Optimize — Process Analytics
Process analytics computed from real run history — success rate, instance duration (average, median, p95), per-activity bottlenecks ranked by time spent, a duration distribution, and DMN decision hit-frequency. See which step is slow and which path your processes actually take.
Once processes are running, the question becomes where is the time going, and how often do things succeed? Optimize answers it — the equivalent of Camunda Optimize, built into Versus. It reads your existing run and step history and turns it into analytics: throughput, duration, bottlenecks, and decision behavior. No new instrumentation, no extra data collection — it computes from what your runs already recorded.
This is part of BPMN Process Orchestration.
The summary
Workflows → Optimize, with a workflow picker (or All workflows), leads with the numbers that matter:
- Total runs and how many were timed (had a start and finish).
- Success rate and failure rate.
- Average and median instance duration.
- p95 duration — the tail. The 95th-percentile run is what your SLA actually feels; an average hides it.
Activity bottlenecks
The core of Optimize is the bottleneck chart: every activity ranked by total time spent across all runs, with its average duration, run count, and failure count. This is how you find the slow step — not the one that feels slow, but the one that, multiplied by how often it runs, dominates your process time. Speed that one up and the whole process gets faster.
In the example, Deploy change, Verify traffic, and Validate config each ran five times; the bars make their relative cost obvious at a glance.
Duration distribution
A histogram of instance durations shows the shape of your process time — tight and predictable, or spread with a long tail. A bimodal distribution often means two real paths through the process (a fast happy path and a slow exception path); the distribution surfaces it.
Decision analysis
For processes that use DMN decision tables, Optimize also reports decision rule-hit frequency — which rules actually fired, and how often. That tells you which branch your processes really take: if the "high risk" rule almost never fires, your rules (or your world) may have changed, and it's evidence to tune them.
SLA and throughput, from real history
Because Optimize computes from actual run records, its numbers are ground truth, not estimates. Point it at a single workflow to tune that process, or at All workflows to see the health of your whole automation estate — success rate trending down, a p95 creeping up, a specific activity that's started failing. It's the feedback loop that turns process orchestration from "it runs" into "it runs well".
Pair it with the Operate console — Optimize tells you what to improve; Operate is where you act on the instances that need attention right now.