Semiconductor equipment
Gas-delivery manifolds, aluminum gas-panel bases, chamber cooling plates, cleanroom subassemblies
- Share of shipments
- 64%
- Engineering changes in 6 months
- 37
- Machining time vs estimate
- 1.12×
Case study · Sample site assessment
Why a plant at record output kept shipping late
Data-center work grew from 12% to 36% of a precision hardware plant's shipments in just over a year. Every department reported record output while on-time delivery fell 28 points. Our site assessment traced the loss to one vacuum furnace and showed how to recover it without a capital request.
Sample deliverable. This case study and report are a composite. Company names, people, locations, and figures have been changed to protect client confidentiality.
MonTueWedThuFriSatSun12 am6 am12 pm6 pm
Show data
| Category | Hours |
|---|---|
| Effective | 56 |
| Empty fixture slots | 40 |
| Second braze cycle | 5 |
| Waiting for a load | 11 |
| Changeover or maintenance | 16 |
| Unstaffed | 40 |
- On-time in-full
- 86%to58%
- Over 13 months, while machining hours rose to 112% of target
- Annual cost of the constraint
- $5.0M
- Order-of-magnitude estimate from operating data
- Furnace output in 90 days
- +48%
- From scheduling, staffing, and fixtures, with no capital request
The situation
A high-mix plant hit by an AI-driven surge
Ridgeline built its business on semiconductor equipment parts. Then liquid-cooling hardware for AI data centers tripled its share of shipments in 13 months, and both product lines ran through the same few resources.
Shared by both lines
- Furnace 2The only furnace long enough for data-center parts
- CMM labFirst-article inspections up from 41 to 128 a month
- Precision cleaningCleaned parts wait up to three days for braze
- PlannersOne due-date schedule for two very different businesses
Data-center liquid cooling
Copper cold plates, coolant manifolds, coolant distribution subassemblies
- Share of shipments
- 36%from 12%
- Engineering changes in 6 months
- 214
- Machining time vs estimate
- 1.34×
The brief
Why is delivery falling while every department reports record output?
Asked by the business unit's VP of Operations, with three follow-on questions:
- Where is the throughput constraint, and how severe is it?
- Can today's staffing, shifts, and equipment carry the data-center growth plan?
- What should the plant fix now, and what should it build toward?
Plant profile
- Facility
- 215,000 sq ft
- People
- 440
- Shipments
- $9.7M a month
- Operating pattern
- 24/5 plus Saturdays
- Our time on site
- 2.5 days, all shifts
What we found
Busy everywhere, stuck at one furnace
Four views of the same plant, from the delivery numbers back to the cause.
1 of 4
Busy and late
Machining hours climbed to 112% of target while OTIF fell from 86% to 58%. Every report the plant ran measured activity, so the loss of flow at the constraint went unseen for most of a year.
OTIF fell from 86% to 58% while machining hours rose to 112% of target
Daily metrics workbook. Values adjusted for confidentiality.
2 of 4
Where the work piles up
WIP shows where work waits. A quarter of the plant's $14M sits in front of the braze cell, the largest share of any area and the clearest sign of the primary constraint. CMM inspection and outside plating queue up behind it.
The braze cell holds 24% of plant WIP, more than any other area
WIP snapshot, September 14, 2026. Values adjusted for confidentiality.
3 of 4
Where the furnace's week goes
Furnace 2 is the only furnace long enough for data-center parts. Only 56 of its 168 weekly hours produce effective loads. The rest go to unstaffed weekends, changeovers, waiting, and loads run partly empty to rush a late order.
Only 56 of Furnace 2's 168 weekly hours produce effective loads
Furnace load log, 13 weeks. Values adjusted for confidentiality.
4 of 4
What feeds it
New data-center parts run far over their machining estimates, and 61% of their engineering changes reach the floor after material is released. Parts arrive at the furnace late and in bursts, and every burst becomes another partial load.
31% of new data-center CNC jobs run at 1.5× estimate or worse, against 7% of mature parts
2,416 CNC operation records, June to August 2026. Values adjusted for confidentiality.
OTIF fell from 86% to 58% while machining hours rose to 112% of target
Monthly, September 2025 to September 2026
Daily metrics workbook. Values adjusted for confidentiality.
Try the recovery
Change how the furnace runs and watch the queue
A simplified version of the capacity model in the report. Start from today's operation, then try the Phase 1 plan.
Today the furnace sits idle from Saturday afternoon to Monday morning.
Each switch between copper and stainless runs costs about 2.4 hours.
Share of fixture slots used. Rushed orders run partly empty.
A part that fails leak test goes back through the furnace for a second cycle.
The next load is staged before the current cycle ends.
Effective loads per week
An effective load is one full furnace load that passes first time. A partial load counts by its fill.
5.2against demand of 6.4
Off track. The braze queue grows by 1.2 loads every week.
Where the furnace's 168 hours go
- Effective
- 55.5 h
- Empty fixture slots
- 40.5 h
- Second braze cycles
- 4.7 h
- Waiting for a load
- 11.2 h
- Changeovers
- 10.1 h
- Maintenance
- 6.0 h
- Unstaffed
- 40.0 h
- Output vs today
- 0%
- Added shipments once the queue clears
- $0 a week
Assumes a 10.6-hour cycle, 6 hours of weekly maintenance, 2.4 hours per changeover, a current braze queue of about 14 effective loads, and about $280K of shipments per effective load. Added shipments count output up to demand of 6.4 loads; capacity above that clears the queue faster. Your plant's model is built from your own furnace data.
The plan
Recover first, then fix what caused it
Each phase has a gate the plant must pass before the next one starts.
- Run Furnace 2 in campaigns by thermal profile, with a 72% minimum fill rule
- Dedicated load-building crew, a Saturday second shift, and a third-shift lead
- Release machining to the furnace schedule and hold a two-day buffer of ready loads
- Second-shift CMM coverage and first-article triage by change class
- Close $1.91M of completed-but-open jobs and fix scan compliance
- Tier meetings and leader standard work on every shift
Gate to Phase 2
Furnace 2 sustains 7 or more effective loads a week for four weeks, and OTIF is measured against the first committed date.
The deliverable
What the 23-page report covers
Every site assessment ends with a report like this one: the evidence, the constraint model, a 90-day plan with owners, and an 18-month roadmap.
- Executive summaryThe finding, the cost, and what to do first, in two pages.
- Background and contextThe business, the site, the systems landscape, and how we gathered evidence.
- Current stateDelivery, the constraint map, the furnace model, upstream drivers, quality, staffing, and customer risk.
- 90-day action planEvery action with an owner, a timeline, and an expected outcome.
- 3 to 18 month initiativesMetrics, change control, time standards, quality systems, and the capital case.
- Implementation roadmapThree phases, five workstreams, and the gates between them.
- AppendicesData sources, methodology, metric specifications, and financial impact.
How a site assessment runs
About five to six weeks, start to finish
These are the Study and Outline stages of our S.O.L.V.E. framework.
- Step 1 · 2 weeks
Before the visit
You send a standard data request: daily metrics, a WIP snapshot, operation time records, and your constraint's logs. We build a first constraint hypothesis and an interview plan.
- Step 2 · 2 to 3 days
On site
We walk material flow end to end on every shift, time the constraint, sit in on tier meetings, and interview 20 or more people, from leadership to operators.
- Step 3 · 3 weeks
After the visit
We quantify each loss, build the capacity and financial models, and deliver the report with a 90-day plan and owners.
Next step
Find out what is setting the pace in your plant
Sample deliverable. A composite, with names and figures changed to protect client confidentiality.