When the Average Looks Fine but the Exceptions Don’t
How process mining exposed long-tail risk in a utility connection and development review process.
The Challenge
A utility-related organization wanted to understand whether its development and connection review process was predictable enough to support applicants, municipalities, utility providers, and internal teams.
At first glance, performance looked relatively stable.
Across 500 cases and 9,128 recorded events, the median case duration was only 6 days.
The problem became visible when the team looked beyond the median. The 95th percentile reached 24.1 days, ten cases exceeded 30 days, and one remained open for more than 60 days.
Different service types also behaved very differently. One high-volume service had a P95 of only 11.9 days, while a lower-volume agreement process reached 45.7 days.
At first glance, performance looked relatively stable.
Across 500 cases and 9,128 recorded events, the median case duration was only 6 days.
The problem became visible when the team looked beyond the median. The 95th percentile reached 24.1 days, ten cases exceeded 30 days, and one remained open for more than 60 days.
Different service types also behaved very differently. One high-volume service had a P95 of only 11.9 days, while a lower-volume agreement process reached 45.7 days.
What We Did
Reconstructed the process across 27 process steps, four service types, 51 municipalities, and 75 actors.
Focused the analysis on provider handoffs, repeated rejection and review loops, document-package completion, service-type differences, and long-running exceptions.
The Outcome
Rather than redesigning the entire workflow, the analysis pointed toward a more focused operating model: service-specific SLAs, provider-response monitoring, rework reason codes, and regular review of long-tail cases.
The key insight was simple: improving predictability required managing the distribution of cases, not just the average.
The key insight was simple: improving predictability required managing the distribution of cases, not just the average.
Impact by the Numbers
- 500Cases
- 6.0Days Median
- 24.1Days P95
See the cases your averages are hiding
Alberta, Canada
© All Rights Reserved. Ampelus Inc.
CustomerSupport@ampelus.ca
CustomerSupport@ampelus.ca