Lead Time Analysis Report — February 2026

Executive Summary: PSI machine ship lead times have nearly doubled since 2022, from a 20-year average of ~180 days to ~330 days. This report identifies the specific stages, departments, and systemic factors driving the increase through data fusion across 11+ data sources spanning 9.6 million records.


The Problem in Numbers

MetricHistorical (2003-2021)Current (2023-2025)Change
Avg Lead Time180 days (6 months)328 days (11 months)+82%
Median Lead Time167 days316 days+89%
Avg Schedule Slip+1 day+17 days+16 days
On-Time Ship Rate78%49%-29 points
Late Projects Avg Slip21 days43 days+22 days

The inflection point was 2022. Lead time jumped +86 days (+49%) in a single year, then added another +66 days in 2023, and has plateaued at ~330 days through 2025.


Year-by-Year Lead Time Trend

Avg Lead Time (days)   Ship Year
350 |                                                    ████ ████ ████
325 |                                                    ████ ████ ████
300 |                                                    ████ ████ ████
275 |                                               ████ ████ ████ ████
250 |                                               ████ ████ ████ ████
225 |            ████                           ████ ████ ████ ████ ████
200 | ████ ████  ████      ████ ████ ████  ████ ████ ████ ████ ████ ████
175 | ████ ████  ████ ████ ████ ████ ████  ████ ████ ████ ████ ████ ████
150 | ████ ████  ████ ████ ████ ████ ████  ████ ████ ████
125 |
100 |
     2000  '05   '10   '15  '17  '18  '19  '20  '21  '22  '23  '24  '25
PeriodAvg DaysMedianProjects/Yr
2003-200717316160
2008-201216716759
2013-201718818164
2018-202119419477
2022-202531130577

Root Cause Breakdown: Where Are the Extra 150 Days?

The ~150-day increase in lead time is distributed across multiple stages. By fusing data from ganttJOB schedules, work order cycle times, labor hours, purchasing data, and quality metrics, the contribution of each area can be estimated:

Stage-by-Stage Contribution

StageEstimated ContributionEvidence Source
1. Purchasing / Long-Lead Items~45-55 days155.6 avg days in ganttJOB; vendor OTD dropped 81%→69%
2. Engineering (Mech + Controls)~30-40 days71-79% of projects past due; engineering is gate for everything
3. Electrical Assembly~25-35 days87 avg days late per WO; hours/job doubled (80→130)
4. Mechanical Assembly~20-30 days70 avg days late per WO; hours/job peaked at 77 (from 50)
5. Quality/Rework Cycles~15-20 daysECN volume, redbook resolution time, late-stage detection
6. Floor Space / WIP Queuing~10-15 daysBay constraints, projects waiting for space
Total Estimated~145-195 days

Detailed Findings by Area

1. PURCHASING — The Longest Phase (~155 days avg)

What the data shows:

  • Purchasing is the longest single phase at 155.6 days average in the ganttJOB schedule
  • 50.6% of active projects have purchasing past due
  • Stated vendor lead times are stable (~20 weeks) — the problem is not longer quotes

The real problem: Vendor on-time delivery is collapsing

YearVendor OTD RateLate Deliveries
202381.2%5,314
202475.2%6,335
202570.8%6,857
2026 (YTD)68.6%1,341

Worst-performing categories:

  • Machining subcontract (MClass 5500): 53.2% OTD — nearly half are late
  • Electrical components (MClass 1000): 60.4% OTD

Worst vendors by volume:

  • Kendall Electric: 73.4% late rate (4,417 late lines in 2024-2026)
  • Preferred Machine LLC: 50.1% late rate (3,121 late lines)

Root cause: Vendors are consistently missing promise dates. The gap between promised and actual delivery is growing. This is a vendor management and accountability issue, not a market lead time issue.


2. ENGINEERING — The Upstream Bottleneck

What the data shows:

Dept% of Projects Past DueAvg Duration (days)
Controls Engineering (110E)79.0%52.5
Mechanical Engineering (110M)71.6%59.5

Engineering is the gateway to everything downstream. When 79% of projects have Controls Engineering running late, it cascades through purchasing (can’t order electrical until designs are done), electrical assembly (can’t wire without drawings), and startup (can’t test without PLC programs).

Labor data shows engineering is NOT under-staffed in total:

  • Dept 110 labor hours have been stable at 47k-58k/year for 8 years
  • Headcount stable at 28-40 engineers
  • Hours per job are stable at ~185-240

The issue is likely:

  • Too many concurrent projects (81 active jobs scheduled)
  • Context switching across projects
  • Scope growth per project (more complex machines)
  • Waiting for customer inputs/approvals

3. ELECTRICAL ASSEMBLY (CC 108) — The Worst WO Performance

Across ALL historical work orders (Fabric dataset):

Dept (CC)Avg Days LateAvg DurationWO Count
108 (Electrical)87.1 days47.8 days26,309
106 (Assembly)70.1 days34.5 days119,832
104 (Weld/Fab)52.0 days20.4 days261,935
102 (Machining)46.1 days24.4 days64,925
1000 (Outside)38.4 days28.4 days44,807

CC 108 work orders average 87 days late — worst of any department.

Labor data confirms the strain:

YearHours/Job (108)Change from Baseline
201892.9baseline
201980.2-14%
202284.6-9%
2023130.9+41%
2024123.3+33%
2025108.4+17%

Hours per job in Electrical Assembly jumped 41% in 2023 and remain elevated. Meanwhile, headcount has barely changed (20-30 range). The department is doing significantly more work per machine with roughly the same staff.

Likely drivers:

  • More complex control systems (more I/O points, HMI screens, robot integrations)
  • Waiting on Controls Engineering drawings (79% late upstream)
  • Panel component delivery delays (Kendall Electric 73% late)
  • Rework from quality issues found during testing

4. MECHANICAL ASSEMBLY (CC 106) — Volume Surge Impact

WO data: 70 days late on average across 119,832 historical work orders.

Labor hour surge:

YearTotal Hours (106)Hours/JobHeadcount
201827,63152.841
202117,04830.125 (trough)
202342,61877.060
202438,51557.555
202526,34143.835

Dept 106 total hours surged 54% in 2023 vs the 2018 baseline. PSI staffed up to 60 people (from ~41), but hours per job still jumped to 77 (from ~50). This suggests machines are genuinely requiring more assembly labor, not just throughput constraints.

The 2025 return to 26k hours and 35 headcount suggests the surge has partially normalized, but per-job labor remains elevated vs pre-2020 levels.


5. QUALITY ISSUES — Compounding Effect

While quality issues don’t directly add 30+ days, they create rework cycles that compound delays at every stage:

Key quality metrics from redbook data:

  • 50,000+ quality tickets since 1999
  • Tickets opened late in the build (Late Build / At Ship) carry 1.5x-2.0x cost multipliers
  • 16 departments can be assigned per quality ticket, creating cross-department coordination overhead
  • Average 18 RFCs per project (based on Job 2399 sample data)
  • ECN generation from quality issues forces engineering back into completed projects

The compounding effect: A quality issue found during electrical assembly might require:

  1. Controls Engineering to redesign (they’re already 79% overdue)
  2. New parts to be ordered (purchasing is the longest phase)
  3. Rework of completed electrical work
  4. Re-testing of affected systems

Each individual issue may add only a few days, but with 18+ issues per project, the cumulative impact on critical-path activities is significant.


6. CAPACITY & WIP CONSTRAINTS

Current portfolio: 81 active jobs across 9 departments (per ganttJOB data)

Floor space: Projects require physical bay assignments in Building 1 and Building 5. When bays are full, new projects queue for space even if parts and labor are available.

Work center capacity (wc-capacity.csv): The 31-row static capacity file dramatically understates real capacity — stated Machine Shop capacity is 269 hrs/wk but actual Baseline median was 485 hrs/wk (see March 2026 Deep Dive below). Effective capacity must be measured empirically, not from this file.

40% of all scheduled tasks are past due (357 of 891 ganttJOB entries), indicating systemic over-scheduling relative to capacity.


Department Performance Summary (Current Active Portfolio)

From Fabric ganttJOB query — average % complete across all 81 active projects:

DepartmentAvg % CompleteStatus
Mech Engineering52.1%Furthest along but 71.6% late
Weld43.1%Moderate
Machine Shop37.3%Behind
Controls Engineering34.1%Critical — 79% of projects late
Electrical Assembly32.2%Behind — 87 avg days late on WOs
Mech Assembly22.4%Early stages for most projects
Startup/Runoff17.6%Expectedly low (late-stage)
ProposalN/ANo % complete tracked
PurchasingN/ANo % complete tracked

What Changed in 2022? (The Inflection Point)

Multiple factors converged in 2022:

  1. Post-COVID demand surge: After a trough in 2021 (69 projects shipped, ~17k-28k hrs/dept), demand rebounded sharply in 2022-2023 with more complex projects.

  2. Machine complexity is increasing: Hours per job in Electrical Assembly jumped from ~85 to ~130 (+53%). Mechanical Assembly jumped from ~50 to ~77 (+54%). Machines are requiring more labor, suggesting increasing technical complexity.

  3. Supply chain disruption persisted: Vendor OTD dropped from 81% to 69% between 2023-2026. While stated lead times normalized, actual delivery reliability continued to erode.

  4. Portfolio loading exceeded capacity: 81 active jobs simultaneously across a facility with fixed floor space and engineering bandwidth. With 79% of projects having late Controls Engineering, the system is overloaded upstream.

  5. No proportional capacity increase in Electrical Assembly: Despite 40% more hours per job, Dept 108 headcount remained flat at 20-30. This is a structural capacity gap.


Recommendations

Immediate Actions (0-3 months)

  1. Vendor accountability program for Kendall Electric and Preferred Machine LLC — These two vendors account for 7,500+ late deliveries. Implement scorecarding, escalation protocols, or alternate source qualification.

  2. Electrical Assembly staffing review — Hours/job have permanently shifted upward. Either add headcount or identify efficiency improvements (pre-built sub-assemblies, better kitting).

  3. Controls Engineering prioritization — With 79% of projects late, this department needs either capacity relief (contract resources) or better sequencing (stop starting new projects before finishing current ones).

Medium-Term (3-6 months)

  1. WIP limits and project sequencing — Consider reducing active projects from 81 to a sustainable level. Theory of Constraints principles suggest limiting WIP to match the bottleneck department’s capacity.

  2. Purchasing OTD tracking dashboard — Build real-time vendor delivery tracking into PSI Explorer or a new Power BI report using purchaseorders + pohist data.

  3. Engineering phase-gate discipline — Ensure mechanical and controls design are substantially complete before releasing to purchasing and production.

Strategic (6-12 months)

  1. Machine complexity analysis — Quantify whether newer machines genuinely require more assembly hours or if there are design efficiency opportunities (standardized panels, modular wiring).

  2. Capacity planning infrastructure — Formalize work center capacity planning beyond the 31-row wc-capacity file. Integrate schedule data with actual labor capacity.

  3. Predictive lead time model — Build a model using machine type, BOM complexity, and department loading to predict realistic lead times at order entry.


March 2026 Deep Dive — Capacity & Queue Analysis

Update (March 6, 2026): A manufacturing-consultant-style re-evaluation was performed using two previously untouched data sources (tslabor2 historical hours + wiprouteline operation-level routing) to measure actual utilization and queue/touch time decomposition. Full methodology and scripts at C:\git\schedule\GROUNDWORK.md.

Capacity & Utilization: wc-capacity.csv Is Fiction

The stated weekly capacity from wc-capacity.csv dramatically understates what PSI actually sustains. By computing “effective capacity” as the Baseline-era (2017-2019) median weekly hours from tslabor2.csv (2.77M timesheets), the true sustainable rate — the level at which PSI delivered 85% OTD — emerges:

DepartmentStated Cap (hrs/wk)Effective Cap (Baseline Median)Understated By
Machine Shop (102)26948545%
Weld/Fab (104)25043743%
Mech Assembly (106)40050621%
Elec Assembly (108)20023615%

The stated numbers likely reflect single-shift base with no overtime, while PSI chronically runs OT/flex staffing.

Utilization by era (vs effective capacity):

DepartmentBaselinePre-Overload (20-21)Overload (22-23)Current (24-25)
Mach Shop100%73%101%86%
Weld/Fab100%79%125%82%
Mech Assy100%75%141%115%
Elec Assy100%83%178%132%

Electrical Assembly was the breaking point — running at 178% of sustainable rate during Overload. Both Elec Assy (132%) and Mech Assy (115%) remain above sustainable in 2024-2025.

Machine Shop barely surged (100% → 101%) — the bottleneck was never in machining.

The Overload Surge in FTE Equivalents

PSI absorbed the equivalent of ~14 additional FTEs of demand during Overload without proportional headcount increases:

DepartmentBaseline → OverloadSurplus hrs/wkAnnual FTE
Weld/Fab437h → 547h (+25%)+1102.8
Mech Assy506h → 712h (+41%)+2075.2
Elec Assy236h → 419h (+78%)+1834.6
Controls Eng483h → 531h (+10%)+481.2

Controls Engineering: Fragmentation, Not Staffing

CE hours/person are rock-stable at ~36-38 hrs/wk across all eras. But projects per person jumped from 2.2 (Baseline) to 2.7 (Overload) — each engineer juggling 23% more concurrent projects with zero additional per-person output. This confirms the bottleneck is multitasking fragmentation, not total hours.

Queue Time vs Touch Time: The Shop Floor Is Fast

Analysis of 928K operation records from wiprouteline.csv reveals that median touch time is 0 days across all departments and all eras. 89% of operations complete the same day they start (saw cuts, CNC ops, welds take minutes to hours). Virtually 100% of operation elapsed time is queue time — parts waiting for a machine or operator.

Median queue days by department and era:

DepartmentBaseline (17-19)Overload (22-23)Current (24-25)
Mach Shop11d7d6d
Weld/Fab6d3d2d
Mech Assy17d24d23d
Elec Assy15d37d26d

Machine Shop and Weld queues actually improved over time. Mech Assembly and Elec Assembly queues grew dramatically during Overload and haven’t fully recovered — matching the utilization findings above.

The tail is where projects die: During Overload, Elec Assembly P90 queue = 120 days (one in ten electrical operations waited 4 months to start). P95 = 150 days.

The Key Insight: Shop Floor Queue ≠ Project Lateness

The most important finding: shop floor queue time does not correlate with project-level DaysLate (Pearson r = -0.006, N=356). On-time projects actually have slightly higher median queue times than late ones (8d vs 6d).

This means the binding constraint is upstream of the shop floor — engineering release timing, purchasing delays, and CE fragmentation. wiprouteline.csv only covers departments 102/104/106/108 (shop floor), not 110E/110M (engineering). The invisible queue at the engineer’s desk, not the visible queue at the machine, determines whether projects ship on time.

Implication: Shop floor flow management helps Elec Assembly specifically (the only dept where late projects queue longer), but the primary lever for OTD improvement is controlling engineering release cadence — CE project limits, staggered release, and the TOC “subordinate” principle.

Scripts & Source Data

ScriptPurpose
C:\git\schedule\capacity_analysis.pyHistorical utilization from tslabor2 + forward-looking load from capacity.csv
C:\git\schedule\analyze_wiprouteline.pyQueue/touch time decomposition from 928K operation records
C:\git\schedule\GROUNDWORK.mdFull groundwork assessment with analytical roadmap

Data Sources Used

SourceRecords AnalyzedKey Contribution
Project1287List.xml1,951 projects25-year lead time trend (order→ship)
ganttJOB.csv891 tasks / 81 projectsDepartment-level schedule, % complete, lateness
openwo.csv519,286 work ordersWO cycle times, days late by cost center
Fabric ProgressiveDataSet9.6M rows / 33 tablesCross-table aggregations, dept completion rates
tslabor2.csv2,773,043 timesheetsLabor hours by dept, headcount trends
purchaseorders.csv108,768 PO linesVendor lead times, on-time delivery rates
pohist.csv268,238 receiptsActual vendor delivery dates
redbook.csv50,057 quality ticketsQuality issue volume, resolution times
Schedule Excel files7 department workbooksPMO priorities, deviation, hours-to-go
wc-capacity.csv31 work center recordsAvailable capacity by department (static, understates — see deep dive)
wiprouteline.csv927,344 operationsQueue time vs touch time per routed operation (March 2026)
capacity.csv23,834 planned load rowsForward-looking dept load by week (March 2026)
comprehensive_dataset.csv2,569 projects × 216 colsFused project-level dataset for correlation analysis (March 2026)
PROJECT.HRS.128781 active projectsBudget vs actual hours by department


Initial analysis: February 26, 2026 Deep dive update: March 6, 2026 (capacity/utilization + queue/touch time) Data as of: February 25-26, 2026 (initial); March 5-6, 2026 (deep dive) Analyst: Claude Code (data fusion across 14 PSI data sources)