Operations & Systems

Missing Dates in Business Charts: Show the Gap Before Drawing the Trend

A line can cross a week the business never measured.

The report contains a value before the gap and another after it. Connect them, remove the empty row, and the chart appears continuous. The drawing now answers a question the source leaves open.

Preserve the missing interval before writing the trend summary. Keep its date, collection status, and reason available to the person building the chart.

The Office for National Statistics (ONS) line-chart guidance recommends gaps where otherwise regular observations are missing. It recommends markers for irregular observations. The handoff method below applies those principles to a fictional workload report. Its fields are an editorial proposal.

Begin With a Calendar and a Measurement Contract

Assume a fictional service team reports completed requests by week. Its agreed window begins Monday at 00:00 UTC and ends the following Monday at 00:00 UTC. The start is inclusive; the end is exclusive.

The regular workload measure counts completion events inside that window. A separate backlog measure counts open requests at a stated instant. The example keeps them in separate columns because they answer different questions.

Week starting Completed requests Collection status Recorded reason Separate backlog observation
2026.09.07 12 Observed Full event window checked None
2026.09.14 0 Observed Full event window checked; no completions None
2026.09.21 Unavailable Missing Completion export not collected None
2026.09.28 18 Observed Full event window checked None
2026.10.05 15 Observed Full event window checked 11 open requests at 2026.10.08 12:00 UTC
2026.10.12 Not applicable Not applicable Fictional pilot suspended for the full interval None
2026.10.19 14 Observed Full event window checked None

All dates, counts, and reasons are invented. The missing export remains missing. The example contains no replacement estimate.

The irregular observation on October 8 is a backlog snapshot. It is not a weekly completion total. Its location inside a row provides context without changing its measurement type.

Leave the Observed Zero on the Chart

The September 14 row reports zero completed requests. In this fictional contract, the event window was checked and no eligible completion appeared.

Plot that observed zero at the zero value. Retain its date in the weekly sequence. The chart can now distinguish an examined interval with no completions from an interval with no usable collection.

The September 21 row has no observed completion total. Its recorded reason is a missing export. Do not plot that row at zero. Doing so would manufacture a week without completions from a collection failure.

The October 12 interval has another status. The pilot's reporting contract does not apply during the suspension. Keep that reason visible. It differs from both the measured zero and the uncollected export.

These distinctions are assumptions established for this example. For a real report, carry the source's documented reason into the chart handoff. If nobody can establish the reason, label it unknown.

A business metric dictionary should define these states before chart production. This review starts after those definitions exist. It checks whether the chart preserves them.

Compare a Bridge With a Visible Gap

Consider the weekly completion line around September 21. The previous observed value is zero for the week starting September 14. The next observed value is 18 for the week starting September 28.

A continuous segment connects those observations across the missing week. On that straight segment, the intervening weekly position lies at nine. Nine is a property of the drawing. The source contains no observed nine-request total.

A visible gap leaves the September 21 position on the time axis and interrupts the line across it. The zero and 18 remain at their observed positions. The caption identifies the missing export.

This treatment follows the ONS recommendation for otherwise regular missing data. It preserves the distinction between observed endpoints and an unobserved interval.

Write the accompanying sentence with the same limit: the available weekly totals are zero and 18, with an uncollected week between them. Do not describe a steady recovery through the missing interval.

This article excludes methods for estimating missing values. A chart author must not invent a measurement to complete the line.

Keep Calendar Distance on the Horizontal Axis

Removing empty rows creates a second risk. A chart may treat the remaining dates as equally spaced category labels.

In the fictional table, September 7 and September 14 are seven days apart. September 14 and September 28 are fourteen days apart. Equal spacing between those three remaining labels would hide that difference.

Use an explicit time axis when elapsed time matters. Retain the expected weekly positions even when the value is unavailable. A reader can then see that one observation is missing between September 14 and September 28.

A label containing a date does not prove that a chart uses elapsed time. Inspect the rendered distances. If a seven-day span and a fourteen-day span occupy the same width, investigate the axis configuration.

Do not infer the behavior of an unfamiliar chart library from its field name. Check its documentation and inspect the rendered output for your configuration. This method specifies the intended result without claiming a universal software default.

When a table is clearer than the chart, publish the table with the period and status columns intact. A continuous line is optional. The missing evidence is not.

Keep Irregular Observations Separate From Weekly Flows

The completion count is a flow across a defined interval. The backlog count is a stock observed at a particular instant. The fictional report records 11 open requests at noon UTC on October 8.

Do not place that 11 at October 5 and call it weekly workload. Do not add it to the 15 completed requests. Open requests at an instant and completion events during a week are different measures.

Use a separate backlog panel or a labeled observation table. Place the snapshot at its actual timestamp. Show its marker so the reader can locate the observation rather than imagine a regularly collected weekly series.

ONS advises using markers for irregular observations to make their timing visible. That guidance does not establish what happened between the observations.

If the next backlog snapshot arrives on October 19, record its timestamp independently. A line between snapshots does not prove the backlog stayed constant or changed smoothly between them.

For this example, retain the backlog as a single labeled point. One snapshot cannot establish a trend.

Document the Window Before Adding Up Periods

A chart handoff needs the boundaries behind each value. “Week of October 5” is insufficient if readers cannot tell whether that date identifies a start, an end, or an observation instant.

For the fictional completion series, record the start and exclusive end of each window. Record UTC as the reporting basis. Keep the backlog timestamp in its own field.

Use the existing time-zone and daylight-saving contract when a real report uses local reporting days. Do not let the chart display silently choose a different period basis.

The observed completion totals in the fictional table add to 59: 12 plus zero plus 18 plus 15 plus 14. That is a subtotal across five observed weeks. It is not a complete seven-week total.

One week lacks collection. Another does not apply. Preserve those two qualifications next to the subtotal. Do not label 59 as all completions across the entire displayed span.

The backlog snapshot of 11 does not enter that addition. Its timestamp and measurement type disqualify it from a completion-event subtotal.

Copy This Chart Handoff Table

Use this proposed table between the data owner and the report author. Retain one row for every expected interval. Add a separate row for an irregular observation.

For this example, place each completion point at its window's start and label it as a weekly total. That position identifies the reporting window. It does not make the total an instantaneous observation.

Interval or instant, UTC Source status Observed value Display treatment Caveat
Sep 7 to Sep 14 Observed weekly flow 12 completions Weekly point End excluded
Sep 14 to Sep 21 Observed weekly flow 0 completions Point at zero Collection checked
Sep 21 to Sep 28 Missing collection Unavailable Gap at weekly position Export not collected
Sep 28 to Oct 5 Observed weekly flow 18 completions Weekly point No evidence for preceding gap
Oct 5 to Oct 12 Observed weekly flow 15 completions Weekly point End excluded
Oct 8 at 12:00 Irregular stock observation 11 open requests Separate dated marker Not a completion total
Oct 12 to Oct 19 Not applicable No applicable value Labeled interruption Pilot suspended
Oct 19 to Oct 26 Observed weekly flow 14 completions Weekly point No continuity claim across suspension

All intervals in this handoff use 2026 dates. The compact labels preserve the full boundaries defined above. In a working handoff, retain full dates and timestamps in the underlying fields.

Add the source snapshot, definition version, chart owner, and review date above the table. These fields let someone reproduce the display without rebuilding the reasoning from a slide caption.

Keep “gap” and “not applicable” distinct in the caption. They can both interrupt a line while recording different reasons. A missing export is not a decision to suspend reporting.

Review the Rendered Chart and Its Caption Together

Review the final display after filters, exports, and layout changes. A correct source table does not prove the rendered chart preserves its missing rows.

Locate September 14. Confirm that the observed zero remains visible. Locate September 21. Confirm that the line stops across the missing interval and the calendar position remains.

Then inspect October 8. The backlog snapshot should appear separately, at its actual observation time. Check that its label says open requests rather than completed requests.

Inspect the suspension interval and the resumed series. The caption should explain why the October 12 period does not apply. It should avoid claiming uninterrupted collection across the suspension.

If a dashboard disagrees, use the dashboard reconciliation process to compare definitions and periods before changing values. One chart may have dropped unavailable rows. Another may use a different reporting window.

Record the discrepancy and the evidence. Do not choose the smoother line because it looks more complete.

State What the Available Observations Establish

For the fictional report, a defensible summary names five observed weekly totals, one missing collection interval, and one interval outside the pilot's reporting contract.

It states that the known totals sum to 59 completed requests across those five observed weeks. It reports the October 8 backlog snapshot separately. It leaves the September 21 total unknown.

Those statements give a decision-maker usable limits. Someone can investigate the export failure or request another backlog observation without treating the drawing as evidence.

Preserve the gap.

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