Nembl
Workflows
Analytics

Workflow Analytics

Nembl measures every workflow instance as it runs — how long each phase takes, which phases stall most often, what percentage of instances complete vs. get cancelled. Analytics lets you turn those measurements into concrete process-improvement insights.

Two views:

  • Per-workflow analytics — /admin/workflows/[workflowId]/analytics — deep detail on a single workflow
  • Company-wide analytics — /analytics — request volume, SLA compliance, and cross-workflow performance

Per-Workflow Analytics

Open a workflow's detail page and click Analytics in the action bar.

Stat Cards

Across the top, five headline metrics:

CardWhat it shows
Total InstancesAll-time instance count for this workflow version, plus a "last 30 days" detail
Completion RatePercentage of instances that finished as COMPLETED. Color-coded: over 80% green, 50–80% neutral, under 50% red
Status BreakdownCounts of COMPLETED, CANCELLED, TERMINATED, and ACTIVE instances
Avg Cycle TimeAverage duration from start → end, completed instances only
Bottleneck PhaseThe phase consuming the largest share of total cycle time, plus its average duration

Completion Rate Donut

A four-category donut chart with consistent colors:

  • COMPLETED — green
  • CANCELLED — red
  • TERMINATED — gray
  • ACTIVE — blue

Phase Duration Bar Chart

One bar per phase, height = average duration in milliseconds. Hover for exact numbers.

Phase Detail Table

Per-phase breakdown, sortable by any column:

  • Phase Name
  • Avg Duration
  • Median
  • P90 — 90th percentile (helps surface long-tail slowness that averages hide)
  • Count — how many instances went through this phase
  • % of Total — share of the workflow's total average cycle time consumed by this phase

Bottleneck highlighting: phases where % of Total > 30% are flagged with a red "Bottleneck" badge in the table. 30% of the cycle on a single step is strong evidence the phase is worth splitting, automating, or reassigning — it's the threshold competitors like Pipefy and Camunda use for their auto-flagged bottleneck reports.

Flow Efficiency

Deferred — flow efficiency (value-add time vs wait time) requires classifying phases as "value-add" vs "wait," which is not yet configurable per phase. Planned for a later phase.

Company-Wide Analytics

/analytics is the cross-workflow roll-up page.

  • Request Volume — 30-day trend line of inbound request counts
  • SLA Compliance — gauge showing percent of requests meeting their SLA
  • Service Popularity — which services attract the most requests
  • Workflow Funnel — start / in-progress / completed / cancelled across all workflows
  • Resolution Time Distribution — histogram of completed-request durations
  • Workflow Performance table — per-workflow: instance count, completion rate, avg cycle time, deep-link to per-workflow analytics

CSV Export

On the per-workflow analytics page, click Export CSV to download the phase detail table:

Endpoint: GET /api/workflow-analytics/export?workflowId=<id>

Columns: Phase, Count, Avg (ms), Median (ms), P90 (ms), Avg (formatted)

Filename: <workflowName>-analytics.csv

Use cases:

  • Hand off to a stakeholder who wants the numbers in a spreadsheet
  • Attach to quarterly process-review documents
  • Diff pre/post against a change to quantify improvement

Date Ranges

Analytics currently uses:

  • Last 30 days for instance count cards
  • All-time for phase-level aggregates (avg, median, P90, count)

Configurable date-range filtering is planned for a future release.

How Metrics Are Computed

  • Phase durations come from WorkflowPhaseHistory rows (enteredAt → exitedAt). Phases still open at query time are excluded from averages.
  • Cycle time is the sum of completed phase durations per instance, then averaged across completed instances.
  • Bottleneck detection sorts phases by avgDuration, picks the top one, and flags it when its share of total average cycle time exceeds 30%.
  • P90 is the 90th-percentile phase duration across all recorded entries for that phase.

Interpreting Common Patterns

  • High P90 with low median — a few extreme outliers; dig into individual instances rather than changing the phase design.
  • Both P90 and median high — the whole phase is slow; consider automating, splitting, or reassigning.
  • High completion rate but long cycle time — process works but is slow; focus on the bottleneck phase.
  • Low completion rate — check CANCELLED and TERMINATED counts; something is routinely failing before the end.
  • Bottleneck is a waiting phase (approval, timer) — consider parallel paths, shorter SLAs, or delegated approval to reduce the wait.

Related

  • Workflow Execution — how phases activate and advance
  • Responsibilities — assigning parties so the right people get phases (affects wait time)
  • Audit Trail — every phase entry/exit is also audited if audit.request_logging is enabled
VersioningTemplates