Choose the quality tool for the question

Choose the quality tool for the question

Choosing a tool starts with the decision, not with a favourite diagram. A check sheet collects observations consistently; a Pareto ranks a stated burden; a histogram displays a distribution; and a scatter plot explores paired variables. A control chart adds time and a baseline model. A fishbone organizes possible causes, while stratification separates sources that pooling may hide. These tools can be used together, but each answers a different question. First write the question in plain language. Then choose the smallest useful display and state what further evidence is needed before acting.

Follow the method

1. Record consistently
2. Rank a burden
3. See a distribution
4. Explore paired variables
5. Organize possible causes
6. Separate mixed sources
7. Monitor over time

Match the question to the evidence and the decision

Fictional case: Vale Dispatch has delivery complaints, measured packing times and a list of possible causes. Coordinator Lina is asked to produce a quality chart for the weekly meeting. The manager wants to know where to begin, while a technician wants to know whether packing time changes with batch size. These are different questions; one attractive chart cannot answer both.

Follow the method

1. Where are most comparable complaints?
2. Does time vary with batch size?
3. Which explanations should be tested?
4. Are sources hiding different patterns?
5. How does behavior evolve?

A pile of records is not yet a question

Turn a vague quality concern into an answerable question, select a suitable evidence display, and state what that display cannot establish.

Fictional case: Vale Dispatch has delivery complaints, measured packing times and a list of possible causes. Coordinator Lina is asked to produce a quality chart for the weekly meeting. The manager wants to know where to begin, while a technician wants to know whether packing time changes with batch size. These are different questions; one attractive chart cannot answer both.

Role: Dispatch coordinator and process analyst

Normal condition

Every proposed display has a stated question, an observation unit, a defined window and an honest limit of inference. Raw records retain their original identities.

The gap

A draft slide mixes complaints, hours and suggested causes into one descending bar chart. Its title says the biggest bar is the root cause.

• The seven-tool overview is a selection aid, not a compulsory sequence.
• A severe isolated complaint can require immediate response even when it ranks low by frequency.

1. Write the decision first

   Lina separates “Which complaint category contributes most occurrences this week?” from “Do larger batches tend to take longer?” She records the intended user of each answer.

   Why: The question determines the evidence and display. Starting from a preferred graphic invites conclusions that the available records cannot support.

   Evidence: Two written questions, each tied to a decision rather than “make a chart”.

2. Check the collection unit

   She defines one complaint occurrence and checks duplicates against shipment IDs. For the timing question she keeps each batch-size/time pair together instead of sorting the two columns separately.

   Why: Comparable counts need a consistent category rule. A paired relationship disappears or becomes fictitious if the observations lose their pairing.

   Evidence: Complaint records retain their window; every time has its matching batch size.

3. Choose the first two views

   She assigns a Pareto chart to complaint frequency and a scatter plot to the paired observations. A check sheet is the next collection method if future categories are inconsistent.

   Why: Pareto ranks a chosen burden; scatter explores association. Neither is a root-cause verdict, and neither replaces the record definitions.

   Evidence: The chart request identifies tool, input and permissible interpretation.

4. Place other questions deliberately

   A histogram can summarize packing-time distribution; stratification can separate routes or product groups; a fishbone can organize competing hypotheses. Time behavior needs ordered data and an appropriate run/control view.

   Why: These tools reveal different structures. Mixed sources, timing and causal hypotheses should not be folded into an unexplained “quality score”.

   Evidence: A selection record covers all seven roles without claiming that every project must use every tool.

5. Approve a bounded next step

   Lina asks the manager to prioritize a category for investigation and asks the analyst to assess the paired data. She removes the root-cause title and records the unresolved checks.

   Why: A useful display ends with a defensible decision and the next evidence needed. Restraint about causation is part of using the tool correctly.

   Evidence: The meeting assigns evidence collection, not a causal conclusion from a bar height.

The manager changes the basis

The manager now asks which complaint type consumes the most recovery hours.

Rebuild the Pareto using documented recovery time for every included category, label hours and retain the original frequency chart separately.

Changing the basis can change the rank. Adding hours to occurrence counts would produce an uninterpretable total.

A revised request names recovery hours and the same declared reporting window.

Run a practice session

Materials

• Blank question-to-tool sheet
• Three evidence cards
• Answer kept separately

1. Frame the meeting request · 4 minutes

   What decision is hidden inside “make a quality chart”?

2. Match and challenge · 8 minutes

   Defend a tool, then name what it cannot tell us.

3. Complete the clinic exercise · 10 minutes

   Which unit belongs on each axis or total?

4. Debrief transfer · 5 minutes

   What raw information must survive the display?

Debrief

• Accept alternative suitable tools when the question and boundaries are explicit.
• Challenge any claim of a cause based only on ranking or consensus.

Individually write each request as a question before looking at the seven-tool overview.

Transfer into the work

Owner: Process analyst with the record owner

Record: Question, collection definition, selected display and interpretation note

Review: At the next evidence review

Evidence: A colleague can reproduce the display and distinguish findings from hypotheses

Return to data definition or collection when the requested conclusion exceeds the evidence.