Skip to content

Reference & Data Model

KPI recalculation (kpi_value)

KPIs are not computed on display but maintained by a background pipeline that reacts to every data change.

Every change to a table that feeds into a KPI (e.g. costs, tasks, risks, orders), plus a clock tick for time-dependent KPIs.

Data change → delta inbox (per affected KPI and scope)
│ every 2 s
▼
Drain: apply delta (±count, ±sum) or recompute the scope completely
│
▼
kpi_value (stored value) → client sees the change live
  • Delta inbox: Every change produces entries “what changes by how much” per KPI and scope (project, program, portfolio, organization). Where a delta cannot be computed safely (e.g. changed project master data, date fields), the scope is marked as “recompute”.
  • Drain: Every 2 seconds the server works through the inbox (max. 8 tenants, 50 scopes per tenant) and writes the new value only on an actual change. Subordinate values (project) trigger superordinate ones (program, portfolio).
  • Clock: Every 5 minutes, time-dependent KPIs (the EVM KPIs) are recomputed, even without a data change.
  • Reconcile: Every 6 hours the oldest scopes are fully recomputed against the raw data to eliminate drift.
  • Snapshots: Daily (UTC) a day value is recorded per scope; from this comes the change “vs. previous day” in percent.
  • First access: A scope never computed before initially shows an empty value (em dash); the computation is triggered on opening or caught up by the backfill (every 6 hours).

Reverted or deleted data runs through the same pipeline as an opposite delta. Archived objects still count; deleted ones no longer do.

Scoreboards and KPI cards on all levels. How current a value is and what “live” means is described in Freshness of KPIs.

  • Values lag the data change by up to a few seconds; breakdown charts, by contrast, are always computed live.
  • A failed computation of a scope is retried after 5 minutes; the old value remains until then.

See KPI overview · EVM.