A solution which turns out cheaper or faster than deduced is fine: only a drift for the worse counts.
KPIL indicators recalculation
The deduction chose the solution for its indicators: duration, cost, distance, strain, wait… as declared by the KPIL script of the project. This strategy replays the same indicators with what the field tells now and compares them with the values the deduction found. When an indicator gets worse than deduced by more than its threshold, the solution is Wrong: another solution, or another schedule, would now do better.
implemented byKpilIndicatorsRecalculationStrategy · gind-npm-strategy-supervision-kpil-indicators
01 · Indicators
Those of the KPIL script, concept Function
The default configuration of a project lists every indicator of the tasks (Function:*) the KPIL script declares, with its unit, its aggregation and its objective (minimize or maximize). On IMT Albi Crisis: cost, distance, duration, strain and wait.
A new indicator added to the script after the configuration was saved is shown below the table as “declared by the KPIL script but not configured here”, with a button to add it.
02 · Replay
What each task weighs now
| Task in the To-Do List | Value used |
|---|---|
| Done (or in error) | Its measured value, or its expected value when it was not measured. |
| Going on (accepted) | Duration: max(expected, time spent since the acceptance). Cost and strain grow with the time: expected × time spent / expected duration once the task takes longer than planned. |
| Rejected or cancelled | 0. |
| Not reached or not accepted | Its expected value. |
The tasks are then combined as the script says: CUMULATIVE (sum), MULTIPLICATIVE (product), MAXBYPATH or MINBYPATH (longest or shortest sum along the nominal flows of the process, from its first nodes to its last ones; exception and event flows are left out). The same aggregation applied to the expected values gives the deduced value.
03 · Score
The worst drift wins
drift = (now − deduced) / |deduced| (deduced − now for an indicator to maximize)
Wrong = max over the checked indicators of drift × 100 / threshold (only when it gets worse)
Wrong = max over the checked indicators of drift × 100 / threshold (only when it gets worse)
With the default threshold of 20 %, a duration of the solution 20 % longer than deduced gives Wrong = 1; 10 % gives half the threshold. The contribution reads, for instance, “Expected duration +25 % (1500 s now against 1200 s deduced)”.
04 · Configuration
One row per indicator
05 · Watch points
Good to know
The Weighted Graph Distance does not weigh positions: the distance and the duration of the moves are judged by this strategy.
This strategy compares the whole solution, done and remaining; the earned value compares the work done by now with the work planned by now.
gind-detection-engine-kpil-indicators-strategy/…/KpilIndicatorsRecalculationStrategy.javaIndicators of the script, projected value of a task, aggregations, drift.gind-npm-strategy-supervision-kpil-indicators/src/kpil-indicators-configuration.component.*The table of the indicators and their thresholds.