The Dashboard Nobody Trusts
A product team reports weekly to leadership from a single dashboard. The dashboard has four charts: (1) tasks by section, (2) tasks by assignee, (3) tasks completed this week, (4) overdue tasks. Leadership has twice challenged the numbers. First, 'completed this week' showed 38 but the team's standup said 22. Second, 'overdue' showed 9 but the team could only find 4. Investigation found: the project has 1,100 tasks, 400 of them completed more than a year ago and never archived; some tasks have no assignee; 'completed this week' counts by completion date but several tasks were completed late and back-dated; and the 'overdue' chart counts tasks in a section called 'Parked - no action' that the team treats as intentionally not-due. The team wants numbers leadership will stop challenging.
1. Diagnose each discrepancy and state the fix, distinguishing between a data-hygiene fix and a filter/definition fix.
Read the worked answer
Two distinct failure classes; conflating them is why the team keeps re-litigating the numbers.
Discrepancy 1 - completed this week: 38 vs 22. Causes: (a) the 400 old completed tasks inflate any unfiltered count; (b) 'completed by completion date' vs the standup's 'what we finished' are different questions - back-dated completions land in the wrong week. This is a definition + filter fix: define 'completed this week' as completion date within the current week and project scope limited to active work, and make sure the team's standup uses the same definition. The 400-task backlog is a hygiene fix - archive completed work older than a threshold so it stops polluting every count.
Discrepancy 2 - overdue: 9 vs 4. The 'Parked - no action' section is intentionally not-due, but the overdue chart has no exclusion. This is purely a filter fix: exclude the Parked section (and tasks with no due date, if those are being miscounted) from the overdue chart. No data is wrong; the chart is asking the wrong question.
Unassigned tasks are a hygiene issue that distorts 'by assignee' - they should either be assigned or explicitly bucketed as 'Unassigned' so the chart sums to the true total.
2. Leadership asks for a single rule: 'make the dashboard match reality.' Explain why no dashboard change alone can achieve that, and what the team must standardize first.
Read the worked answer
A dashboard is a query, and a query can only be as trustworthy as the definitions and the underlying data. Two things must be fixed before any chart is edited:
- Definitions. 'Completed this week' and 'overdue' must have one written meaning each, agreed by the team and leadership. Until 'overdue' excludes parked work by policy, every chart is a coin flip on interpretation. The fix is a short data dictionary: metric name, definition, filters, exclusions.
- Hygiene. Archive stale completed tasks; ensure every active task has an assignee or an explicit 'Unassigned' bucket; stop back-dating completions or accept that completion date means 'date marked complete.'
Only then do you edit the charts - and edit them to implement the dictionary, not to chase the last complaint. The reason leadership keeps challenging is that the team has been patching charts reactively; each patch changes a number without changing the shared definition, so the next number gets challenged too. Standardize definitions and hygiene, then the dashboard becomes a faithful readout rather than a negotiation.