Almanac
Microsoft/power-automatePower Platform

Consultant KB for Microsoft Power Automate end to end: cloud flows, desktop flows and RPA, connectors and integration, AI and agent flows, process and task mining, approvals and human-in-the-loop, ALM and solutions, governance and security, licensing and capacity, monitoring and troubleshooting, plus cross-cutting decision guides. Implementation notes, configuration decisions and the gotchas that bite on real projects. Populated by the daily author agent from the Power Automate release plans, docs repo and product blog, plus the author's own consultant notes.

power-automate/05-process-mining/feature-mining-analytics-and-recommendations.md
feature-mining-analytics-and-recommendations.mdv1 · history
CurrentApplies to Process miningUpdated last monthSource Microsoft Learn ↗

What it does

Once a process is analysed you get a process map, a set of KPIs and a group of prebuilt visualisations for reading it. The map layers frequency, performance and rework metrics over the same activities and transitions, and the analytics pages let you filter down to variants, activities and time periods.

Key facts

  • The process map shows activities as nodes and transitions as edges. Three metric layers: Frequency (total count, case count, maximum occurrence in case), Performance (total, mean, maximum, minimum duration), and Rework.
  • Rework metrics separate self-loops (an activity directly followed by itself) from loops (repetition with at least one activity in between), plus loop inflow, loop outflow and net loop gain. Net loop gain flags which activities start loops and which close them.
  • Default KPIs: Average Case Duration, Self-Loop Cases %, Loop Cases %, Rework Cases %. Median Case Duration and Resource Count are Premium.
  • Premium features need your own Power BI workspace or a Fabric workspace (preview) connected. That includes the case filter, start date filter, custom attribute filters, the Average duration of cases over time chart, and the Time Analysis view.
  • Without a connected workspace the Filters pane offers the activity selector only.
  • Variant analysis comes as Variants by frequency and Variants by time. Selecting variants refilters the map, the KPIs and the other visuals.
  • Variant DNA is available in both default and premium views. It colour-codes and abbreviates the activity sequence per variant, which is the quickest way to spot non-compliant paths and loops.
  • The standard Power BI report applies a filter of ViewID = 1 — usually the view named Default — across all pages unless changed.
  • Automation recommendations appear as blue icons on process map activities. +Automate activities takes you into the flow designer with connector recommendations for those activities.

When to use / skip

The analytics are where the engagement earns its fee, so plan for the analysis sessions rather than treating them as a readout. Rework and loop metrics are the fastest route to a finding a client will act on — "18% of your cases go through this step twice" lands harder than a duration chart. Skip the deep time analysis if you haven't got the premium views wired up, because without the case and date filters you can't drill into a bottleneck far enough to say anything useful.

Configuration decisions

  • Whether to connect a Power BI or Fabric workspace, which gates roughly half the analytical capability.
  • Which custom attributes were brought in at ingestion, since those become the custom filters — you can't add them later without reloading.
  • Whether the standard report is used as shipped or customised in your own workspace, and who then owns that report.
  • Which KPI the client's improvement target is expressed in, agreed before the analysis rather than after.

Gotchas

  • The premium/default split isn't obvious until you go looking for a filter that isn't there. Confirm which side of the line the client's licensing puts them on before promising an analysis approach.
  • Average case duration is badly distorted by outliers. The median is the more honest number and it's a Premium KPI, which is an awkward conversation.
  • The ViewID = 1 report filter catches people who build extra views and wonder why the numbers don't move.
  • Automation recommendations are connector suggestions based on the applications in the process, not a judgement about whether automating that step is a good idea. They're a prompt, not a business case.
  • Filtering to a small variant makes the KPIs recalculate on that subset. It's easy to quote a number from a filtered view and forget the filter was on.

Consultant notes

  • Run the analysis as workshops with the process owner in the room. The tool finds the loops; only the business can tell you why they exist, and the why is what the recommendation hangs on.
  • Lead with rework and variant counts, not with process maps. Executives glaze over at the spaghetti diagram and engage with "there are 340 ways this process runs".
  • Be careful attributing time to activities when the timestamps are system events rather than human work. Long durations often mean waiting, not working, and the fix is different.

Worth another look once Fabric workspace support leaves preview and the premium/default line moves again

Was this accurate?
Related
Process mining overview05-process-miningFrom mining to automation05-process-miningFlow analytics10-monitoring-and-troubleshootingTask mining05-process-miningProcess mining data ingestion05-process-miningApplication Insights integration10-monitoring-and-troubleshooting