2025 NTD Annual Data Is Arriving: A Five-Step Model Refresh Plan

The 2025 annual NTD release will give agencies and consultants a new validated baseline for service, ridership, finances, assets, and agency characteristics. Used carefully, it can strengthen calibration and peer comparisons. Used as a drop-in replacement, it can quietly break time-series consistency.

Why this release deserves a controlled update

FTA says data collection for Report Year 2025 closed in mid-July 2026 and that the annual products are scheduled for publication in October. That timing makes the release immediately relevant to fall forecasting work, annual performance reviews, grant support, and model-validation plans.

The release also sits alongside reporting changes that apply across Report Years 2025 and 2026. FTA has described changes affecting station and facility reporting, asset categories, selected safety fields, reduced-reporting eligibility, and voluntary-reporter identification. Teams should therefore expect more than new rows: definitions, form relationships, and field availability may also require review.

1. Freeze the current baseline before importing anything

Archive the exact files, transformation code, lookup tables, and model parameters used in the current production run. Record the retrieval date and file version. A reproducible “before” state is essential because FTA may revise historical values and because annual and monthly products are designed for different purposes.

Create a short baseline note listing the current NTD year, the months included in any monthly overlay, known exclusions, and every locally derived field. This turns later differences into traceable decisions instead of unexplained model drift.

2. Compare schemas and definitions before comparing values

Start with the publication guide and data dictionary released with the new files. Compare column names, types, units, reporter coverage, mode and service identifiers, and null-value conventions against the prior release. Do not assume that a familiar label still has an identical reporting basis.

Pay special attention to assets and facilities. FTA's reporting updates consolidate station and facility reporting and add asset detail intended to improve accessibility and state-of-good-repair analysis. Even when a forecasting model does not consume those fields directly, related joins and agency summaries can change.

3. Reconcile annual and monthly series explicitly

FTA's monthly product is published on roughly a two-month lag and is available in both raw and complete forms. The raw file contains reported observations without adjustments or estimates for missing submissions. The complete file can include estimates for recent missing reports and historical adjustments.

Use the annual release to anchor validated levels and agency context. Use monthly data to extend direction, seasonality, and recent service changes beyond the annual period. Keep flags for estimated observations, and do not splice the two series without documenting how overlapping months are resolved.

4. Recalculate the relationships that drive the forecast

Update more than total unlinked passenger trips. Recompute service productivity using Vehicle Revenue Miles and Vehicle Revenue Hours, examine mode-specific changes, and test whether agency or peer-group relationships have shifted. A ridership increase caused by restored service has a different forecasting implication from growth achieved at stable service levels.

APTA's quarterly report is useful as a directional check, but its totals are estimates based on participating agencies and extrapolation. Treat it as a reasonableness test alongside the NTD—not as a substitute for agency-level validation.

5. Run a controlled calibration and publish the change log

First run the existing model with only the data refresh. Then introduce any revised behavioral parameters, socioeconomic inputs, or network assumptions in separate runs. This isolates the effect of the NTD update from unrelated changes.

Publish a concise change log with the resulting forecast: source files, retrieval dates, schema changes, excluded records, treatment of estimates, recalibrated parameters, and the quantitative effect of each major adjustment. Reviewers should be able to distinguish a data revision from a modeling decision.

The practical takeaway

The 2025 annual NTD products are an opportunity to improve model credibility, not merely update a dashboard. A disciplined refresh will show whether a forecast changed because the evidence changed, the reporting structure changed, or the modeler's assumptions changed.

Orion Synergix supports agencies and consulting teams that need experienced transit-data, ridership-forecasting, travel-demand-modeling, GIS, and model-governance capacity.