LAUNCH BOOKST 000 · COVERCS-11 · SHEET 1 OF 7

Battery-module assembly line · launch year

Pragmatos case study CS-11 · industry archetype · one module serial from cell lot to pack · every result modelled

Case study CS-11 · automotive e-drive and battery-module assembly · industry archetype

One module serial, read from cell lot to pack, with the launch clock running

A battery-module line in a DACH plant is nine weeks past start of production. Five automated stations, an end-of-line test bench, three shifts, a new module generation, and an OEM launch team that reviews units per shift every Thursday. This launch book follows one module serial through the line and shows what the plant can read, record and decide when every station and the test bench report into one record, and what that is worth in the plant's own quantities: first-pass yield at the bench, weeks from SOP to rate, seconds against takt, minutes per genealogy lookup, serials inside a containment scope.

This is an industry archetype, not a customer. The plant, the line and every figure are modelled from how such lines run; the inputs sit on sheet 7 where a reader can replace them with their own. Pragmatos reads the line; it writes nothing to a station, a bench or a controller.

Line status board · launch week 9

Stations

010 screwing · 020 press-fit · 030 dispensing · 040 laser welding · 050 EOL test

Takt

100 s

plan at rate · bottleneck 040 runs at 104 s

Units per shift

142

plan at rate 200 · three shifts · week 9 of 26 on the ramp plan

FPY at EOL

85 %

first pass through the bench · the rest reworked or scrapped

Record depth

cell lot → module serial → pack serial · retained life of part + 1 year

Who reads this book, and on what number

Plant Manager	units per shift against the ramp-up plan, OEE per line and shift, for the Thursday launch review with the OEM.
Head of Quality	first-pass yield at the bench, ppm at the customer, open findings, for the IATF 16949 surveillance audit and the OEM's VDA 6.3 process audit.
Ramp-up lead	weeks from SOP to rate, station cycle time against takt, and the day a field finding arrives and the containment scope has to be named in serials within the 24 hours the 8D's D3 step allows.

What the module is

Cells arrive in supplier lots. At 010 the carrier gets its serial as a data-matrix code and the cell-lot IDs are married to it. 020 presses the contacts, 030 dispenses the thermal adhesive, 040 welds cells to the copper busbar, 050 tests insulation, open-circuit voltage, contact resistance and leak. Packs downstream take the module serial into their own record.

Pragmatos Live plant overview on a monitor in the launch office: machine state counts, power now, energy today, lines with their running machines, open faults with age

Pragmatos Live on the launch-office monitor: machine states across the plant, power and energy today, every line with how many of its machines run, open faults by age. The demo plant on the screen is the product's own sample data, shown as recorded at 1×.

GENEALOGYCELL LOTMODULE SERIALPACK SERIAL

Crius Software · Pragmatos · launch book CS-11 · rev A 2026-11

1 / 7

LAUNCH BOOKST 010–050CS-11 · SHEET 2 OF 7

Station walk · what each station reports

and the read path Pragmatos uses for it · the line keeps its controllers, its bench software and its parameters

010–050Five stations, one bench, one record per serial

The walk follows the module in build order. For each station: what it measures, what Pragmatos reads from it, and which part of the platform does the work. Every read is one-directional: Loom, the connection software on the plant's own server, polls or receives; it never writes a parameter, a setpoint or a result back. A supervisor approves each connection before it carries data, and a station that misses three polls is shown as stale rather than assumed fine.

010SCREWING

Carrier serial is born; cell lots are married to it

The station scans the carrier's data-matrix code, scans the cell-lot labels of the cells loaded, and records torque and angle for each screw position. Loom reads the station's cell controller over OPC UA and takes the lot-marriage record as the station's CSV export. Pragmatos Live opens the part history for this serial: the record every later station reports into, kept for the life of the part plus one year.

READ PATH · OPC UA (read) · CSV export · store-and-forward if the network drops

020PRESS-FIT

Force–displacement curve per contact

Peak force and press-in depth per contact are the characteristics the OEM's control plan names. Pulse keeps a control chart per characteristic with run rules (AIAG/VDA 7-point, Western Electric 8, Nelson 9) and shows Cpk against the plan's limit (1,33 running, 1,67 at PPAP). A characteristic that trips a rule is flagged "process not stable" on the chart; the station is not touched.

READ PATH · S7 (read) from the press controller · one sample per contact

030DISPENSING

Bead weight, adhesive lot, pot-life clock

Dispensed mass per module and the adhesive lot in the dispenser go into the part history, so the genealogy later answers "which modules carry adhesive lot X". Orbis watches the bead-weight series and raises a drift card when the trend leaves the band; the card says what it saw and how much of the series it saw (the confidence value is a data-coverage figure, not a defect probability).

READ PATH · Modbus TCP (read) from the dispense controller · adhesive lot by barcode scan

040LASER WELDING

Weld result per busbar joint, cycle time against takt

The welding cell reports OK/NOK per joint from its own seam monitor and the cycle time per module. Pragmatos Live shows OEE per station and shift (ISO 22400-2) and the station-to-takt comparison on the line screen: 040 is the one that runs at 104 s against a 100 s takt. The ramp-up lead sees the loss in seconds per module, per shift, before the Thursday review.

READ PATH · OPC UA (read) from the cell PLC · cycle-time stamps per serial

050EOL TEST

Insulation, open-circuit voltage, contact resistance, leak

The bench writes a result file per serial: four test steps, pass/fail, measured values, an NOK code. Loom takes the file over the bench's REST export; first-pass yield per shift and per NOK code appears in Pragmatos Live. When Orbis finds an out-of-control pattern across serials, Decisions puts a proposed containment scope in the quality engineer's queue (a list of serials, nothing more), and Flow carries the hold, the rework order and the release once two signatures are on it.

READ PATH · REST export from the bench PC · CSV fallback · results per serial

Where the software runs

On premise, in the plant's own cloud tenancy, or both. Loom is software (a service on the plant server or a VM near the line), not hardware on the line. No station parameter changes because Pragmatos is present.

Who approves what

Every connection: a supervisor. Every hold: one named role. Every release: two signatures. Approval is never granted by elapsed time; a proposal that nobody answers stays a proposal.

What the record keeps

Cell lots, adhesive lot, every station result, bench values, who held and who released, with timestamps: the per-serial record an IATF 16949 auditor asks to see, and the one the 8D's D3 step needs within 24 hours.

GENEALOGYCELL LOTMODULE SERIALPACK SERIAL

Crius Software · Pragmatos · launch book CS-11

2 / 7

LAUNCH BOOKST 050 · NOKCS-11 · SHEET 3 OF 7

NOK records · what the launch is failing on

five fail records in the format of a bench printout · each with its public source · the ramp-up curve in units per shift

NOKThe line fails on records, not on welds

STEP 01EOL · insulation / leak · first passNOK-YIELD

ResultFPY 85 % at the bench in week 9. The fail reason is on the bench PC; the rework decision is made at shift end from a printed list.

SourceFraunhofer FFB on battery production ramp-up: scrap rates of 15 to 30 % in the first years, around 10 % after five years, and each percentage point of scrap costing about €30,000 a day in cell production ^S1. Module assembly scrap is lower than cell scrap; the shape of the curve is the point.

STEP 02Launch review · units per shiftNOK-RATE

Result142 of 200 units per shift. The Thursday number is counted from the bench log into a spreadsheet; which station lost the seconds is argued, not read.

SourceSame ramp-up study: a new battery line takes years, not months, to reach planned output and yield ^S1; the OEM's ramp plan on this line allows 26 weeks.

STEP 03Field finding · containment scopeNOK-GENEALOGY

ResultA field finding names one cell lot. Cell lot → module serial → pack serial sits in three systems; a lookup takes 45 minutes per serial, and the scope stays "all modules of that month" for three days.

Source532 vehicle recalls in Germany in 2025 against 185 in 2010; 2,84 million vehicles affected in 2024, +48 % on the year ^S2. One PHEV battery recall: about €400 million for about 30,000 vehicles ^S6; an EV battery recall in autumn 2026 reaches almost 100,000 vehicles ^S7. The KBA's recall process starts from the manufacturer's own scope definition ^S3.

STEP 04OEM 8D · D3 containment within 24 hNOK-RECORD

ResultD3 evidence (which serials, which lots, which stations) is assembled by hand from four exports. Audit preparation runs 24 hours per audit, four times a year.

SourceVDA's 8D method sets containment (D3) as an immediate step with the scope documented and the affected parts identified ^S4.

STEP 05Battery passport · data per moduleNOK-DATA

ResultFrom 18 February 2027 each battery above 2 kWh placed on the EU market needs a battery passport. The module's manufacturing data per serial has to exist before anyone can fill one.

SourceRegulation (EU) 2023/1542, battery-passport articles, as summarised by the Battery Pass consortium ^S5. Pragmatos holds the per-serial manufacturing record; the passport document itself is on the product outlook, not in the product today.

Ramp-up curve · units per shift · weeks from SOP

Both curves are the model's inputs (sheet 7). The plan line is the OEM's 26-week ramp as the plant has it; the modelled line is the same shape four weeks shorter, which is what sheet 5 asks the line to do. The hatched area is counted once, at the module margin of C-12.

Tablet at an assembly station of a sample automotive plant showing a Pragmatos control chart with the Cpk value and run-rule flags; the station and a worker in the background

The same machine record, three tabs: the control chart with its run rules is Pulse, the containment proposal under it is Decisions, the hold-rework-release queue is Flow. A quality engineer reads all three at the station on one screen. Demonstration plant.

What the five records have in common

Each is a record fail: the measurement exists at the station or the bench, but not against the serial, not in one place, not within the plan's time. The platform holds the record, shows it where the three buyers already look, and puts proposals, never actions, in front of the person who decides.

GENEALOGYCELL LOTMODULE SERIALPACK SERIAL

Crius Software · Pragmatos · launch book CS-11

3 / 7

LAUNCH BOOKONE RECORDCS-11 · SHEET 4 OF 7

One record · the line as Pragmatos reads it

five read taps, one part history per serial, the genealogy a field finding walks, and the three words a quality engineer sets

RECThe U-line, its read taps and the record they fill

What a field finding looks like on this record

A cell-lot number arrives from the OEM's warranty desk at 09:10. The quality engineer opens the lot in Pragmatos Live: the genealogy returns every module serial built with it, the station results each one has, and the pack serials they went into: one screen, under five minutes, instead of 45 minutes per serial across three systems. The scope is 120 serials from two shifts, not the month's 2,400.

The three words

Sperren	one named role sets the hold on the 120 serials; Flow opens a quality lot hold (held / inspect / rework / scrap) and the work orders that follow it.
Nachmessen	re-measurement at 050 for the held serials; the bench values land against each serial in the same record.
Freigeben	two signatures release; nothing releases on elapsed time. The release, who signed and when, stays in the part history for the life of the part plus one year.

D3 of the 8D (containment scope, affected serials, where they are) is a print of this screen, inside the 24 hours the OEM allows. Record completeness on the line screen shows what share of serials carry a complete record; a serial with a missing station result is visible as such, not assumed complete.

Monitor in a quality office showing the Pragmatos records screen of a sample automotive plant: proof records and inspection deadlines by record type, with overdue and due-this-week counts and the responsible role per record

Quality office, OEM audit week. The records screen lists each proof type the plant has to show (IATF 16949 internal audits, layered process audits, VDA 19.1 cleanliness, CQI-9 heat treatment, DGUV inspections), with its interval, the responsible role, what is overdue and what falls due this week. The plant on the screen is the product's own sample data, shown as recorded at 1×. The platform keeps the records an audit asks for; a competent person performs the inspection and the auditor forms the judgement.

Containment scope · serials per event

600 → 120

Today the scope is "all modules of that month": 600 serials. With cell lot → module serial in one record it is the modules that carry the named lot: 120. C-20, C-21 · R-05 · MODELLED

Genealogy lookup · minutes per serial

45 → 5

Today three systems and a spreadsheet join; modelled, the part history on one screen in Pragmatos Live. 300 lookups a year at the quality engineer's loaded rate. C-16, C-17, C-18 · R-04 · MODELLED

Audit preparation · hours per audit

24 → 8

Today four exports joined by hand; modelled, the records screen and the part history printed per sample. Four audits a year: OEM process audit, IATF surveillance, two customer visits. C-19 · R-04 · MODELLED

GENEALOGYCELL LOTMODULE SERIALPACK SERIAL

Crius Software · Pragmatos · launch book CS-11

4 / 7

LAUNCH BOOKSTAIRCASECS-11 · SHEET 5 OF 7

Launch staircase · five treads to rate

the rollout drawn as the maturity levels a ramp-up lead already uses · weeks on the risers · what is live on each tread

L1–L5Each tread is checkable on a screen before the next one starts

The staircase runs inside the OEM's 26-week ramp plan. No tread changes a station; each one adds a read, a record or a decision path, and each is approved by the plant's own people before the line leans on it. The modelled rate week is 22 (C-10); the plan's is 26 (C-09).

Who signs each tread

L1: the automation supervisor, per connection. L2: the launch lead, when the record is complete for a full shift. L3: the quality engineer, per characteristic. L4: the head of quality, for the hold and release roles. L5: the plant manager, for the ERP hand-over.

What stays with the plant

Controllers, bench software, parameters, the control plan, the people who act. On premise, in the plant's own cloud tenancy, or both; the record is kept for the life of the part plus one year, up to 15 years for this product class.

What is not on the staircase

No setpoint, no station parameter, no bench limit is ever written by the platform. The battery-passport document for 2027 is not a tread: the per-serial data it needs is, the document itself is on the product outlook.

What the Thursday launch review reads at each tread

Tread	Number on the screen	Who reads it	Module
L1 · Read	Five connections, each live or stale; last poll per station	Automation supervisor	Loom
L2 · Record	Record completeness per shift: share of serials with every station result present	Launch lead	Pragmatos Live
L3 · Chart	Cpk per control-plan characteristic; run-rule flags per shift; drift cards with coverage	Quality engineer	Pulse · Orbis
L4 · Decide	Open proposals; serials under hold; hours from proposal to two signatures	Head of Quality	Decisions
L5 · Hand over	Units per shift against the ramp plan; OEE per station against takt; 8D steps open	Plant Manager	Pragmatos Live · Flow

GENEALOGYCELL LOTMODULE SERIALPACK SERIAL

Crius Software · Pragmatos · launch book CS-11

5 / 7

LAUNCH BOOKTAKT BOARDCS-11 · SHEET 6 OF 7

Takt board · six results, today against modelled

one column pair per result, both bars from zero in the result's own unit · the better bar touches the takt line · € a year at the foot

R-01–06Every bar is a modelled figure; its inputs are on sheet 7

€2 685 060

a year, the six results summed once each · MODELLED · the weakest is R-02: it rests on a four-week shortening of a ramp curve nobody has measured on this line yet; at half the shortening it is €252 000, at double €1 008 000.

How to read the board

Each pair is scaled to its own unit, so bar heights compare today with modelled inside a pair, never across pairs. The € figures do compare: R-01 and R-02 carry two thirds of the total; R-04 is the smallest and the surest, because it is hours of a named person. R-06 is the field exposure per event multiplied by a quarter of an event a year (C-24): the one figure on this sheet that the plant manager would set higher or lower from the plant's own warranty history.

Shown, not counted

Warranty cost per vehicle, OEM audit findings and their follow-up, the battery-passport data set, and a second line on the same records. Each would add to the total; none is in it. The scrap-per-point figure of the cell-production study (S1) is not applied to module assembly.

What moves each bar, and who moves it

Result	Lever (the input that changes)	Who moves it	Where it shows
R-01	First-pass yield at the bench, 85 → 87 % (C-04, C-05): the NOK code lands against the serial and the station, so the recurring cause is visible per shift	Quality engineer, station owners	Pragmatos Live FPY card · Pulse per characteristic
R-02	Weeks from SOP to rate, 26 → 22 (C-09, C-10): the Thursday number is read from the record, not argued from a spreadsheet. The weakest figure on the board	Launch lead	Live line screen · units per shift against plan
R-03	Station 040 cycle, 104 → 100 s (C-14): the station-to-takt comparison names the loss per module, per shift	Ramp-up lead, cell owner	Live OEE per station (ISO 22400-2)
R-04	Lookup 45 → 5 min, audit preparation 24 → 8 h (C-17, C-19): the part history replaces the join of three systems	Quality engineer	Live part history · records screen
R-05	Containment scope 600 → 120 serials (C-20, C-21): cell lot → module serial on one record	Head of Quality holds, two signatures release	Decisions proposal · Flow quality lot
R-06	Field exposure per event, €1,5 m → €0,3 m (C-25 × scope): the same scope reduction priced at the field rate; the plant sets C-24 and C-25 from its own warranty history	Plant Manager, warranty desk	Calculation sheet 7, not a screen

GENEALOGYCELL LOTMODULE SERIALPACK SERIAL

Crius Software · Pragmatos · launch book CS-11

6 / 7

LAUNCH BOOKCALCCS-11 · SHEET 7 OF 7

Calculation sheet · inputs, formulas, sensitivity

every number on sheets 1 to 6 is derived from the inputs on this sheet · replace a C-value with the line's own figure and the board moves

Inputs ASSUMED: archetype values, replace with the line's own

ID	Input	Value
C-01	Planned rate, units per shift	200
C-02	Shifts a year (3 shifts, 230 days)	690
C-03	Modules a year at rate (C-01 × C-02)	138 000
C-04	First-pass yield at EOL, today	85 %
C-05	First-pass yield at EOL, modelled	87 %
C-06	Rework cost per NOK module	€150
C-07	Scrap cost per NOK module	€1 500
C-08	Share of NOK that is reworked (rest scrap)	75 %
C-09	Weeks from SOP to rate, plan	26
C-10	Weeks from SOP to rate, modelled	22
C-11	Units per shift at SOP	60
C-12	Contribution margin per module	€120
C-13	Shifts a week during ramp	15
C-14	Station 040 cycle vs takt, s	104 / 100
C-15	Steady-state shifts a year (after week 22)	360
C-16	Genealogy lookups a year	300
C-17	Minutes per lookup, today / modelled	45 / 5
C-18	Loaded hourly cost, quality engineer	€65
C-19	Audit preparation hours, today / modelled, × 4 audits	24 / 8
C-20	Containment scope, serials, today	600
C-21	Containment scope, serials, modelled	120
C-22	Containment events a year (in plant)	6
C-23	Handling cost per contained serial	€60
C-24	Field events a year (one in four years)	0,25
C-25	Field cost per serial in scope	€2 500

Sources PUBLIC: archived as text under _evidence/

S1	Fraunhofer FFB, battery-production ramp-up: scrap 15–30 % in the first years, ~10 % after five; ~€30 000 a day per scrap point in cell production
S2	ADAC / KBA recall statistics: 532 recalls in 2025 vs 185 in 2010; 2,84 m vehicles in 2024, +48 %
S3	KBA, product-recall procedure: scope defined by the manufacturer
S4	VDA, 8D method: D3 containment with documented scope
S5	Regulation (EU) 2023/1542, battery passport from 18 Feb 2027 for batteries above 2 kWh, via Battery Pass consortium summary
S6	Trade press, PHEV battery recall: ~€400 m for ~30 000 vehicles
S7	Trade press, EV battery recall autumn 2026: almost 100 000 vehicles

Results DERIVED: formula · value · at half / double the lever

ID	Formula	Value	½ / 2×
R-01	(C-05 − C-04) × C-03 × [C-08 × C-06 + (1 − C-08) × C-07] = 2 760 × €487,50	€1 345 500	673 k / 2 691 k
R-02	½ × (C-01 − C-11) × (C-09 − C-10) × C-13 × C-12 = 4 200 × €120, weakest	€504 000	252 k / 1 008 k
R-03	(200 − 200 × 100 / 104) ≈ 8 units × C-15 × C-12	€345 600	173 k / 691 k
R-04	C-16 × (45 − 5) / 60 × C-18 + 4 × (24 − 8) × C-18	€17 160	10,7 k / 30,2 k
R-05	(C-20 − C-21) × C-22 × C-23 = 480 × 6 × €60	€172 800	86 k / 216 k*
R-06	(C-20 − C-21) × C-24 × C-25 = 480 × 0,25 × €2 500	€300 000	150 k / 600 k
Σ	R-01 + R-02 + R-03 + R-04 + R-05 + R-06, each counted once	€2 685 060

* R-05 at double is capped: the scope cannot fall below zero serials. "Lever" is the input that moves: FPY gain (R-01), weeks taken off the ramp (R-02), units regained per shift (R-03), lookups a year (R-04), serials taken out of scope (R-05), field events a year (R-06). MODELLED

What the deployment is built on

Modules: Loom, Pragmatos Live, Pulse, Orbis, Decisions, Flow; Fabric packs Quality/SPC, Traceability, Automotive.
Read paths: OPC UA, S7, Modbus TCP, REST, CSV; store-and-forward; stale after three missed polls; each connection approved by a supervisor.
On premise, in the plant's own cloud tenancy, or both. Retention: life of the part plus one year, up to 15 years. Designed for the EU AI Act's duties on transparency and human oversight: every card says what it saw, and a person decides.
Not in this study: battery-passport documents (product outlook), any writing to a station or bench, any statement about a recall's outcome.

GENEALOGYCELL LOTMODULE SERIALPACK SERIAL

Crius Software · Pragmatos · launch book CS-11 · rev A

7 / 7