To develop a manufacturing process, describe the sequence of operations, equipment, materials, operating conditions, and quality checks, then test the process in practice. Record the test results, address any problems identified, and prepare clear instructions for employees.
This plan helps turn an idea or a finished prototype into a repeatable manufacturing process. This article explains how to describe a process step by step, check whether it is feasible, and document it so products can be made consistently.
| What to check | What to record | Decision |
|---|---|---|
| Hypothesis | Action, measurable result, rationale | Run a trial or refine the hypothesis |
| Materials and equipment | Process inputs, equipment condition, limitations | Prepare for testing |
| Trial run | Sequence of actions and inspection results | Repeat the iteration or move on |
| Quality and costs | Results against criteria and cost items | Decide whether the process is ready to scale |
- 3 parts of a hypothesis: action, measurable result, and rationale
- 4 types of data to analyze equipment: temperature, vibration, load, and past failures
- 4 elements of a business process: goal, sequence of actions, people responsible, and KPIs
How do you turn a manufacturing idea into a testable hypothesis?
Action, result, rationale
Frame a manufacturing idea as a testable hypothesis: “If [action], then [measurable result], because [rationale]” — a structure proposed in an article by the Russian School of Management. Instead of a general goal such as “improve output,” specify the operation and the change to be made: for example, change how raw materials are fed in and check whether this affects product quality, cycle time, or material consumption.
What exactly should you test?
A hypothesis should link a single change to a clear metric and an explanation of the expected effect. For a raw-material feeding operation, the logic might be: changing the feeding method should affect consumption because it eliminates losses at a specific stage. This is an example of a structure, not a promise of results: the rationale must be tied to the actual process and the observations available.
How do you measure the result?
Before testing, record the baseline and how you will measure it, then use the same method for the trial process. For quality, decide which product metric to record; for cycle time, specify where the measurement starts and ends; for material consumption, define which materials and operations to include in the accounting. Comparable baseline and trial results make it possible to verify the effect; without them, you cannot separate the change from differences in how data was recorded.
Do not set a target in percentages or rubles before collecting baseline data and defining the calculation. First choose a metric — such as cycle time or material consumption — and specify its unit of measurement. It makes sense to set a numerical threshold only once it is clear exactly what is being measured and how results will be compared.
What should you check about materials, equipment, and documentation before testing?
Before testing, make sure the process inputs and outputs, material availability, equipment readiness, participants, and inspection procedure are all described. This preparation makes it possible to compare iterations using recorded criteria rather than trying to reconstruct the trial from memory.
For materials, record exactly what goes into the operation; for the output, specify what the product should be and which verifiable characteristics determine whether it passes. Then check that the material is available and the equipment is in suitable condition for the specific operation: having the material on hand does not, by itself, confirm that the equipment is ready for the trial. Separately record any limitations that could prevent the test from being repeated, such as unavailable materials or equipment unsuitable for the planned operation.
Who is responsible, and what should be recorded?
Involve process engineers, engineers, and the technical lead in preparation: an article on R&D projects links the technical lead’s responsibilities to the engineering solution, documentation, and manufacturability. Before the trial, document the procedure and inspection criteria; the process should include a goal, sequence of actions, people responsible, and metrics for assessing the result.
The records should make it possible to compare iterations: specify which materials were used, what equipment the operation was performed on, what was done, and which characteristics were used to assess the product. If any of these details changed, record that separately — otherwise, it will not be possible to confidently link differences in results to a specific change.
How do you run a trial and determine whether the process works?
A trial will show whether a process works if you test it on a real manufacturing task, follow one consistent sequence of actions, and compare the result with criteria defined in advance. This approach is recommended by Worksection’s algorithm: test the process as described in actual work rather than relying on a diagram or hypothetical example.
Iterate instead of scaling too soon
For each run, record the actual conditions, any deviations from the sequence, and the quality inspection results. Compare your observations with the success criteria set before testing, such as specified quality parameters or the target process outcome. Do not change the criteria after seeing the results; otherwise, the comparison will not show whether the hypothesis was confirmed.
- Criteria met: record the test results and assess whether the process is ready for the next stage.
- Criteria not met: refine the hypothesis based on the deviations identified and repeat the test as the next iteration.
Keep the engineering documentation and conclusions together with the inspection results: an account of R&D projects identifies complete documentation and manufacturability as conditions for moving to the next stage. The test record should therefore make clear the sequence of operations, actual conditions, deviations identified, and reasons for deciding to continue, make changes, or test again.
How do you assess quality and production costs before scaling?
Assess the quality and production costs of a manufacturing process across a repeatable series of trial runs: calculate the actual costs of materials, equipment operation, and necessary operations, then compare the results with the baseline using the same metrics. One successful product is not enough: it does not show whether the process works consistently or whether the result can be reproduced.
What to record during a trial run
- Costs: record material consumption, equipment operating time, and the cost of each operation needed to produce the item separately. Do not include costs unrelated to the process being tested.
- Quality: choose metrics in advance for comparing the trial result with the baseline, and use the same inspection method each time. Record defects and inspection results alongside the stage where they were detected; this makes it easier to determine whether a problem is related to the materials or an operational failure.
- Repeatability: run several trial cycles under the documented conditions. Compare the results with one another, not just with the best-performing product.
Scale the process when the sequence of operations, operating conditions, quality checks, and cost calculations are documented well enough to be repeated and verified. If the materials, equipment settings, or order of operations change, record that separately; otherwise, it will be difficult to explain differences between trials.
Where can digital manufacturing and AI help, and where can’t they replace testing?
Digital manufacturing helps monitor quality as work progresses, while AI can analyze equipment data. But neither digital monitoring nor an algorithm can, by itself, prove that a new process meets requirements: the results must be checked against measurements defined in advance.
SAP contrasts digital monitoring with the traditional approach, in which quality is checked periodically, documentation is kept manually, and defects are addressed only after they are found. Digital monitoring makes it possible to track a process as it runs, but any deviation still requires a responsible specialist to make a decision and verify the result itself.
Data should support a specific decision
An article by the Higher School of Economics describes predictive maintenance as one use of AI: a model can analyze equipment readings and quality inspection results. For such a model, it is important to have data related to the condition and history of the specific equipment:
- Temperature, vibration, and load — readings a model can use to look for signs of changes in operating conditions.
- Past failures — records for comparing current readings with known breakdowns.
- Quality inspection results — data linking equipment performance to a verifiable outcome.
Before introducing a digital tool, describe the process: which parameters are measured, who assesses the results, and which decision depends on each metric. If the goal is to confirm the quality of a new process, an algorithm can help flag signals for review, but it cannot replace acceptance criteria and hands-on inspection.
Why might a process fail when transferred to production?
A process may fail when transferred to production if its effect was not measured in advance, it was tested using incomplete data, or it was scaled before the process and quality criteria were documented. Without these foundations, it is impossible to reliably distinguish the process’s effect from a random deviation, and a trial run does not guarantee that the effect will be reproduced under different conditions.
A testable hypothesis links an action to a measurable result and explains why that result should occur: for example, “if we implement this process, the selected metric will change because…”. This metric and the success criterion must be recorded before testing. If the trial does not account for the characteristics of the materials and equipment, its conclusions cannot automatically be applied to other batches, settings, or production conditions.
Quality control and readiness to scale
Manual inspection detects a defect only after it has appeared, so it can be difficult to determine from the finished product at which stage the problem occurred. Digital monitoring can track quality in real time, but only using measurements selected in advance; AI-based predictive maintenance, for example, requires data on temperature, vibration, load, and past failures.
Before scaling, document the sequence of operations, people responsible, records, and quality criteria, then test the described process on a real task. Otherwise, ambiguity that might be acceptable in a one-off trial will spread across the entire production run: employees may perform the same operations differently, and deviations will be harder to compare.
Frequently asked questions
Where should you start when developing a manufacturing process?
Who should be involved in developing a manufacturing process?
When can you scale a trial process?
What data does AI need to predict equipment maintenance?
Sources
- Russian School of Management blog — “How to Test Ideas at Work: Hypothesis Testing, Rapid Iterations, and a Management Approach”
- habr.com — “From Hypothesis to Production: How R&D Projects Work / Habr”
- worksection.com — “How to Create a Business Process: A Step-by-Step Guide with Examples — Worksection Project Management Blog”
- sap.com — “What Is Digital Manufacturing?”
- online.hse.ru — “AI for Business: Why Companies Should Introduce Artificial Intelligence into Their Processes Now, with Examples”
