The book is being typeset, please wait a moment.
SEM: Thinking Under Pressure
Strategic Engineering Matrix for complex industrial escalations
Ingo Konrad Dentz
Publisher: IKAR Solutions OÜ
Last revised: V5, October 17, 2026
© 2026 IKAR Solutions OÜ. All rights reserved.
This work is protected by copyright. The copyright in the text and graphics is held by IKAR Solutions OÜ. Reprinting, reproduction, and distribution, in whole or in part, require the written permission of IKAR Solutions OÜ. This also applies to storing it in databases and to using it for training AI systems.
All case examples are anonymized. The technical content reflects real projects.
SEM (Strategic Engineering Matrix) is the intellectual property of IKAR Solutions OÜ.
Typeset in Source Serif, Inter and JetBrains Mono. A5 format.
Prologue
The supplier quality engineer (SQE) of the vehicle manufacturer calls. Of the rear lamps from the night shift, eighty percent have broken in the alcohol test, and the part is still warping. The team has been working on the fault for three weeks, and the production launch is in five weeks.
In the first week the customer wanted to understand what was going on. In the third, the customer wants a date. The project manager has none. What the project manager has is a list of trials that all changed something and explained nothing.
At eight the meeting begins, eight people from project management, process, mold shop, quality, purchasing, logistics and development. An Ishikawa diagram takes shape on the flip chart. The items named are the new material lot, a nozzle-side mold half that runs three degrees warmer than the ejector side, the hold pressure and the cooling time. Each of these changes has helped for two days at some point. Then someone says: “If the parts cool down in the fixture, they come into tolerance.”
That puts the explanation back in the room that has held since the first week: residual stress from laser welding. The weld brings heat into the part, and the stress remains; under the cleaner the part cracks where the stress is highest, and as soon as it is unconstrained it warps. The explanation is plausible: it explains the fracture and the warpage, and it explains why the fixture helps. On the right edge of the flip chart it now says “residual stress,” with a line underneath.
From then on every proposal is measured against this explanation: fixture time even longer, change the packaging once more, cooling time once more. Since the second week the toolmaker has noticed that the parts already warp at demolding, and so before the welding process. The toolmaker said so twice; both times the fixture fit better with what was already on the flip chart. After ninety minutes there are six new “actions” on the flip chart, each with a name and a target week. For none of them is it settled what one would have to see if it does not work. The next trial on Wednesday changes three parameters at once, because there is no time for three separate trials.
Nobody in this meeting did poor work. What was missing was one question: Which other mechanisms carry the same symptom, and which finding would tell them apart? That the fixture helps fits residual stress. It excludes nothing, because a fixture holds every part in shape that is still moving after demolding, for whatever reason. And if the parts from Wednesday's trial are better, nobody knows which of the three parameters had the effect.
That is the difference between trial and error and fact-finding. A trial for which it is settled beforehand which result supports which mechanism and which result excludes it brings the team step by step to the mechanism that carries the symptom, and to an intervention that eliminates it. A trial without this question can deliver today's parts. Whether it prevents the next escalation, nobody knows afterward, because it remains open what had the effect.
SEM, the Strategic Engineering Matrix, gives this work its order. The team first records the situation, the context. It describes the symptom, collects findings and keeps several testable mechanisms side by side. Before every measurement it settles which result supports or excludes which mechanism. Where delivery is at risk, an interim intervention secures it from the first day, even while the mechanism is still open. The intervention meant to eliminate the mechanism is chosen only when one mechanism leads after the findings, and the verification checks after stable production whether the mechanism is eliminated.
The book shows this sequence on injection molded parts and on escalations from series production. Two everyday cases, a door and a damp wall, introduce the terms, because everyone knows the physics there and can direct full attention to the method. The rear lamp from this prologue is worked through step by step in chapter 26. The cases from real projects are anonymized. Most of them were worked before SEM had a name; the method is derived from this practice, and one of these cases already ran with it. The rest are everyday and teaching cases, built so that one step of the method becomes visible on its own.
I
Part I
Thinking
Chapter
1
An everyday case introduces the method before its terms are worked out. Everyone knows the physics of a door; that keeps attention on the method. The symptom is the described disturbance, without interpretation. A mechanism is a pathway that explains how the symptom arises. Findings are results of measurements, inspections and trials; they make a mechanism more likely or less likely. All mechanisms in question form the workspace. The mechanism that is in front after the findings is the hypothesis. The context is the situation in which the decision is made; it determines which intervention fits. At the end, the verification checks whether the intervention acts on the mechanism as planned: whether it eliminates it or whether the system tolerates it safely.
Since July the living room door has been sticking; it dates from 1978, like the house. It rubs with its lower corner on the lock side, over the last fifteen centimeters before the stop. At the top there are three millimeters of clearance, as there have been for years, at the bottom zero. The rub mark in the paint of the door frame sits at the very bottom of the frame, at the height of the lower corner of the door leaf.
That is the symptom. “The wood has swollen” would already be an interpretation, and it does not explain why the door rubs only at the bottom. The description itself is the first finding: every mechanism will later have to explain what it contains.
Right away, the resident also notes the situation, the context: own house, the heating season starts in ten weeks, the hand plane is in the basement. The context is applied only at the intervention.
Four mechanisms are in question. They form the workspace.
| No. | Mechanism | What it predicts |
|---|---|---|
| M1 | The door leaf has absorbed moisture and become wider. | Less clearance over the whole height, at the top as well. |
| M2 | The upper hinge has loosened, and the door sags. | Play when the door is lifted by the handle. |
| M3 | The frame, set in mortar in 1978, has shifted. | Frame out of plumb, cracks where it meets the plaster. |
| M4 | The wall is moving. | Other doors and cracks in the same wall. |
Next to the house there has been an excavation pit since April. M4 is the uncomfortable mechanism: if the wall is moving, nobody in the house can change it, and that is why nobody likes to say it out loud. It still belongs on the list. What is not in the workspace cannot be tested by any finding.
Now measurements are taken, and for each one the question is: Which mechanism becomes more likely, and which less likely?
What remains leads: the frame has settled. That is the hypothesis. It is in front because two findings support it and the other mechanisms are excluded or deferred. Whether the frame keeps moving is still open.
The context was noted from the start, and it is applied only now: the house belongs to the resident, the heating season starts in ten weeks, and a new frame costs EUR 700 and two days of dust. The hand plane is in the basement.
Planing the door leaf down takes twenty minutes and cannot be undone. If it was the wood after all, the door has three millimeters of clearance at the bottom in winter. If the frame keeps settling, there will be more planing in the spring. A hinge washer is reversible, a steel ring for EUR 2 that sits on the hinge pin. Two rings raise the door by two millimeters, the lower corner clears the spot where the frame leans inward, and the door opens freely. The rings compensate and eliminate nothing; the frame is still crooked. That is fine as long as one knows it.
That the door opens freely in October proves nothing. With the heating on, wood dries, and a swollen door would have opened freely in October without rings, too. So the check targets the mechanism: has the frame come to rest, and do the rings carry the rest? Once a month the clearance at the bottom is measured with the same ruler. If the value holds through the heating season and frost, the frame is no longer moving. After a winter without renewed sticking, the frame is supported as the mechanism. The verification is complete after the next summer, when the door also stays free in humid air. If the value drifts, the frame is still moving, and the wall is back on the table.
| Step | At the door |
|---|---|
| 0 Context | Own house, heating season in ten weeks, new frame €700 |
| 1 Symptom | Rubs at the bottom on the lock side, three millimeters of clearance at the top, zero at the bottom |
| 2 Findings | Clearance at the top unchanged; no play at the hinge; frame two millimeters out of plumb with a crack; neighboring doors free |
| 3 Workspace | M1 wood, M2 hinge, M3 frame, M4 wall |
| 4 Hypothesis | M3, the frame has settled |
| 5 Intervention | Two hinge washers, reversible, removable in five minutes |
| 6 Verification | Clearance at the bottom monthly; supported after the winter, complete after the summer |
This is the movement this book teaches. On an injection molding machine it is the same movement, with more mechanisms and more expensive findings. Chapter 3 works out the terms.
Stop. A door, a device, a machine that has “always been like that” where you are. Which four mechanisms carry the symptom, and which of them has nobody ever said out loud?
Chapter
2
An 8D report is on the table, the customer's form for a complaint in eight disciplines. D3, the interim containment actions, has been reported to the customer; now D4 is due, the cause of the defect. A crack appears on the part. Three explanations are in the room: the new material lot, the mold and the mold temperature control. The deadline pushes toward entering the lot. The supplier changed the formulation; that is in the data sheet. Whether the change causes the crack is not proven by that.
The report has to go to the customer on Friday. In the same week, three things are changed: the remaining stock of the old lot is run, the temperature control is changed, and the mold is reworked at the crack location. The scrap rate drops, and the report calls that the verification of effectiveness. This proof does not say which of the three changes had the effect. The customer closes the complaint.
When the remaining stock is used up, the new formulation runs again. A residual share of cracks has remained; no field explains it. Then the temperature control is adjusted a second time, and the crack is back as on the first day. The complaint is reopened, and the customer asks why D7 was signed.
The form is fine. According to the VDA volume on 8D, D4 requires investigating all applicable causes; several are explicitly provided for. A cause is proven when its pathway switches the defect on and off in a targeted way. D4 also asks why no inspection found the defect before the customer did. D5 selects the permanent corrective actions and proves their effectiveness before introduction, D6 introduces and validates them, D7 prevents recurrence on this part and on parts of the same design, and D8 closes. In D4 there was an explanation without proof, in D5 a scrap reduction after three simultaneous changes. That is an 8D narrowed under deadline pressure, and it is common.
Ishikawa diagram, 5 Whys and 8D are tools for structure and documentation, and they do what they are meant to do. The Ishikawa diagram collects possible causes broadly by category: man, machine, material, method, milieu, measurement. 5 Whys follows one branch into depth and is linear in its basic form. The 8D organizes the whole process, from the team through the containment action to closure.
Under deadline pressure, application often narrows to one cause. The Ishikawa diagram is broken off as soon as one branch convinces, as in the prologue. D4 contains the cause that can be proven by Friday. For this narrowed practice, this book uses the chain symptom, cause, action. It describes how the methods are often run under pressure; the methods themselves allow more.
With 5 Whys, the narrowing shows in the form of the answers. A 5 Whys text on delamination at a two-component PMMA part answers every why with a claim; no answer rests on a measurement, an inspection or a trial. The chain ends with training and a work instruction, a corrective action without a single physical finding. Formally it is complete, since every question has an answer. It cannot predict whether the delamination will recur afterward, because none of its levels is tested.
The narrowed chain holds when the problem is linear. For that, four conditions must hold together:
If the four hold, one cause, one corrective action and one check are enough.
In an escalation, the four conditions rarely hold together, and time and clear responsibilities are usually missing as well. The crack shows this.
| Assumed | At the crack |
|---|---|
| One dominant cause | At least two mechanisms acted at once, the lot and the temperature control. How much each contributed could no longer be separated after three simultaneous changes. |
| Complete, reliable data | The crack rate was available as a weekly value, without assignment to shift, cavity or lot. |
| Free choice of the best intervention | The supplier delivered the old formulation only as remaining stock, and the mold belonged to the customer. |
| Time for a clean analysis | D3 had to be in place within 48 hours, before the defect pattern was described. |
| Clear responsibilities | Material supplier, molder and customer each had part of the answer. |
On top of that comes a confounding factor: earlier interventions mask the signal. When the complaint opened for the second time, a different lot, a different temperature control and a reworked mold were running. Nobody could say which crack was still the old one.
The chain knows only 0 or 1 at each step. The field carries several mechanisms, and each finding acts on all of them.
The graphic shows the difference the book builds on. The chain knows only yes or no at each step: if the chosen path is wrong, the work is lost. The field carries several mechanisms side by side, M1, M2, M3, and each finding acts on all of them. An arrow shows how a finding raises or lowers a mechanism, and the cross excludes it. Chapter 5 introduces this notation.
SEM is not the first method to go beyond the chain, and it takes over what has proven sound. The book introduces the terms of the right-hand column from chapter 3 onward.
| Method | What SEM takes over | What SEM adds |
|---|---|---|
| Ishikawa | the broad collection by category, as a checklist for the workspace | fact-finding by discriminating power, the standing of each mechanism |
| 5 Whys | the question into depth, until the pathway is named physically | several pathways side by side, the counter-check for each pathway |
| 8D per VDA | team, containment action in D3, investigation of all applicable causes, verification of effectiveness, preventing recurrence | the running standing of each mechanism on the way to D4, robustness as its own intervention type, the scale of the context, the verification on the mechanism |
| Kepner-Tregoe | is and is-not as negative evidence, musts and wants as exclusion criteria and weights | the context before the criteria, the uncomfortable mechanism, the context-shifting intervention |
| Shainin | the comparison of good and bad parts | the workspace before the comparison, interventions according to the context |
| DMAIC, A3, PDCA | measuring before analyzing, the one page, the cycle with a way back | leading with incomplete data, the interim intervention, deferred mechanisms with a trigger |
| Toyota Practical Problem Solving, MSA | genchi genbutsu, going to see at the part; checking the gauge before the finding | the question of which mechanisms a measurement separates |
| IATF 16949 | the requirement for a documented problem-solving process (section 10.2.3) | a procedure that can be documented as this process |
What SEM itself contributes is four things. First, SEM keeps mechanisms and findings running with a standing: at least three mechanisms, each with its assessment, and each finding moves them visibly. Second, fact-finding follows discriminating power: the finding that separates the most comes first. Third, the interventions are staged according to the context: interim, robustness, optimal solution, reversible first. Fourth, the verification checks on the mechanism whether it is eliminated or whether only the symptom has gone.
The 8D does not end at the cause, and SEM does not replace the 8D. In the language of the loop, D3 corresponds to the interim intervention, which secures delivery before the mechanism is established. D6 usually corresponds to the optimal solution, which eliminates the mechanism. In between stands robustness: with the mechanism known, it widens the process window as long as the optimal solution cannot yet be implemented. The context measures which intervention fits the situation, with criteria, weights and exclusion criteria (chapter 6). And the verification asks whether the mechanism is eliminated. At the crack it would have shown this: a scrap reduction after three simultaneous changes proves none of them.
The form stays, and it can be filled with SEM. D2 carries symptom and defect pattern, D3 the containment action, kept internally as an interim intervention. A mechanism goes into D4 only when a finding switches the defect on and off with it; until then, D4 reads “under analysis,” with the next finding and its date. D5 and D6 carry the permanent corrective action with its target week, D5 the proof of its effectiveness, D6 its validation; the team keeps the SEM type and the verification on the mechanism internally alongside. D7 takes up the transfer to parts of the same design. The field-by-field mapping is in the appendix “SEM operating guide.” For reading up on the disciplines, use the VDA volume on the 8D method (VDA QMC, Problem Solving in 8 Disciplines).
The question at the start of every escalation is therefore: Do the four conditions still hold? If they do, you work linearly, and SEM shrinks to its linear form: one mechanism, one finding, one intervention, one verification. If one of the four fails, the field opens: several mechanisms, findings that exclude, interventions according to the context.
Stop. An escalation from your own practice: the symptom came back after an intervention. Which of the four conditions did not hold at the time, and which finding would have separated the simultaneous changes from one another?
You have read the prologue and the first two chapters. The complete book has 26 chapters in three parts, plus the epilogue, glossary and appendices.
It is included in the SEM Academy course. You receive access with the invitation from your course leader and then read the complete book in the app. A purchase option for the book is in preparation.
sem.ikarsolutions.com/buch