Measured on 10 real 8D cases.
From January to March 2026, eight Slovenian manufacturers solved 10 real 8D cases twice, first the usual way and then with SolveR. Problem definition and root cause analysis (D2 and D4) took 36 minutes instead of 283, 7.8 times faster.
The study is published in the proceedings of the 10th International Scientific Conference on the Development of Industrial Engineering, Novo mesto 2026, pages 77 to 82.
What each number is based on.
D2 and D4 took 36 minutes instead of 283, 7.8x faster, average of 10 cases.
A full 8D drops from 43.1 to 22.5 engineering hours, 48% less, stated as 45%. Only D2 and D4 measured.
Every closed case stays in your knowledge base with its causes and actions.
Same case, same engineers, a stopwatch.
We chose 10 representative 8D cases with real defect data: casting porosity, weld defects, missing components and incomplete assembly. Each case was solved twice by the same engineers, first manually and then with SolveR. We timed the two steps that take longest and depend most on experience: D2, problem definition, and D4, root cause analysis.
A root cause analysis counted only when it named at least one technical cause for why the defect occurred, one for why it was not detected, and an organisational cause for each of the two, four causes in total. Qualified engineers checked every proposal before it counted.
- Real 8D caseReal defect data from the partner
- Manual runStopwatch on D2 and D4
- SolveR runSame case, same engineers
- CheckEvery proposal checked by engineers
n = 10 cases at 8 manufacturers. Source: the paper, Table 2.
Show the numbers as a table
| Step | Manual | With SolveR | Faster |
|---|---|---|---|
| D2 Problem definition | 43.5 minRange: 27.0-55.5 | 4.6 minRange: 3.7-7.1 | 9.6x |
| D4 Root cause analysis | 239 minRange: 171-295 | 31.7 minRange: 21.4-40.1 | 7.5x |
| D2 + D4 together | 283 min | 36.3 min | 7.8x |
The manual run was mostly searching.
Engineers spent on average 47 minutes per case only finding old reports, data and notes before the analysis could start, and the searching went on during the root cause analysis. They opened PDF files and old Excel sheets and called colleagues. SolveR finds similar past cases in seconds and proposes possible causes with their sources. Which cause is real, and which test proves it, stays the engineers' decision.
per case, only to find old reports and data before the analysis starts
Quality check
SolveR proposes 8 to 16 possible causes per case. Engineers kept 20 to 30 percent of them, and in every test so far the real root cause was among the ones they kept.
Illustration: 12 proposed, 3 kept, the root cause among them
Where the 45% comes from.
We found no published figure for how many engineering hours one 8D takes, so we built the model step by step. D2 and D4 are team work. With five people, the low end of a usual 8D team, 283 minutes are 23.6 engineering hours. With SolveR, 36 minutes are 3.0 hours. The other steps stay at today's effort in the model, because we have not measured them. A full 8D drops from 43.1 to 22.5 engineering hours, 48% less. We say 45% to stay on the safe side.
48% less, we say 45%
| Step | Today | With SolveR |
|---|---|---|
| D1 Team Not counted | - | - |
| D2 + D4 Problem and root cause Measured | 23.6 h | 3.0 h |
| D3 Containment Estimate | 3.0 h | 3.0 h |
| D5 Corrective actions Estimate | 15.0 h | 15.0 h |
| D6 Validation Estimate | 0.5 h | 0.5 h |
| D7 Prevention Estimate | 1.0 h | 1.0 h |
| D8 Closure Not counted | - | - |
| Total | 43.1 h | 22.5 h |
SolveR also supports D3 and D5 to D8, but the model counts no saving there, because we did not measure it.
What it means for your team: an 8D takes 22.5 instead of 43.1 engineering hours, so the same team closes almost 2x more cases in the same hours (1.9x; 1.8x at the stated 45%). Or it closes the same cases and puts the hours into prevention, improvement and development. Calculated from the model, not measured.
The next engineer starts where the last one finished.
At one pilot partner, a Tier-2 supplier in CNC machining, a team spent three months looking for the cause of a dimensional deviation. The same problem had been documented in an 8D report two years earlier. The engineer from that time had left the company, and the newer staff did not know the report existed. In SolveR every closed case stays in your knowledge base with its causes, actions and evidence, so the next similar case can find it. That is what 100% knowledge preserved means: every closed case can still be found, even after the people who solved it have left.
- Case closed
- Knowledge base
- New case
- 1 similar case found
Eight manufacturers, ten real cases.
- Slovenian manufacturers
- 8
- employees per company
- 70 to 1,700
- real 8D cases
- 10
- pilot sessions
- Jan-Mar 2026
Industries
- Tier-1 and Tier-2 automotive suppliers
- Die casting and CNC machining of aluminium
- Electrical engineering
- Injection moulding of polymers
- Machine building
Defects in the cases
- Casting porosity
- Weld defects
- Missing components
- Incomplete assembly
We do not publish the names of the pilot partners. They reviewed and confirmed the results.
What these numbers do not show.
- Ten cases show a clear direction. They are not enough for a statistical generalisation.
- The manual run always came first, so the engineers already knew the case in the SolveR run. That can favour SolveR.
- All partners are in Slovenia; other countries were not part of the pilot.
- Only D2 and D4 were timed, on the 10 pilot cases. About 90 more cases were tested internally; the timings do not include them.
- The author of the study co-founded Galileon, which makes SolveR, and says so in the paper.
Read the full paper.
Kalčič, M. (2026). Z umetno inteligenco podprt organizacijski spomin: Avtomatizacija inženiringa kakovosti in preprečevanje ponavljanja napak v proizvodnji [AI-Powered Organizational Memory: Automating Quality Engineering and Preventing Defect Recurrence in Manufacturing]. In T. Perme, S. Muhič and L. Galič (eds.), Development of Industrial Engineering: Opportunities, Potentials and Challenges for Green Transition through the Use of Artificial Intelligence. Proceedings of Reviewed Papers of the 10th International Scientific Conference, Novo mesto, 14 May 2026, pp. 77-82. Faculty of Industrial Engineering Novo mesto. ISBN 978-961-7097-15-3, COBISS.SI-ID 277670147.
The paper is in Slovenian with an English abstract. It was presented at the conference on 14 May 2026.
Questions
How long does an 8D take?
In our pilot, D2 and D4 alone took 283 minutes on the stopwatch per case when done manually, and 36 minutes with SolveR. Our model puts a full 8D at 43.1 engineering hours today and 22.5 with SolveR. For a complex 8D, the manual work up to the root cause is often spread over several one-hour meetings across days; with SolveR it fits into one session. That is what we mean by days become hours. It is our experience, not a measurement.
Which 8D steps did you measure?
D2, problem definition, and D4, root cause analysis. SolveR also guides the other steps and writes the report, but we did not time them, so the 45% counts no saving there.
Do we need old 8D reports to start?
No. SolveR works from the first case. Suggestions get better as your knowledge base grows: with fewer than about 20 past cases they are thinner, and companies usually have a useful base after a few weeks of use.
Will we see the same results?
That depends on your cases and on how scattered your documents are today. A pilot runs on your own cases, so you measure the difference in your own plant.
Is the study independent?
No. The author, Matej Kalčič, co-founded Galileon and says so in the paper. The pilot partners reviewed and confirmed the results, and the paper was published as a professional paper in the conference proceedings of reviewed papers.
Measure it on your own cases.
In 30 minutes we show SolveR on a case like yours and agree how to measure the difference in your plant.