The 2025 fiscal year holds a major milestone for agencies invested in automatic passenger counters (APCs): the NTD’s triennial re-certification.

Agency leaders responsible for NTD re-certification likely have their own deadlines circled on the calendar already. Whether as a first-time effort or a repeat certification, there are common data quality challenges that can threaten a successful outcome.

As the former co-founder and CEO of Hopthru and current product manager for ridership solutions at Swiftly, I wanted to share some lessons from helping a multitude of transit agencies achieve successful certifications of their APC data.

To learn more about Swiftly's full-service NTD certification solution, schedule a meeting with a ridership data expert.

Every ride counts with APC data

First, let’s discuss why agencies are moving to APCs as the source of ridership data for NTD reporting.

Ridership data from APCs is more accurate than fare-based ridership, which can undercount actual ridership figures. Passengers with monthly or free passes don’t necessarily tap their cards at every boarding, especially at busy stops where operators wave passengers on to stay on schedule. Moving to APC data for NTD reporting can help agencies access the full resources available to them through federal funding.

For example, San Diego’s North County Transit District (NCTD) found that fare-based data ended up significantly undercounting ridership, in part due to fare-free days and a program for free student bus passes. Comparing fare-based data to APC data processed with Hopthru, NCTD tallied a 7.6% increase in ridership over the same period!

APC data also includes information on alightings, which fareboxes typically do not capture. This metric is necessary for automatically calculating passenger miles traveled (PMT), a required element of NTD reporting.

To hear directly from agencies that have moved to more accurate ridership data with APCs, including NCTD, watch our recent webinar: Make every trip count: Reporting and analyzing ridership for 2025.

Common challenges to NTD certification

Achieving NTD certification for APC data poses its own challenges.

By comparing APC data against manual samples, Full-Reporters must confirm APC data is 95% accurate when calculating unlinked passenger miles (UPT) and passenger miles traveled (PMT). There is also a data coverage requirement. If collecting data for less than 98% of trips, which applies to nearly every agency, the agency needs to apply statistical factors to ensure complete reporting counts. These algorithms, in turn, must receive approval from a qualified statistician.

Within these requirements, we see agencies struggle with common challenges. Successful NTD certification efforts typically include the following:

Proactive hardware monitoring

Agencies cannot expect to gather accurate ridership data without ensuring APC hardware is functioning effectively. Offline or malfunctioning hardware can lead to large gaps in data. To ensure the most robust dataset possible, agencies need to proactively monitor APC hardware health.

Some solutions only monitor the initial connection between the APC and the vehicle logic unit (VLU) at the start of a block. This opens the possibility for multiple types of equipment malfunction that will not be detected until the vehicle returns to the yard and data has already been lost.

Instead, agencies should proactively monitor not only the hardware’s connection status, but also anomalies in data that indicate hardware issues. For example, discrepancies between boardings and alightings, observed APC events without any boarding or alighting activity, and other issues can serve as early warnings that APC hardware is not functioning properly.

Swiftly monitors raw data and hardware status to proactively notify agencies of issues. This helps to avoid gaps in ridership data that may compromise NTD certification.

Precise data processing

In order to meet the 95% accuracy threshold, agencies need to account for APC “noise” that might be falsely detected as boardings and alightings. APC hardware on its own is unable to differentiate between actual passengers and all other activity that might trigger an APC event. Events such as driver activity, passengers getting on and off to ask for directions, movement during layovers, and other activity can result in miscounts.

This is one of the most difficult challenges with APC data, but it’s imperative that agencies get it right. Even slight inaccuracies can have drastic effects on both the final UPT and PMT counts. Swiftly’s algorithms have been tuned to very precisely predict which APC events should be counted as actual passengers vs “noise” that should be discarded.

Additional considerations for data accuracy: Discarding data

Given that APCs count all activities, agencies will always need to discard some problematic data. But how the processor approaches discarding data can have a huge impact on the resulting data completeness.

Swiftly refers to this concept as “coverage scores,” or the percentage of trips run that also collected APC data. Some solutions discard data at an aggregate level. This means that as a part of their processing, they are discarding data from entire trips and blocks, which can result in extremely low coverage scores.

This imprecise approach to discarding data does not necessarily impact an agency’s ability to get certified, but it proves very problematic when trying to use this dataset for planning efforts. Instead, agencies should look to discard data at the individual event level to preserve as much accurate data as possible.

Data expansion

Despite agencies’ best efforts to capture accurate data, expansion will most likely be necessary to ensure the agency is reporting complete counts. Calculating data expansion requires detecting trips not found in APC data and extrapolating the results based on the historical samples for which we do have data.

This step can turn agency staff into data scientists, frustrating certification efforts. If using a custom algorithm, the agency must also receive approval from a qualified statistician.

Swiftly provides data expansion and statistician approval out of the box as part of our NTD certification solution, streamlining the process for agency staff.

A commitment from the processing vendor

Many agencies complete all of the manual counts necessary for NTD certification and then find, to their dismay, that they are not hitting the 95% accuracy thresholds for UPT and PMT. In this case, the agency is wholly reliant on their vendor to fix the processing issues that may exist.

It’s vital that the agency has the vendor’s commitment to achieve the accuracy thresholds necessary to meet the NTD’s standards. Hopthru has never failed a certification across all completed implementations.

Swiftly, your partner in NTD certification

After acquiring Hopthru in 2024, Swiftly now works with agencies to transform APC data into actionable ridership insights and streamlined NTD reporting.

Specifically, Hopthru Cleanse by Swiftly turns raw APC data into accurate, reliable ridership counts that satisfy NTD reporting requirements. Alongside Hopthru Cleanse, Swiftly offers a full-service NTD certification solution that includes:

• Generating a randomized sampling plan that samples each portion of service
• Data expansion approved by a qualified statistician
• An option to manually count passengers with security footage
• A final NTD submission package available for agency approval before submitting

Finally, Swiftly’s team of engineers and data scientists are not only experts in ridership data and the NTD certification process—we’re accountable for your agency’s success. We’ve helped numerous agencies achieve NTD certification and intimately understand how riders are using their system.

Interested in NTD re-certification for the 2025 financial year? Send us an inquiry right away and we will consult with you based on your agency’s financial year.