PhoneTech Assembly Plant - Process Mining Dashboard

Total Cases Processed

4,218

+8.3% vs. prior period

Avg Throughput Time

6.4 hrs

Target: 6.0 hrs | +6.7% over

On-Time Completion Rate

81.2%

Target: 90% | below target

Rework Rate

12.4%

Target: <8% | 523 cases reworked

Active Cases (In Progress)

Capacity: 120 | 72.5% utilization

R Code

library(bupaverse)

# Load the event log
phone_log <- read_xes("phonetech_assembly.xes")

# Key metrics
phone_log %>% n_cases()                                  # Total cases
phone_log %>% throughput_time("log", units = "hours")  # Avg throughput
phone_log %>% number_of_repetitions("log")              # Rework rate
phone_log %>% filter_processing_time(
    interval = lubridate::dhours(0), lubridate::dhours(6)
) %>% n_cases() / n_cases(phone_log)                   # On-time rate

Process Map — Frequency (Absolute)

Fast / On Target

Normal

Bottleneck / Slow

Nodes show activity frequency (absolute & relative to total cases). Dashed red lines indicate rework loops. QC frequency >100% due to rework re-entries. SW Flashing is the primary bottleneck (avg 68 min, target 45 min).

R Code

# Frequency process map (absolute counts)
phone_log %>% process_map(frequency("absolute"))

# Performance process map (mean processing time)
phone_log %>% process_map(performance(mean, "mins"))

# Combined: node = relative-case frequency, edge = performance
phone_log %>% process_map(
    type_nodes = frequency("relative-case"),
    type_edges = performance(mean, "mins"),
    rankdir    = "LR"
)

# Export
phone_log %>% process_map(render = FALSE) %>% export_map("process_map.pdf")

Throughput Time Analysis

Overall Throughput (hours)

Actual: 6.4h

Target: 6.0h

By Product Variant

Standard 5.2h (target: 5.5h)

Pro 6.8h (target: 6.5h)

Ultra 8.1h (target: 7.5h)

Throughput Trend (Last 24 Weeks)

Sep '25Nov '25Jan '26Feb '26

R Code

# Throughput time at log level
phone_log %>% throughput_time("log", units = "hours")

# By product variant (case-level attribute)
phone_log %>%
    group_by(product_variant) %>%
    throughput_time("log", units = "hours")

# Plot throughput time distribution
phone_log %>% throughput_time("log") %>% plot()

# Weekly trend (augment + aggregate)
phone_log %>%
    throughput_time(level = "case", units = "hours") %>%
    mutate(week = lubridate::floor_date(first_event, "week")) %>%
    group_by(week) %>%
    summarise(avg_throughput = mean(throughput_time))

Activity Performance

Activity	Freq	Rel %	Avg Proc	Avg Idle	Util %
PCB Assembly	4,218	100%	38 min	3 min	88%
Display Install	4,218	100%	52 min	8 min	91%
Battery & Power	4,218	100%	29 min	5 min	79%
Housing Assy	4,218	100%	44 min	6 min	84%
SW Flashing	4,512	107%	68 min	14 min	97%
Quality Control	4,741	112%	85 min	22 min	95%

Bullet graphs: dark bar = actual processing time; black marker = target. Gray bands = poor | satisfactory | good ranges. = exceeds target threshold.

R Code

# Activity frequency
phone_log %>% activity_frequency("activity")

# Processing time per activity
phone_log %>% processing_time("activity", units = "mins")

# Idle time per activity
phone_log %>% idle_time("resource", units = "mins")

# Resource frequency per activity
phone_log %>% resource_frequency("resource-activity")

# Performance process map
phone_log %>% process_map(performance(mean, "mins"))

Trace Analysis — Top Variants

PCB

Display

Battery

Housing

SW Flash

#	Trace (Activity Sequence)	Cases	Rel %	Cumul %	Note
1	P D B H S Q	2,847	67.5%	67.5%	Happy path
2	P D B H S Q S Q	548	13.0%	80.5%	SW rework x1
3	P D B H S Q D S Q	312	7.4%	87.9%	Display rework
4	P D B H S Q S Q S Q	189	4.5%	92.4%	SW rework x2
5	P D D B H S Q	142	3.4%	95.7%	Display self-loop
6	P D B B H S Q	98	2.3%	98.0%	Battery self-loop
7	Other (14 variants)	82	2.0%	100%	Rare paths

P=PCB, D=Display, B=Battery, H=Housing, S=SW Flash, Q=QC. Faded segments = rework iterations. 6 variants cover 98% of all cases.

R Code

# Trace variants with frequencies
phone_log %>% traces()

# Trace coverage (cumulative)
phone_log %>% trace_coverage("trace") %>% plot()

# Visual trace explorer (top 80% coverage)
phone_log %>% trace_explorer(coverage = 0.80)

# Detect rework: repetitions and self-loops
phone_log %>% number_of_repetitions("activity")
phone_log %>% number_of_selfloops("activity")

# Filter only traces with rework
phone_log %>%
    filter_activity_frequency(interval = c(2, Inf))

Resource Utilization — Workstations

PCB Line (WS-01..03)

88%

Display Line (WS-04..06)

91%

Battery Station (WS-07..08)

79%

Housing Line (WS-09..11)

84%

SW Flash (WS-12..13)

97%

QC Bay (WS-14..16)

95%

Handover-of-Work Matrix (Frequency)

	PCB	Display	Battery	Housing	SW Flash	QC
PCB	-	4,218	0	0	0	0
Display	0	-	4,218	0	0	0
Battery	0	0	-	4,218	0	0
Housing	0	0	0	-	4,218	0
SW Flash	0	0	0	0	-	4,512
QC	0	229	0	0	294	-

Matrix shows handover count from row activity to column activity. QC row shows rework flows back to Display (229) and SW Flash (294).

R Code

# Resource frequency
phone_log %>% resource_frequency("resource")

# Resource involvement per case
phone_log %>% resource_involvement("resource")

# Resource specialisation
phone_log %>% resource_specialisation("resource")

# Handover-of-work network (visual)
phone_log %>% resource_map()

# Handover-of-work matrix (tabular)
phone_log %>% resource_matrix() %>% plot()

# Process matrix (directly-follows)
phone_log %>% process_matrix(frequency("absolute")) %>% plot()

Conformance & Quality — Rule Checking

Process Rule Compliance

	Rule	bupaR Function	Conformant	Violations
✓	All cases start with PCB Assembly	starts("PCB")	4,218 (100%)	0
✓	All cases end with Quality Control	ends("QC")	4,218 (100%)	0
✓	PCB always precedes Display Install	precedence("PCB","DI")	4,218 (100%)	0
✓	Battery always followed by Housing	response("Bat","Hous")	4,218 (100%)	0
✗	SW Flashing occurs at most once	contains_exactly("SW",1)	3,481 (82.5%)	737 (17.5%)
✗	QC occurs at most once	contains_exactly("QC",1)	3,695 (87.6%)	523 (12.4%)
✗	Throughput time < 8 hours	filter_throughput_time()	3,426 (81.2%)	792 (18.8%)

Rework Detection Summary

Rework Type	Cases Affected	Rate	Trend (24 wk)
Repetitions (rework loop)	1,049	24.9%
Self-loops (immediate repeat)	240	5.7%

Bottleneck Cases (Top Offenders)

Case ID	Throughput	Activities	Bottleneck Activity	Wait Time
CASE-3847	14.2 hrs	11	QC (x3 iterations)	4.1 hrs
CASE-2156	12.8 hrs	10	SW Flashing (x3)	3.6 hrs
CASE-0891	11.5 hrs	9	Display + SW rework	2.9 hrs
CASE-4102	11.1 hrs	9	QC (x2) + wait	3.3 hrs

R Code

library(processcheckR)

# Check multiple rules at once
phone_log %>% check_rules(
    rule1 = starts("PCB Assembly"),
    rule2 = ends("Quality Control"),
    rule3 = precedence("PCB Assembly", "Display Install"),
    rule4 = response("Battery & Power", "Housing Assembly"),
    rule5 = contains_exactly("SW Flashing", 1),
    rule6 = contains_exactly("Quality Control", 1)
)

# Rework detection
phone_log %>% number_of_repetitions("activity", type = "all")
phone_log %>% number_of_selfloops("activity", type = "all")
phone_log %>% size_of_repetitions("activity")

# Filter non-conformant cases
phone_log %>% filter_rules(contains_exactly("SW Flashing", 1), reverse = TRUE)

# Bottleneck cases (longest throughput)
phone_log %>%
    throughput_time(level = "case", units = "hours") %>%
    arrange(desc(throughput_time)) %>%
    head(10)

Dotted Chart — Case Timeline (Sample: 30 Cases)

PCB

Display

Battery

Housing

SW Flash

0h 3h 6h 9h 12h+

Sorted by case duration (shortest first). X-axis = relative time since case start. Red-labeled cases exceed the 6-hour target. Faded dots indicate rework iterations. Cases C-0015 and C-0011 are clear outliers with multiple rework cycles.

R Code

# Dotted chart (relative time, sorted by duration)
phone_log %>% dotted_chart(
    x     = "relative",
    sort  = "duration",
    color = activity_id(phone_log)
)

# Absolute time view
phone_log %>% dotted_chart(x = "absolute", sort = "start")

# Intra-day pattern analysis
phone_log %>% dotted_chart(x = "relative_day")

# Performance spectrum (detailed)
phone_log %>% ps_detailed()
phone_log %>% ps_aggregated(bins = 30)

Appendix: Event Log Setup in R (bupaR)

The code below shows how to create the PhoneTech assembly event log from raw factory data using bupaR's activitylog format. This log structure powers all the analyses shown above.

library(bupaverse)
library(dplyr)
library(lubridate)

# --- 1. Read raw factory MES data ---
raw_data <- read.csv("phonetech_mes_export.csv") %>%
    mutate(
        start    = ymd_hms(start_timestamp),
        complete = ymd_hms(complete_timestamp)
    )

# --- 2. Create bupaR activity log ---
phone_log <- raw_data %>%
    activitylog(
        case_id     = "case_id",
        activity_id = "activity",
        resource_id = "workstation",
        timestamps  = c("start", "complete")
    )

# --- 3. Validate and inspect ---
summary(phone_log)
n_cases(phone_log)       # 4,218
n_activities(phone_log)  # 6
n_resources(phone_log)   # 16 workstations
n_traces(phone_log)      # 20 unique trace variants

# --- 4. Enrich with product variant ---
phone_log <- phone_log %>%
    left_join(product_info, by = "case_id") %>%
    re_map(mapping(phone_log))

# --- 5. Data quality checks ---
library(daqapo)
phone_log %>% detect_time_anomalies()       # Negative durations?
phone_log %>% detect_incomplete_cases(
    activities = c("PCB Assembly", "Quality Control")
)
phone_log %>% detect_duration_outliers(
    "SW Flashing", sd_threshold = 3
)

PhoneTech Assembly Plant — Process Mining Dashboard | Built with bupaR (bupaverse) | Data: Fictional | Design: Stephen Few principles