The Attendance Dashboard That Hid Its Own Gaps
A school's wellbeing team builds a dashboard tracking unexplained absences by year level. Attendance data is exported weekly from three systems: the timetable system (period attendance), the front-office sign-in log (late arrivals), and a paper-based excursion register that a teacher photographs and emails. The dashboard shows Year 9 absences spiking in Term 3. The principal wants to redeploy a wellbeing officer to Year 9 immediately. A teacher notes that Term 3 is when Year 9 has the most excursions, and that the paper register is often entered a week late. The dashboard's y-axis starts at 40% rather than zero, and the 'absence rate' is calculated as unexplained absences divided by total enrolments rather than by total possible attendance days.
1. Evaluate whether the Year 9 spike is a real signal or an artefact. Identify each contributing factor and its direction of effect.
Read the worked answer
The spike is plausibly an artefact, and each factor has a direction:
- Late paper register (timeliness). Excursion days are recorded a week late, so during Term 3 the system under-counts attendance for exactly the cohort with the most excursions. Absences appear higher until the register catches up. Direction: inflates apparent absences.
- Wrong denominator. Dividing by total enrolments rather than possible attendance days ignores that excursion days, part-time students, and enrolment changes alter the base. If Year 9 has more excursion days, its denominator is wrong in a way other year levels are not. Direction: distorts the rate unevenly across year levels, making cross-year comparison invalid.
- Truncated y-axis. Starting at 40% magnifies small differences visually. A 2-point change looks dramatic. Direction: exaggerates the perceived size of any real difference.
- Integration gap. Three systems with no shared key mean late sign-ins may be double-counted or missed, and excursion attendance may not reconcile. Direction: unpredictable, but increases noise.
The honest conclusion: the dashboard cannot yet distinguish a real wellbeing signal from a recording artefact. The correct action is to fix timeliness and the denominator, then re-examine - not to redeploy a staff member on the current chart.
2. Propose the specific data preparation and visualisation changes needed before the dashboard is used for a resourcing decision.
Read the worked answer
Data preparation:
- Establish a single student-day key (student ID + date + session) and reconcile the three sources against it; flag unmatched records rather than dropping them.
- Replace the paper register with a structured digital capture (or at minimum a standard template) so excursion attendance enters the same pipeline.
- Add a timeliness indicator: every dashboard view should show 'data complete to [date]' and suppress or grey out periods with more than X% pending records.
- Recompute the metric as unexplained absences ÷ possible attendance days, excluding approved excursions and adjusting for enrolment changes.
- Run validation checks: no absence on a day the student was on an approved excursion; no negative or >1 rates.
Visualisation:
- Start the y-axis at zero for rate comparisons, or use a dot plot with an explicit note if a truncated axis is unavoidable.
- Show uncertainty or completeness alongside the point estimate, and annotate excursion weeks.
- Compare Year 9 against its own prior terms and against other year levels on the corrected metric.
Only after the corrected metric is stable across two terms should the principal use it to justify redeployment.