How it works

The maths, in plain words.

Due Date Check uses no machine learning and no estimates. It counts what your team already did. Every step below is the same code that runs inside Jira.

1. Read the history

When you open a ticket, the app reads the project's 500 most recent finished tickets (anything with a resolution date), as you, so it only sees tickets you can see.

2. Find similar work

Each finished ticket gets a similarity score against the one you are looking at:

Tickets scoring 4 or more are "similar". It keeps the 30 closest. If fewer than 5 are similar, it falls back to tickets of the same type, then to the whole project, and tells you which it used. With fewer than 3 finished tickets it says there isn't enough history yet rather than guess.

3. Measure them in working days

For each match: working days (Monday to Friday) from created to resolved, and whether it met its own due date.

If your ticket is already under way, the app only counts past tickets that took at least as long as yours has so far. A ticket open for 4 days can't finish in 2, so faster work stops counting.

4. Turn them into dates

The verdict follows the chance: on track at 70% or more, tight from 40%, otherwise likely to slip. Confidence is high with 20+ matches, medium with 8+, low below that.

5. The project page

The same numbers, added up: the share of due dates met overall and per month, each kind of work (by component, else work type) with its promised versus real median days, and every open ticket with a due date ranked by its chance.

6. The Rovo agent

Due Date Coach calls two actions, one for a ticket (optionally with a date you suggest) and one for a project. They return the numbers above as text with ticket keys, and the agent's instructions forbid stating any number they did not return.