DMAIC Analyze Phase: Finding Root Causes of Defects in Plastics Production

Production analytics connecting injection molding process conditions with defect trends

Part 4 of our DMAIC series

A defect is a signal, not a diagnosis. Short shots, flash, splay, dimensional variation, and other injection molding defects can have several possible causes. The Analyze phase of DMAIC helps teams move beyond plausible explanations to evidence-backed root causes.

The goal is practical: identify what is driving the defect, where and when it happens, and what evidence supports that conclusion. Then your team can improve the process with greater confidence.

Start with a focused investigation

Before selecting an analysis tool, define the problem clearly. Specify the defect, the product or process affected, the time period, and how the defect is measured. Confirm that inspection criteria and measurement methods are consistent. If teams count or classify defects differently, the analysis can point in the wrong direction.

Next, map the potential sources of variation. In plastics production, those often include:

Machine: temperature control, pressure, injection speed, cycle behavior, or equipment condition.
Material: resin lot, moisture, contamination, regrind, or changes in material handling.
Mold or die: wear, venting, cooling balance, alignment, cavities, or flow paths.
Method: setup, process parameters, changeovers, work instructions, or operator adjustments.
Environment: ambient temperature, humidity, or other conditions around production.
Measurement: inspection method, gauge, sampling frequency, or recording practices.

This list frames the investigation. It does not prove that any one factor is responsible. A familiar defect can have an unexpected cause, and several process conditions can interact. In injection molding, for instance, defects that seem to call for opposite adjustments may be linked through the wider process. Plastics Technology explains why verifying the cause matters before changing a process.

Use classic tools to narrow the search

Root cause analysis does not have to begin with complex statistics. A few familiar quality tools can help teams organize what they know and decide where to look next.

Pareto analysis prioritizes the problem. Rank defect categories by frequency, cost, or another meaningful measure. If a small number of categories account for most of the impact, begin there. Then stratify the data where it helps: a plant-wide view may look different from a breakdown by machine, product, cavity, or shift. Aggregation can hide a problem concentrated in one place.

A fishbone diagram structures hypotheses. Place the selected defect at the “head” of the diagram, then group possible causes under categories such as machine, material, mold, method, measurement, and environment. For example, if a team is investigating splay, it might consider material moisture, drying practices, contamination, and process conditions. These are candidate causes, not conclusions.

The 5 Whys helps trace a cause chain. Ask why the defect occurred, then why the next condition occurred, continuing until the team reaches a cause it can test or control. If a short shot is linked to insufficient filling, ask what conditions affected filling and whether the evidence supports that explanation. Avoid ending at labels such as “operator error” or “machine issue.” Those describe a broad area, not a verified root cause.

Fishbone-style root cause analysis organized around plastics process variables

Test hypotheses against production data

Once the team has a shortlist, turn each candidate into a testable question. Did defect rates change by machine? Were they higher on one shift, with a particular material lot, or after a tool passed a certain age? Did process measurements differ between good and bad periods?

Compare like with like. A useful comparison might examine the same product and tool across two time windows, or the same material lot across shifts. Where several factors change at once, note them and plan the analysis to account for them. Otherwise, an apparent relationship may reflect a different variable.

Look for the time-stamped process context around the defect. For injection molding, that may include melt temperature, pressure, injection speed, cycle conditions, or other available measurements. A defect record without its production context can tell you what happened, but not much about why. Linking quality outcomes with process measurements helps engineers see whether a suspected condition changes alongside the defect.

The same principle applies to extrusion and other plastics processes: align the quality observation with relevant machine, product, material, and process data. Compare periods, machines, shifts, or lots based on the question being investigated, not simply because a dashboard makes the comparison easy.

Production trend comparisons across shifts, material lots, and tool age

Separate special causes from common causes

A change in the data does not always mean the same thing. Special-cause variation is linked to a specific, unusual event or condition. Common-cause variation is the variation inherent in the current process.

That distinction matters because the response should differ. A one-off process interruption or unusual material event may call for a targeted investigation. If variation is part of the process’s normal behavior, repeatedly adjusting settings in response to individual readings may add instability rather than solve the problem.

Use process context and appropriate statistical methods to distinguish a meaningful shift from routine fluctuation. Avoid treating a correlation as proof: data can help identify a likely cause, but a controlled trial or further verification may be needed to establish that the cause is real.

Avoid two common traps

Too little data. A handful of defects may not represent the process. Make sure the sample covers enough production to examine the conditions that matter, including relevant shifts, machines, products, and material lots.

Too much aggregation. Combining different products, tools, or process conditions may mask a concentrated cause, or create a misleading average. Break the data into sensible groups, but keep the investigation tied to a clear question.

Also check that the measurements are trustworthy. If defect classification, sampling, or process-data capture is inconsistent, improve that foundation before drawing strong conclusions. Reliable root cause analysis depends on reliable, appropriately connected evidence.

How iDOO supports the Analyze phase

The Analyze phase asks teams to connect production conditions with quality outcomes. The iDOO Portal brings production services, KPIs, machine status, and analyses together, with the ability to integrate existing applications and data. With relevant data available and aligned, teams can compare periods, machines, and products and investigate patterns across production.

The Industrial Data Manager supports production-data recording and dashboards. When process measurements and quality outcomes are available together, engineers can examine defect trends in context instead of piecing information together from disconnected records.

Analysis also depends on experience. The Experience-based Assistant helps make operational knowledge more accessible, so teams can look for how similar issues were investigated and resolved before. When teams document their interventions and outcomes, they can turn a successful response into reusable knowledge rather than starting from scratch the next time a defect appears.

Industrial assistant interface for making process knowledge easier to reuse

We bring more than digital tools to this work. Our team draws on decades of plastics manufacturing expertise through the CiTEX Group to help you frame the investigation, connect the right information, and assess what the evidence supports. iDOO is a capable partner, not a substitute for your engineering judgment or process knowledge.

Make the root cause actionable

A useful Analyze-phase conclusion is specific and evidence-based. It states the condition associated with the defect, explains how that condition affects the process or product, and identifies where or when the effect appears. The team should also record what evidence supports the conclusion and what still needs validation.

A quick check for your team: Can you link each defect to a time, machine, product, and relevant process conditions? Can you compare good and bad production periods? Do you know whether the observed variation is unusual or part of the normal process? And can operators and engineers find records of how similar issues were resolved?

Your answers can help identify the next step: improving data availability, strengthening analysis, or moving to corrective trials.

Ready to move from analysis to improvement?

If you are investigating recurring injection molding defects or another plastics quality issue, talk with the iDOO team. Share the defect you are targeting, the process data you can access, and where your investigation is getting stuck. We can explore how connected production data and plastics experience can support your next step.

In Part 5 of the DMAIC series, we’ll move into the Improve phase: testing and implementing changes that address verified root causes.