Project delivery · next-quarter outlook

Spot delivery risks before they escalate.

Enter a project’s current conditions to see how the selected model scores its chance of elevated risk next quarter.

Prototype trained and evaluated on synthetic data

These predictions and reported metrics do not represent actual Pinsout performance or proven accuracy on real projects.

Interactive risk assessment

Score a project scenario

Defaults use synthetic training-data values. Add project management fields below to prefill signals where possible.

Scenario inputs

Current project conditions

21 model features

Saved scenarios stay in this page only and clear when you close or reload it.

Scenario workspace

Compare project scenarios

Compare saved scores against the current adjustable threshold.

0 scenarios 0 flagged
No scenarios saved yet

Enter a scenario above and choose “Add to comparison”.

Practical follow-up

Risk mitigation suggestions

Suggestions are generated by transparent rules from the current inputs, not by the trained model.

Rule-based guidance

Read-only connectors

Bring in project signals

Connect Jira or Trello to prefill issue, blocker, due-date, and team-size signals. You can review and edit every value before scoring.

No changes sent to Jira or Trello

Jira

Issues, assignees, blockers, and overdue work

Not checked yet. Select “Check Jira connection” to begin.

Requires the Pages secrets JIRA_BASE_URL, JIRA_EMAIL, and JIRA_API_TOKEN.

Trello

Cards, members, labels, and due dates

Not checked yet. Select “Check Trello connection” to begin.

Requires the Pages secrets TRELLO_API_KEY and TRELLO_TOKEN.

Review imported signals. Imported counts are proxies: overdue issues/cards stand in for missed milestones; Jira issue links stand in for dependencies; labels and issue types identify defects, blockers, or changes. These mappings may not match your team’s conventions. Review and edit them. Unavailable values stay unchanged; blank model inputs use the model’s median.

Model evaluation

How the selected model performed

All figures below are from generated project records and are not real-world performance claims.

Synthetic test set
At threshold 0.425

Synthetic test confusion matrix

1,280 records
Counts of predicted versus generated test labels
Generated outcomeBelow thresholdFlagged
Not elevated——
Elevated——
Across all synthetic training rows

Global feature importance

This impurity-based summary describes the fitted model overall. It is not causal and does not explain an individual scenario.

Selection basis: Gradient Boosting was selected by validation F1.