, because events can arrive late: some pipelines load a warehouse only once a day. Before each run, DataFramer also imports new AI traces from every project, so the traces made near the end of a journey are included. If a project’s imports keep failing, it stops holding runs back after 24 hours, and an amber box names it: “Could not import AI traces from project X since [date]: [reason]. Journeys analyzed since then do not include these traces.” The box goes away once the project imports again. A journey larger than 25 million tokens is skipped; the status line under the session shows how many were skipped and why.

​

Track a finding

Open a finding and choose Track. A tracked finding keeps the question and filters it was found with, and every tracking refresh adds the new journeys that match it. Tracking never removes a journey and never creates new findings. On the Tracking tab:

• Choose how often tracking refreshes: every hour, day, or week, or off. Refresh now runs it at once.
• Each tracked finding shows its journey count over time and how many journeys the last refresh added.
• Turn on the bell to get a Slack message when a finding gets new journeys. Connect Slack under Settings first.
• Unpin stops tracking a finding.

​

Changing the linking rule

Changing the linking rule groups all your events and AI traces into journeys again, so findings would point at journeys that no longer exist. When you change it, DataFramer asks first, then:

• keeps your discovery sessions, with their questions and settings, but clears their results;
• deletes your tracked findings, with their alerts and trend history.

Changing the paths of a field the rule groups journeys by does the same, and DataFramer asks first in the same way. These fields are the Journey id field, plus the match field (linking by a field) or the User field (linking by user and time gap). Run your sessions again once the journeys have been linked under the new rule.

​

Customer-hosted deployments

If your data stays in your own cloud account, Journey Findings runs in your data plane: the events and AI traces are read there, and the descriptions and explanations stay in your bucket; your browser reads them from your gateway.

​

Next steps