Who benefits

Road travel is the earliest honest reading of the real economy.

People drive before they spend, and they stop driving before any survey records it. Everything on this platform is built from that one measurement how far Americans actually drove, where, and for what.

Below are the seven desks that get the most out of it. Each one starts from a question that is genuinely hard to answer, then names every model that answers it and where in the platform it sits.

282months

of state-level VMT, Jan 2003 to Jun 2026

51jurisdictions

50 states plus the District of Columbia

9scopes

US total and 8 PADD regions, including 1a/1b/1c

6end uses

retail, grocery, parks, transit, workplaces, total

109events

68 hurricane, 28 cold, 13 wildfire, with measured loss

Dec 2027

forecast horizon on VMT and gasoline demand

Coverage verified against the production database, 18 August 2026.

The models below live inside the platform. Everything named on this page is built and running opening it needs an approved account, so the listings here are descriptions rather than links.

Gasoline and refined-product traders

You trade RBOB, gasoline cracks or physical barrels, and you are positioned into the Wednesday EIA release.

The question

Is this week's demand number going to surprise, in which direction, and which region is driving it?

The weekly EIA figure is the thing you are trading against, so it cannot also be your early read on it. Waiting for it means you learn at the same time as everyone else.

Why here

Demand here is built up from measured road travel across 51 jurisdictions, not sampled and grossed up. That means when the national number moves you can open the PADD that moved it, in the same view, before the release confirms it.

What answers it · 5 models

Total US Gasoline DemandMeasured demand with a forecast running ahead of the print.
Fuel Demand Intensity IndexDemand per mile driven separates more driving from thirstier driving.
Seasonally adjusted demand and seasonal factorsWhether a move is the calendar or genuinely new.
Anomaly detectionWhich weeks broke their own pattern, and by how much.
Retail and wholesale price spreadsActual against predicted spread, with the seasonal component split out.

Refiners, blenders and product marketers

You run supply for a refining system or a marketing book and you plan around outages, storms and the summer/winter spec change.

The question

If a hurricane lands in PADD 3, how much demand do I lose, for how long, and how much of the loss is people not driving rather than plant down?

Storm post-mortems usually give you one aggregate loss figure. That is no use for planning, because supply loss and demand loss need opposite responses.

Why here

The demand/supply split is measured, not assumed and it corrects a common error. A cold snap raises heating fuel demand but cuts gasoline demand, because people stop driving; roughly 70% of the total gasoline hit in a cold event is demand, not supply. Every cold week on record moves the same way.

What answers it · 5 models

Weather Impact demand loss in kbdLoss split into demand-side and supply-side, per event and per PADD.
Hurricane Seasons76 seasons of ACE, landfalls and measured loss, coloured by ENSO phase.
Polar Outbreak28 recorded cold events with historical analogs and forward risk by PADD.
ENSO PlaybookHow a La Niña year runs West to East fires, then cold, then Atlantic storms.
Seasonal and regulatory calendarGasoline formulation changes, clock changes and holidays, dated.

III

Macro investors, rates desks and economists

You are trying to date the cycle, and you would rather not wait a quarter for GDP to tell you what already happened.

The question

Is this slowdown broad, or is it three large states dragging the average down?

A national aggregate cannot tell those apart. Both look like the same decline until you can see how many states are falling and how far apart the fastest and slowest have travelled.

Why here

Breadth and magnitude are kept as two separate measurements instead of one blended score, so you can see a narrow decline turn broad. And with 282 months of state history the series covers 2008-09, so the current reading has a real recession to be judged against rather than only 2020.

What answers it · 6 models

Diffusion IndexHow many states or categories are rising, on the familiar ISM 0-100 scale.
Dispersion bandHow far apart the top ten and bottom ten states have travelled.
Mobility Stress and recession signalDiffusion and the dispersion gap combined, switchable across 2020+, 2007+ and 2003+ history.
Composite VMT Leading IndexThe travel-based leading index against NBER recession dates.
Recession probability319 months of modelled probability, back to January 2000.
GDP by PADDRegional growth beside regional travel, on one timeline.

Equity analysts consumer, restaurants and retail

You model quarterly revenue for consumer names and you need the quarter before it is reported.

The question

What did traffic actually do this quarter for the names I cover, and is the softness company-specific or the whole category?

Card panels tell you spend, not visits, and they rarely break out by state. Company guidance arrives after the quarter it describes.

Why here

Because the same travel series drives both the company forecasts and the category view, you can tell a single name's problem from a category-wide one without reconciling two vendors who disagree.

What answers it · 5 models

Restaurant RevenueRevenue history and forecast for MCD, YUM, SBUX, QSR, CMG and DPZ.
Revenue seasonality and guidance comparisonModelled seasonality per name, checked against issued guidance.
Retail & Recreation travelTravel to retail and leisure, by state and by PADD.
Grocery & Pharmacy travelThe defensive category, for read-across when discretionary softens.
Workplaces travelCommuting, which is what weekday lunch traffic depends on.

Auto, EV and fuel-efficiency analysts

You forecast vehicle sales, fleet turnover or the point at which electrification starts to bite gasoline volumes.

The question

How much of the change in gasoline demand is electrification, how much is an ageing fleet, and how much is simply price?

Those three move together, so a scatter plot of any one against demand mostly draws the common trend and tells you very little about the individual driver.

Why here

The fleet-age charts are partial regressions, not scatters: each driver is plotted with the others already accounted for, so the slope you read is the one that variable actually carries. On a simple scatter the same data gives the wrong sign.

What answers it · 5 models

EV Share ForecastMeasured and forecast EV share, nationally and by PADD.
EV fleet and total fleetStock rather than sales what is actually on the road.
Fleet age, partial regressionsFleet age against real gasoline price and real income with the other drivers held fixed.
MPG panelFuel efficiency by region, the bridge from miles driven to gallons burned.
Work-from-home demand dragThe commuting miles that have not come back.

Quant researchers and alt-data teams

You evaluate datasets for a systematic book and most of your time goes on cleaning rather than testing.

The question

Can I get a state panel with a long enough history to backtest, and can I find out what was done to it before I trust it?

Most mobility datasets start in 2020, which gives you one regime and no recession. Most also arrive without provenance, so you cannot tell a real move from a vendor revision.

Why here

Construction is documented, including the awkward parts. The 2003-2006 history is spliced from an FHWA workbook and eight corrected cells are listed on the page with their original values; the Jan 2003 coverage break is the stated reason the panel starts in 2003 rather than 2000. You can audit the series instead of taking it on trust.

What answers it · 5 models

State VMT panel282 months across 51 jurisdictions, monthly, no gaps.
Diffusion and dispersion seriesBreadth and spread as standalone series, both exportable.
Seasonal decompositionTrend, seasonal and residual on a centred 12-month window.
Diffusion histogramThe regime split as a distribution, so the two-state assumption is testable.
Chart-level exportsEvery series on every chart downloads as the numbers behind it.

VII

Transportation, infrastructure and public-sector planners

You forecast traffic for a state DOT, an authority or a consultancy, and your baseline was set before 2020.

The question

Is my state's travel recovering in line with the country, or falling behind it and by how much?

National totals hide the answer, and comparing states directly is unfair unless every state is measured from the same starting month.

Why here

Every state is indexed to one shared base month, so a comparison across states is like for like. Where the underlying federal data changes definition as FHWA did in January 2003 it is handled explicitly rather than spliced over silently.

What answers it · 4 models

VMT Executive SummaryLevel, month-on-month, year-on-year and forecast for any state or PADD.
Position in dispersion bandWhere your state sits between the slowest and fastest, as a percentile.
VMT by end useWhether the shortfall is commuting, retail or transit.
Forecast to Dec 2027A forward path with bands, not just history.

What it does not do.

A dataset is only useful if you know where it stops. Each of these is already stated on the panel it affects, and none of them is a defect being disclosed late they are properties of the underlying data.

Diffusion is measured to Oct 2022, modelled after that

The published end-use diffusion series stops being observed in October 2022; later months, including the current reading, are model output. The 2007+ and 2003+ views avoid this by rebuilding diffusion from state total VMT instead, which is measured throughout.

The Mobility Stress weights are illustrative, not fitted

The 0.65 / 0.35 blend and the 75/25 mix into recession probability come from the specification as written. Nothing in that score is calibrated to our own history, and the card says so.

Recession probability is already a blended figure

The base probability is itself an enhanced, blended number, so adding Mobility Stress puts a second blend on top of a first. It is presented as a research view, not as a replacement for the underlying model.

State VMT starts in 2003, not 2000

FHWA widened reported road coverage in January 2003 by a different factor in every state 9.5x in Rhode Island against 1.7x in South Dakota. Earlier months exist but cannot support a cross-state comparison, so they are excluded.

The dispersion gap level depends on the base month

Each state is indexed to the first month of the window shown, so the gap's level is not comparable between the 2020+, 2007+ and 2003+ views. Diffusion is comparable across all three.

Not sure which of the seven you are?

Most desks span more than one of these. The dashboard carries them all on a single set of filters, so you can hold a state or a PADD fixed and move between travel, price spreads, recession risk and weather without rebuilding the view each time.

Request access Sign in See pricing