APC Spike Before the NTD Report
A mid-size agency uses automatic passenger counters on 120 buses. The monthly ridership report is due to the National Transit Database, and the raw APC data shows an 18% increase over the same month last year. Manual ride checks on 20 representative trips show only a 4% increase. GPS logs show that 12% of APC records lack a valid stop event, and a firmware update six weeks ago changed the door sensor debounce setting. The CFO wants to report the higher number because it supports a budget request. The NTD reporting officer wants an accurate submission. The data team has two weeks before the deadline and limited staff to recheck every trip.
1. Diagnose the likely cause of the discrepancy and explain why the raw number is not reliable.
Read the worked answer
The raw APC number is not reliable because two data-quality signals point to a measurement problem rather than a real ridership surge. First, 12% of records lack a valid stop event, which means the system cannot reliably associate boardings and alightings with stops; those records may be duplicated, assigned to the wrong trip, or counted without matching exits. Second, the firmware update changed the door sensor debounce setting six weeks ago, which is close to the start of the anomaly. A debounce change can cause the counter to register multiple events for one door cycle or miss events depending on timing. The manual checks, while a small sample, provide an independent estimate closer to 4%. The 18% figure therefore contains a likely instrumentation bias and should not be submitted without validation.
2. What should the agency report and how should it document the decision?
Read the worked answer
Report a validated estimate, not the raw 18%. Use the manual sample to adjust the APC data, and disclose the method: sample size, routes, time periods, and the adjustment factor. If the sample is too small for a formal expansion, report the lower validated figure with a note that APC data are under review, or use the prior year's method consistently. Do not submit a number chosen to support a budget request; NTD reporting must be accurate and consistent with the agency's documented methodology. Document the firmware change, the missing stop-event rate, the manual check results, and the corrective action plan. Notify the NTD reviewer if an amended report may be needed. Preserve the raw data and the audit trail.
3. Recommend immediate and longer-term corrective actions.
Read the worked answer
Immediate: roll back or reconfigure the firmware on a test group, compare counts against manual checks, and validate the door debounce setting. Prioritize routes with the largest variance and the highest ridership. Longer term: add automated data-quality checks for missing stop events, impossible passenger loads, and door-cycle anomalies; require a pre-deployment validation period for firmware changes; and maintain a standing manual-count sample for calibration. Assign ownership of APC data quality to a named analyst and create a monthly dashboard that flags variance from manual checks. Train maintenance and operations staff on how firmware changes affect data. Finally, separate the budget advocacy process from the ridership reporting process so that reporting decisions are not influenced by funding goals.