Truxx Freight Intelligence Methodology

One-paragraph summary

What Truxx Freight Intelligence measures

Truxx Freight Intelligence publishes quarterly aggregate statistics computed from actual booked-load data on the Truxx.AI dispatch network. We only publish closed quarters — once a calendar quarter ends, we generate a snapshot and never modify it. Each metric (mileage-weighted average rate per mile, sample size, top corridors) is calculated from the same booked-load fact table that powers our internal dispatch tooling, with the deliberate exclusions described below.

Sample definition

Each snapshot's sample is the set of loads that satisfy all of the following:

• Booked through the Truxx.AI network (not load board listings, not unbooked quotes).
• Equipment class matches the insight series (cargo van for the cargo-van series, etc.). Equipment is taken from the carrier's posted equipment type at booking time.
• Pickup date falls inside the closed quarter (UTC quarter boundaries). A load picked up on June 30 counts toward Q2; a load picked up on July 1 counts toward Q3.
• Reached "delivered" status. Cancellations, no-shows, and still-in-transit loads are excluded.
• Has a non-null booked rate and non-null paid loaded miles. Loads with missing rate or mileage data are dropped to keep the calculation honest, not imputed.

Calculation method

Average RPM (the headline figure)

The headline rate is mileage-weighted: avg_rpm = Σ(total_booked_rate) / Σ(paid_loaded_miles) across all loads in the sample. We weight by mileage rather than averaging each load's per-mile rate because short local moves are priced as flat minimums — a 30-mile load for $550 computes to $18/mi, which is real revenue but a meaningless "rate per mile." Mileage-weighting lets long hauls dominate the denominator so those short moves don't inflate the number. Deadhead miles are excluded from the denominator.

Per-load distribution

Alongside the headline average we publish the per-load p25 / median / p75 (each load's rate / miles as one observation), computed with a standard linear-interpolation quantile. This distribution skews above the mileage-weighted average because short hauls carry high per-mile rates — it's context for the spread of individual loads, not the headline.

Quarter- and year-over-year

qoq = (avg_rpm_current − avg_rpm_prior) / avg_rpm_prior, expressed as a decimal (0.034 = +3.4%). YoY compares against the same quarter one year earlier, which controls for seasonality better than QoQ. Both track the mileage-weighted average.

Trip length

Trip length uses paid loaded miles from the booking confirmation, not broker-quoted miles or actual driven miles (which include deadhead).

Top states / corridors

Origin and destination states come from the pickup and drop ZIP codes mapped to the US state. Corridors are state-to-state pairs (no city-level resolution to avoid identifying specific shippers or receivers).

What's excluded

• Cancelled loads — they distort accepted-rate signal.
• TONU and detention line items — these are real revenue but they attach to broken bookings and would inflate the per-mile number.
• Loads with zero or missing mileage — you can't compute a rate per mile without a denominator, so these are dropped rather than imputed.
• Loads under 25 paid miles — these micro-moves are priced as flat minimums, not per mile; their computed per-mile rate is an artifact, not a rate.
• Per-load rate outliers below $0.50/mi or above $30/mi — almost always data-entry errors, so they're trimmed before any aggregation.
• International loads (cross-border to Mexico or Canada) — we restrict to loads with both pickup and delivery in the US, since cross-border freight follows different pricing dynamics.
• Thin quarters — a trailer type's quarter is only published once it clears a minimum of 20 qualifying loads; below that we omit the quarter rather than publish a number the sample can't support.
• Carrier-specific rate adjustments (volume discounts, contracted bonuses) are applied as part of the booked rate, but private side agreements outside the booking system are not visible to us and not included.

Refresh cadence

We only publish closed quarters. A new snapshot is generated shortly after a quarter ends (typically the first week of the following month) and appended to the time series. There is no mid-quarter refresh, no "quarter-to-date" partial number, and no backfill — once a quarter is published, those numbers are frozen and do not change.

This is deliberate. Rate data is noisy week-to-week, and partial quarters compress seasonality into misleading shapes. Publishing only closed quarters keeps each number defensible and reproducible.

Known limitations

Read these before citing the index in commercial or contractual contexts.

• Sample skew toward dispatched-platform users. Carriers booking through Truxx.AI may have different lane mix and customer profiles than the open spot market. The index is best read as "what comparable network carriers accepted," not "the market-wide spot rate."
• Geographic coverage is uneven. Lanes where we have higher carrier density (Texas, California, Florida origins) carry the most signal. Sparse lanes produce noisier averages and we cap published corridor lists at the top-N to avoid over-reading thin samples.
• Equipment classification depends on carrier-posted equipment. "Cargo van" includes vehicles that some operators would call sprinters; we don't separately segment by GVWR within the cargo-van class.
• Sample size matters more than the average. Always look at the sample-size column before quoting a quarter — a 1,500-load quarter is robust; anything under 300 is directional at best.
• Past performance is not predictive. Rate trends reflect what happened, not what will happen. Use the index to benchmark your own historical rates, not to forecast future ones.