Operations & Systems

Control Limits vs Targets: Separate Process Behavior From the Result a Business Wants

Three horizontal lines appear on the service report. One comes from a statistical baseline. One comes from a planning decision. One comes from a customer requirement.

The legend calls all three “limits.”

Give each line its own name and source. A report should explain what the process appears to do, what the business wants, and what a customer requires. Those statements need different evidence.

The worksheet below is a proposed reporting method. Every observation, baseline, requirement, and decision in the worked example is hypothetical. The numbers illustrate line meanings. They validate no control chart or capability claim.

Separate the Three Meanings Before Reading the Chart

Use target for the result an authorized business owner chooses to pursue. Record the measure, desired value, scope, owner, and effective period. In this example, the target is to complete each eligible review within twenty minutes.

Use specification limit for a stated boundary of acceptable output. Identify the document and authority establishing it. Here, an invented customer agreement sets a maximum of thirty minutes for each eligible review. The example supplies no legal interpretation or real contract term.

Use control limit for a boundary associated with a statistical monitoring method. NIST's control-chart guidance describes limits around an in-control process and a center line representing its mean. Their meaning depends on the chart method and underlying process assumptions.

Record who owns each line. The business sponsor owns the target decision. The requirement owner identifies the applicable specification. The statistical method owner documents how the process limits are established and used.

One person might hold multiple roles. Keep the roles distinct anyway. Authority to revise an internal goal does not establish authority to revise an agreed requirement. Neither decision supplies a statistical baseline.

Before adding lines to an operator dashboard, make the legend explicit. “Internal target: twenty minutes” tells the reader more than a red band called “acceptable.”

Require Comparable Measures for Every Line

A line can have a valid source and still belong on a different chart.

Suppose observations measure individual review durations. A monthly average target measures a different quantity. Comparing each individual point with that average target would answer a question the target never defines.

Write down the observation unit, start event, end event, elapsed-time basis, eligible population, and exclusions. Then check each line against that definition. Use the business metric dictionary to keep the shared meaning explicit.

For this example, one point is one standard request's review duration in elapsed minutes. Review start begins the interval. First completed response ends it. The internal target and customer maximum refer to that same interval for each eligible request.

The hypothetical statistical limits also refer to individual durations for this process definition. They are not limits for daily means, percentages, or totals. No calculation method is taught or inferred from the numbers here.

If a chart instead monitors a subgroup statistic, name that statistic. Do not borrow limits from individual observations because their units happen to match. Ask the method owner whether the sampling and aggregation match the baseline.

When two reports disagree about the line's meaning, preserve both versions. Dashboard reconciliation starts with definitions and underlying records. Refreshing a chart will not resolve a disagreement about what its threshold represents.

Walk Through Hypothetical Pre-Established Limits

Assume an earlier, separate baseline has supplied a center line of twenty-four minutes. Its lower control limit is sixteen minutes. Its upper control limit is thirty-two minutes.

These limits are stipulated for teaching. This article provides no baseline observations, estimate, or evidence that they are statistically suitable. The tiny sample below does not establish or recalculate them.

The internal target remains twenty minutes. The invented customer specification remains a maximum of thirty minutes. Keep both requirement lines distinct from the process limits.

Observation Review duration, minutes Within hypothetical control limits? Meets internal target of at most twenty? Meets customer maximum of thirty?
1 23 Yes No Yes
2 25 Yes No Yes
3 22 Yes No Yes
4 26 Yes No Yes
5 24 Yes No Yes
6 23 Yes No Yes
7 25 Yes No Yes
8 24 Yes No Yes

All eight observations fall inside the stipulated process limits. All eight exceed the chosen internal target. All eight meet the fictional customer maximum.

The durations total one hundred ninety-two minutes. Dividing by eight gives twenty-four minutes. The observed range is twenty-two through twenty-six minutes. These calculations describe only the invented sample.

The values look clustered around the hypothetical center line. That stable-looking appearance does not establish statistical control. The table has too little context to evaluate the process history, sampling design, or all applicable signal rules.

The useful business statement is still clear: these eight observations miss the target even though they lie within the assumed control limits. Process behavior and desired performance remain separate questions.

Keep Capability Claims Behind Their Assumptions

NIST's capability guidance compares the output of a stable process with specification limits. Its conventional Cp, Cpk, and Cpm indices assume normally distributed data. The guidance also discusses sample-size and independence considerations.

A capability interpretation therefore needs more than a chart with quiet points. Establish a suitable stable process baseline, comparable observations, an appropriate distributional method, and enough data for that method. Document the specification being evaluated.

This article performs no capability study. The eight fictional values cannot validate the hypothetical baseline. They cannot establish a future proportion meeting the customer maximum. No capability index appears in the example.

Avoid copying a normal-distribution formula into a service report because the spreadsheet offers it. Ask whether the method fits the measurement and observations. If another method is appropriate, have its owner state its assumptions and limits.

Keep descriptive results available while that work remains incomplete. “Eight of eight fictional observations meet the stated maximum” is a count. It is not evidence that every future observation will meet it.

Also keep the internal target separate from the specification used for a capability assessment. A business can pursue a tighter aim than its current customer requirement. The report should show which boundary each conclusion concerns.

Change a Target Through a Recorded Decision

Suppose the fictional sponsor changes the internal target from twenty to twenty-five minutes for a future reporting period. Under that revised target, seven of the eight historical observations would meet the threshold. Observation four, at twenty-six minutes, would still miss it.

That counterfactual count does not mean the process improves. Every recorded duration remains identical. The hypothetical center line and control limits also remain unchanged by the target decision alone.

Keep the effective date and approval beside the new target. Retain the earlier target in the prior report. If you present a historical comparison under the new threshold, label it as a re-evaluation using the revised rule.

Do not present the seven-of-eight count as newly achieved performance. It results from changing the comparison boundary. A later process change needs evidence from the work and its observations.

The customer maximum also stays thirty minutes unless the applicable requirement changes through its own authority. An internal planning meeting cannot silently amend that source document.

If the process itself changes, refer the baseline question to the method owner. Record the change point, retained history, and review decision. Do not recalculate limits merely to absorb an inconvenient point or make the plot look quiet.

Investigate a Signal Without Naming a Culprit

Now add a separate hypothetical observation of thirty-three minutes. It exceeds the stipulated upper control limit of thirty-two. It also exceeds the invented customer maximum of thirty. Record those as separate findings with separate next actions.

NIST's control-chart guidance treats a point beyond control limits as grounds for investigation. It also warns that points inside the limits can still form a nonrandom pattern. Being inside the boundaries alone is insufficient to establish control.

First confirm the source observation and its eligibility. Check the timing events, request classification, and any recorded change in collection. Preserve the point while you investigate it. Do not remove it solely because it changes the appearance of the report.

Then investigate the conditions around the work. A tool interruption, different request type, or changed procedure could matter. Each remains a hypothesis until its source evidence is checked.

An employee name beside the request does not establish that employee caused the duration. Keep the observation, investigation question, and supported finding in separate fields. Any response to the customer requirement belongs to its own owner and process.

The thirty-three-minute example teaches threshold meaning. It makes no claim about false-alarm probabilities or the suitability of these invented limits.

Copy a Limit-Provenance Worksheet

Use one record per line. The following compact table keeps each line's meaning visible without combining three authorities in one ambiguous “threshold owner” field.

Field Internal target Customer specification Process control limits
Measure Individual standard review duration Same defined duration Same defined duration
Units Elapsed minutes Elapsed minutes Elapsed minutes
Line type and value Target maximum: twenty Specification maximum: thirty Lower: sixteen; center: twenty-four; upper: thirty-two
Source Hypothetical sponsor decision Hypothetical agreement clause Hypothetical pre-established baseline record
Baseline or effective scope Current planning period Eligible requests covered by agreement Earlier baseline, not supplied here
Method owner Business sponsor Requirement owner Statistical method owner
Assumptions to verify Each request, same timing definition Applicable request class and requirement version Chart method, stability, sampling, dependence, and distribution as applicable
Review trigger Authorized target revision Requirement or scope change Process, measurement, sampling, or signal-review decision

In a real working record, replace every hypothetical source with its controlled document identifier and version. Add the effective date and evidence location. Leave missing baseline details unresolved instead of inventing them from the plotted lines.

Before distributing the report, ask each owner to confirm the line they govern. Keep those confirmations with the report version. A later reader should be able to reproduce the interpretation without interviewing the original chart maker.

Close the report with two separate statements: what the observations support, and which desired or required boundary they meet. Let each line keep its meaning.

Sources

Topic map: Business reporting and statistics

Articles in this cluster, then the 3 clusters that share the most topics with it.

This cluster: Business reporting and statistics

Neighbor: Spreadsheets, CSV files, and data handoffs

Neighbor: Records, access, and continuity

Neighbor: APIs, webhooks, and automation