Taxi trip data: every regulated ride, one row each

Both cities mandate the reporting, and that single fact explains why the record beats anything built on sampling: there is no opt-out, no recall bias, no panel decay. A demand model trained on it trains on the population.

Illustrative rows in the delivered shape - first a yellow-cab trip:

tpep_pickup_datetime : 2026-05-01 08:12:00   tpep_dropoff : 2026-05-01 08:29:00
PULocationID         : 138                   DOLocationID : 230
trip_distance        : 3.2 mi                passenger_count : 1
fare_amount          : 18.5                  payment_type : 1 (card)
tip_amount           : 3.7   tolls : 0   congestion_surcharge : 2.5
total_amount         : 24.7                  cbd_congestion_fee : 0.75

And a high-volume for-hire trip, the Uber-class record:

hvfhs_license_num    : HV0003 (Uber)         dispatching_base_num : B02617
request_datetime     : 2026-05-01 08:05:00   on_scene_datetime : 2026-05-01 08:11:00
trip_miles           : 4.1                   trip_time : 1020 s
base_passenger_fare  : 16.4                  driver_pay : 13.9
shared_request_flag  : false                 wav_match_flag : false

Read the anatomy rather than the digits - each value above is the documented example for its column. Three habits separate careful users from casual ones. First, payment_type governs the tipping story: card payments auto-populate tip_amount, cash never does, so any tip comparison needs the payment mix stated. Second, geography arrives as zone codes, not coordinates - 258 TLC Taxi Zones in New York, community areas and census tracts in Chicago - which is what keeps the record publishable per trip and makes origin-destination matrices a group-by rather than a GIS project. Third, hvfhs_license_num turns platform share into arithmetic: HV0003 Uber, HV0005 Lyft, HV0004 Via, HV0002 Juno.

All thirty-three cataloged New York fields are verified against the official per-class dictionaries - the reason that record scores a perfect ten on our field-documentation rubric against a catalog average of 7.81 across 1,744 datasets. Chicago's taxi ledger carries twenty-three verified fields; its ride-hail companion twenty-four.

The money columns deserve emphasis. Because fares, surcharges, tolls, taxes and tips arrive as separate fields rather than one blended total, a congestion-pricing study reads directly off rows logged years before the policy existed.

That ladder is the point. The same base serves a two-block late-night pickup audit and a citywide mode-share panel without reconciliation, because both read off identical trip-level rows.

Get a sample of taxi trip data scoped to your zones and months before anything else - the shapes above are what arrives.

Four boundaries define the record, and four named companions fill them.

• Two cities, not a country. Trip-level reporting exists where ordinances created it. For national ground-travel behavior, National Household Travel Survey (NHTS) contributes weighted household, person, vehicle and trip microdata across waves from 1969 to 2022 - its 2022 wave ships 31,074 weighted trip records - the behavioral layer the city ledgers cannot supply.

• No transit context. Curb competition only makes sense against the alternatives. MTA Daily Ridership Data: Beginning 2020 tracks nine modes day by day in the same city as the taxi ledger, including the two congestion-relief-zone counters, while CTA Ridership - Daily Boarding Totals holds 9,312 daily bus-versus-rail rows for Chicago since January 2001 - small enough to open in a spreadsheet.

• No federal benchmark layer. National Transit Database (NTD) Monthly & Annual Ridership Datasets carries roughly 370,000 agency-by-mode-by-month rows since January 2002 across 834 NTD IDs - the denominator for any modal-share claim.

Urban economists and policy analysts treat the congestion fee and surge windows as natural experiments - the fare decomposition lets them isolate policy components from market movement. See market researchers use cases.

Investors and quant researchers read platform share, airport-run volumes and late-night demand as monthly fundamentals for mobility and real-estate theses; the workflow lives at investors quants use cases.

Data scientists and ML engineers get billions of labeled trips with timestamps, geography and outcomes - training substrate for ETA prediction, demand forecasting and dynamic pricing models; see data scientists use cases.

Competitive intelligence teams track operator share and fleet utilization from mandatory filings rather than app audits; see competitive intel product teams use cases.

Journalists and academics cite filed figures instead of press releases - mandated reporting makes attribution trivial.

Datadory handles them upstream of delivery: per-family schemas normalized into one frame, privacy constraints surfaced explicitly rather than discovered mid-analysis, zone and community-area lookups shipped beside the codes they decode, and extracts shaped to the field dictionary above whether they land as API responses, scheduled files or warehouse tables.

Name your zones, months and operators when you request a sample and it arrives already cut to that scope - the production feed follows the same shape, so anything prototyped on the sample survives delivery intact.

Start with the passenger-ground-transportation data hub, which browses all fourteen cataloged records in the industry as products, and the best passenger ground transportation datasets list scoring the top ten side by side.

For depth on the anchors named here: Chicago Taxi Trips vs TNP Ride-Hail Trips Compared settles the taxi-versus-ride-hail question inside one city, NHTS vs NYC TLC Trip Records Compared weighs survey breadth against census depth, the NYC TLC Trip Record Data (Yellow/Green Taxi & FHV/HVFHV Uber/Lyft) product page documents all thirty-three fields class by class, and Chicago Taxi Trips shows how a pseudonymized billion-row ledger reads in practice.