Is the risk score produced by AI?

No. It is arithmetic over stage medians, board state and days to deadline. Every term that changes the score adds a reason with its points, and the reasons sum to the score.

What does a project with no deadline score?

Zero, with the reason listed. There is nothing to be late against.

Why does my score say some stages have no history?

A stage needs at least three recorded intervals to use its own median; below that it borrows the organisation median. The reason is listed with zero points so you know how much to trust the estimate.

How often is the score updated?

Once a day, by the analytics job, which stores a snapshot with the reasons. The risk tab shows the newest snapshot per project.

High at 70 or above, watch from 40 to 69, low below 40.

A risk score you can explain to a client

A project's risk score in Tasked is a number from 0 to 100 built from reasons you can say out loud: how much work is left according to the median times of your own stages, how many days remain to the deadline, and three facts about the board today. Each reason carries the points it added, and the points add up to the score. There is no model call and nothing hidden.

Published September 15, 20265 min read

Three inputs a person can check by hand

The score uses exactly three inputs. First, the median time tasks in your organisation have historically spent in each stage. Second, where each open task of the project sits on the board right now, and how long it has been there. Third, how many days are left until the project's deadline.

Nothing else goes in. No text is read, no external model is called, and the same inputs always produce the same score. The model is versioned, and every stored score records which version produced it, so a change to the arithmetic can never be mistaken for a change in your projects.

How much work is left

For each stage the model takes the median of its recorded intervals across the organisation. A stage with fewer than three samples borrows the organisation-wide median instead, and an organisation with no history at all uses eight hours per stage until it has some.

For each open task, the expected remaining time is what is left of its current stage (the stage's median minus the time already spent there, never below zero) plus the full median of every stage still ahead of it. The stages' totals are added across all open tasks and divided by six productive hours per calendar day, not twenty-four: a deadline three days away is three working days of capacity.

The pressure line

Slack is days to the deadline minus expected remaining days. The base score is 50 when the expected finish lands exactly on the deadline, and it moves 10 points for every day of slack in either direction, clamped between 0 and 100. Two days of slack is 30; two days behind is 70. The line is straight on purpose: a curve nobody can invert is not explainable.

Three cases short-circuit the line. A project past its deadline with open work scores 100, with the number of days late attached. A project with no deadline scores 0, because there is nothing to be late against. A project with no open work scores 0 whatever the calendar says, because nothing left to do cannot be late.

Three modifiers and one zero

After the base, three observable facts about the board add points. Each one appends a reason with a count so the reader can verify it on the board.

Overdue tasks: at least one open task past its own due date adds 15 points.
Blocked tasks: at least one open task marked blocked adds 10 points.
Stale approval: at least one approval pending for more than 3 days adds 10 points.
No history: when some stages borrowed the organisation median, a reason is listed with 0 points. Uncertainty is not risk; it tells you how much to trust the number without changing it.

Bands, and where the score shows

A score of 70 or more is high, 40 to 69 is watch, below 40 is low. The three thresholds are the whole taxonomy; every screen colours from these names and none re-derives its own.

A daily job scores every project and stores a snapshot with its reasons. The risk tab in analytics shows the newest snapshot per project, riskiest first, with the reasons expanded. The Monday owner report names the three riskiest projects with their score and band. The reasons are stored as data and rendered in Hebrew or English from the same numbers, never written into the database as sentences.

Saying it to a client

Because every reason carries its points, the score reads as a sentence. A project at 75 might be: base 55 because we are half a day behind pace with 2.5 expected days of work and 2 days to the deadline; plus 10 because one task is blocked; plus 10 because your approval has been pending for four days. Add the three and you get 75, which is high.

That sentence has no black box in it. The client can act on the third line today, and the producer can act on the second, and both can watch the score move tomorrow when the daily snapshot runs.

Questions that keep coming up

Is the risk score produced by AI?: No. It is arithmetic over stage medians, board state and days to deadline. Every term that changes the score adds a reason with its points, and the reasons sum to the score.
What does a project with no deadline score?: Zero, with the reason listed. There is nothing to be late against.
Why does my score say some stages have no history?: A stage needs at least three recorded intervals to use its own median; below that it borrows the organisation median. The reason is listed with zero points so you know how much to trust the estimate.
How often is the score updated?: Once a day, by the analytics job, which stores a snapshot with the reasons. The risk tab shows the newest snapshot per project.
What are the bands?: High at 70 or above, watch from 40 to 69, low below 40.