A duration report says the middle task takes twelve minutes. The person reviewing the queue asks a different question: how many completed tasks took twelve minutes or less?
Answer that question with the observed count and its denominator. An empirical cumulative distribution function, or ECDF, repeats the calculation across the duration scale.
For every threshold, show the share of measured observations at or below it. Keep equal durations together. Preserve the records that produced each step.
Every task, group, and reporting decision below is fictional. The tables are static examples. The worksheet is a proposed review method, with no claim about future completion times.
State the Threshold Convention
Use this definition throughout the example: F(t) = count of measured durations less than or equal to t / measured count. The threshold t is a duration in minutes. F(t) is a fraction from zero to one.
The R stats documentation for ecdf defines the function using this less-than-or-equal convention. It describes a step function whose jump at each duration reflects the number of tied observations divided by the measured count.
A threshold of twelve minutes therefore includes every task recorded at twelve. A strictly-less-than question would exclude those tied tasks and produce a different answer.
Put “at or below” in the chart label. Do not leave readers to infer whether a task on the boundary counts. A review threshold may have a separate business rule. If that rule means strictly under twelve minutes, calculate and label that count separately.
The ECDF asks for a share at a supplied duration. A sample quantile asks for a duration at a selected share. Preserve that distinction when choosing the report. A cumulative percentage beside twelve minutes answers the first question directly.
Define the Completed Records
For this example, one observation is one distinct completed standard task. Duration measures elapsed minutes from recorded start to recorded completion. Waiting stays inside the measurement.
Use the completion window from 2026.10.20 00:00 UTC to before 2026.10.27 00:00 UTC. The extract contains eleven eligible completed tasks. Ten have usable start and completion timestamps. One has no usable start timestamp.
Keep that missing record visible. The distribution below describes the ten measured tasks. It does not describe all eleven eligible tasks as if the missing duration were known.
| Task | Elapsed minutes | Treatment |
|---|---|---|
| E01 | 3 | Measured |
| E02 | 5 | Measured |
| E03 | 5 | Measured |
| E04 | 9 | Measured |
| E05 | 12 | Measured |
| E06 | 12 | Measured |
| E07 | 12 | Measured |
| E08 | 20 | Measured |
| E09 | 28 | Measured |
| E10 | 40 | Measured |
| E11 | Unknown | Missing start; outside measured denominator |
The sorted measured values are 3, 5, 5, 9, 12, 12, 12, 20, 28, 40. Keep all three twelve-minute records. A distinct-value list supplies plotting positions, but it cannot replace the observations used for counting.
The current R-devel documentation says its ecdf function ignores missing values. That implementation behavior does not decide a business reporting policy. This worksheet requires eligible, missing, and measured counts to remain visible regardless of the software used.
Build the Cumulative Table
Start below the smallest measured duration. At zero minutes, no measured tasks qualify. The cumulative count is zero out of ten.
At three minutes, E01 qualifies. At five, E02 and E03 join it. Each distinct duration adds its tied records to the cumulative count. Divide that count by ten each time.
The following table contains the full set of step values. It can accompany a chart or serve as the inspectable report itself.
| Duration threshold, minutes | Tasks exactly here | Tasks at or below threshold | Cumulative share |
|---|---|---|---|
| 0, reference below minimum | 0 | 0 | 0 / 10 = 0% |
| 3 | 1 | 1 | 1 / 10 = 10% |
| 5 | 2 | 3 | 3 / 10 = 30% |
| 9 | 1 | 4 | 4 / 10 = 40% |
| 12 | 3 | 7 | 7 / 10 = 70% |
| 20 | 1 | 8 | 8 / 10 = 80% |
| 28 | 1 | 9 | 9 / 10 = 90% |
| 40 | 1 | 10 | 10 / 10 = 100% |
The exact-duration counts sum to ten. The cumulative count finishes at ten. Each cumulative share uses the same denominator. These checks make the table traceable to the raw list.
The jump at twelve minutes is thirty percentage points because three of ten observations have that duration. The jump at five minutes is twenty percentage points because two observations are tied there.
Never add the cumulative shares down the column. They repeatedly include earlier observations. Their sum has no role in this worksheet. Sum the exact-duration counts when checking that every measured task is represented once.
Read Between the Recorded Durations
At ten minutes, only the values three, five, five, and nine qualify. The cumulative share is forty percent. At eleven minutes it remains forty percent.
At twelve minutes, all three twelve-minute tasks enter together. The share becomes seventy percent. The function stays at seventy percent until the next recorded duration of twenty minutes.
At nineteen minutes, the answer is still seven out of ten. Do not draw a rising diagonal between twelve and twenty and read an invented intermediate share from it. This observed table has no measured duration inside that gap.
For a step chart, put duration on the horizontal axis and cumulative share on the vertical axis. Show the vertical scale from zero to one hundred percent. Keep a marker or clear convention at each included boundary.
Label the chart “Measured completed task durations: share at or below each duration.” Put “ten measured, one missing” in its note. Use the chart axis review before placing it beside another distribution.
Below three minutes, the share is zero. At forty minutes and above, it is one hundred percent for these ten measured observations. That final step states the largest observed value. It sets no ceiling on future work.
Answer the Operating Question Precisely
Suppose the fictional report owner asks whether completed tasks usually fit below a twelve-minute review threshold. Replace “usually” with the count the extract supports.
Seven of ten measured tasks finished at or below twelve minutes. Four of ten finished strictly below twelve. The three tied observations explain the difference between seventy percent and forty percent.
Three of ten measured tasks exceeded twelve minutes: twenty, twenty-eight, and forty. Those records are available for a separate evidence review. The cumulative table does not tell the reader why any duration occurred.
The missing eleventh task limits the all-eligible statement. Among all eleven eligible tasks, seven are known to meet the threshold. Depending on the unknown duration, the final qualifying count could be seven or eight.
That means the all-eligible share could be 7 / 11, about 63.6 percent, or 8 / 11, about 72.7 percent. These are possible completed-data results under the example's one missing duration. They are not confidence limits or a prediction.
For the current report, use the measured-subset sentence and retain the missing count. Ask for the missing timestamp if an all-eligible conclusion is needed. Do not turn the unknown record into a presumed fast or slow task.
Compare Groups With the Same Measurement Boundary
A comparison needs more than two lines on one chart. Match observation type, duration definition, completion window, and missing-record treatment before interpreting their positions.
Use two invented review groups, A and B, working within the same stated window. Group A has the ten measured durations above and one missing duration. Group B has ten eligible tasks, all measured: 4, 6, 6, 10, 14, 14, 18, 22, 30, 35.
Both groups use start-to-completion elapsed minutes, including waiting. The grouping is a fictional label, with no staffing or treatment claim. Report the measured subsets and their different missing counts.
| Threshold | Group A measured count and share | Group B measured count and share |
|---|---|---|
| 5 minutes | 3 / 10 = 30% | 1 / 10 = 10% |
| 12 minutes | 7 / 10 = 70% | 4 / 10 = 40% |
| 20 minutes | 8 / 10 = 80% | 7 / 10 = 70% |
| 35 minutes | 9 / 10 = 90% | 10 / 10 = 100% |
At twelve minutes, A has a larger measured share at or below the threshold. At thirty-five minutes, B has the larger share. Those comparisons follow directly from the supplied values.
Avoid declaring one group faster across every duration. The ordering changes between these displayed thresholds. Pick the question that matters and state the corresponding observed count.
The table cannot establish that an intervention caused the difference. It also cannot establish that future work will preserve these shares. Investigate task composition and measurement evidence separately before explaining a group difference.
Keep the Distribution View Matched to the Question
Use an ECDF table when a reader needs to inspect shares below selected duration thresholds. It avoids choosing intervals for those threshold counts. Each observed value can be checked against the original records.
Use a histogram when the question concerns counts inside duration intervals. Record the histogram's edges and boundary rules. Its bars answer a different reading task.
Use the mean and median review for central summaries. A median can accompany this table, but it does not replace the count at twelve minutes. A single middle value cannot display all the steps.
None of those views preserves the order in which tasks occurred. Keep the original timing records if the next question concerns sequence. A distribution can describe completed values while leaving the operational chronology unresolved.
Copy the Cumulative Distribution Worksheet
Use this proposed worksheet before writing the descriptive conclusion. Replace each fictional entry with a checked reporting definition.
| Field | Entry for Group A |
|---|---|
| Eligible observations | Eleven distinct completed standard tasks |
| Completion window | 2026.10.20 00:00 UTC to before 2026.10.27 00:00 UTC |
| Measure and unit | Elapsed start-to-completion minutes, including waiting |
| Missing records | E11 lacks usable start; duration unknown |
| Measured denominator | Ten valid durations |
| Threshold convention | Include every measured duration less than or equal to t |
| Source version | synthetic-ecdf-v1; preserve E01 through E11 |
| Distinct plotted durations | 3, 5, 9, 12, 20, 28, 40 |
| Exact counts | 1, 2, 1, 3, 1, 1, 1 |
| Cumulative counts | 1, 3, 4, 7, 8, 9, 10 |
| Plotted shares | 10%, 30%, 40%, 70%, 80%, 90%, 100% |
| Software record | Hand calculation; no installed R version assumed |
| Descriptive conclusion | Seven of ten measured tasks finished at or below twelve minutes |
| Limit and next question | One eligible duration is unknown; inspect its timestamp evidence |
Store the worksheet with the raw values and the published table. If a missing duration becomes available, create a revised extract and recalculate every affected share. Record the new denominator before explaining the change.
Count the observations. Name the boundary. Keep the unknown record visible.
Sources
- R stats: Empirical Cumulative Distribution Function. Less-than-or-equal definition, tied jumps, and missing-value behavior. R-devel documentation identifies package stats version 4.6.0 when checked on 2026.10.28. No installed-version claim is made. The datasets, reporting rules, tables, and worksheet are original editorial examples.