Fuel & idle intelligence
Stop guessing where your fuel is going.
Idle detection, fuel analytics, card matching, and battery monitoring in one dashboard. Built for small and growing fleets. Included in Take Command at $28/vehicle/month.
Answers that used to take a spreadsheet
Illustrative data. Your fleet's actual numbers appear in the dashboard after install.
Why Threshold-Based Fuel Monitoring Fails
Catching waste, theft, and problems early only works if the alerts are real. A fuel sensor is noisy, fluctuating with road grade, cornering, temperature, and fuel slosh. Cheap systems just watch for a fixed drop, so they cry wolf or miss the real thing.
A vehicle rounds a corner, the sensor sloshes, and the system fires an alert. Every time.
The vehicle moves for 90 seconds while filling, and a threshold system logs it as a trip, not a fill.
Without a calibrated baseline, raw MPG numbers aren't actionable.
Every reading is treated equally, so a solid calibrated reading looks the same as a rough guess.
Fleet Command Center runs a five-stage signal pipeline on every fuel reading. It knows the difference between a hill and a theft, and it tells you how confident it is in every alert, so you act on the ones that matter.
The River + Lake Pipeline
Two systems for two jobs. The River processes every reading in real time through five stages before storing it. The Lake rolls those events up hourly and daily against the immutable River ledger.
The River (Real-Time) · 5 Stages
Sample Normalization
Raw fuel percentages are validated against physical limits. Readings outside 0-100%, deltas >30% in <60 seconds, and future-dated timestamps are rejected before they corrupt your metrics.
Tank Calibration
Percentages become gallons using a four-tier confidence hierarchy: fuel-card-confirmed calibration curves, then your stored tank capacity, then a generic tank-size default, then linear estimation. Each source carries a confidence rating.
Kalman Fusion
A Kalman filter fuses GPS motion, engine RPM, throttle %, and road grade to smooth noise while preserving real spikes. Refuel detection runs BEFORE smoothing, since a true fill would otherwise be averaged away.
Environmental Enrichment
Each sample is annotated with NWS weather (temperature, humidity), road grade (elevation delta between GPS points), and load when available. This tells a mountain pass apart from hard acceleration.
Event Detection
A per-vehicle sliding window runs continuous anomaly detection across refuel, theft, sensor error, tank switch, and mechanical warning, each with a confidence score.
| Event | Trigger Condition | Signal Strength |
|---|---|---|
| Refuel Detected | Fuel level increase while vehicle is stationary | High |
| Theft Suspected | Fuel loss with engine off | High |
| Theft Suspected | Rapid fuel loss in short window | Elevated |
| Sensor Error | Physically impossible reading pattern | High |
| Tank Switch | Multi-tank level change pattern | High |
| Config Missing | Tank capacity unavailable | Flagged |
| Mechanical Warning | Sustained MPG decline vs. baseline | High |
The Lake (Batch Analytics)
Hourly and daily jobs run against the immutable River ledger, producing ready-to-read fleet MPG, consumption, idle cost, and efficiency scores. Lake jobs are idempotent, so running the same aggregation twice gives identical results.
More Than Raw Numbers
Fuel Level (Calibrated, Not Raw)
We don't report raw OBD percentages. We convert to gallons using calibrated tank data, smooth it, and attach a confidence score. A 0.97 reading came from a calibration curve checked against fuel-card fills; a 0.55 reading came from a generic tank-size estimate. That difference matters when you act on the number.
Burn Rate (Per Vehicle, Per Trip)
Burn rate, in gallons per hour, tracks how fast fuel is being used and updates continuously with a moving average that smooths spikes but stays responsive to real change. It anchors fuel-out prediction and driver attribution.
Efficiency Score
Each vehicle's MPG is scored against a blended baseline: EPA lab data, EPA real-world data, your own same-model vehicles, and that vehicle's own 30-day history. Above 1.0 = better than expected, below 1.0 = underperforming. More useful than raw MPG because it accounts for vehicle type, terrain, and expected performance.
Theft. Refuel. Mechanical Decline.
Three kinds of anomaly, each with its own detection logic, evidence, and severity tiers, so you catch problems early.
Every alert includes before/after fuel levels, estimated gallons lost, GPS evidence (last 5 samples with timestamps), and engine state.
Every refuel record logs estimated gallons added and station coordinates. Connect a fuel card and these events are auto-matched against card transactions to flag mismatches.
When MPG drops more than 10% vs. the prior 30-day average (same route and load), it raises a mechanical warning, the kind of signal you see with a failing O2 sensor, dirty injectors, a clogged air filter, or transmission slip.
Not Just Who Used It. Why.
Each behavior is priced against current regional fuel cost to give you a dollar figure, so you see which drivers cost the most and exactly what to coach.
| Behavior | Measurement | Fuel Cost Attribution |
|---|---|---|
| Idling | Burn rate × idle seconds | Gallons wasted at idle |
| Speeding | MPG penalty per 10 mph over limit × miles | Gallons lost to aerodynamic drag |
| Hard acceleration | Per event count | 0.05 gal/event |
| Hard braking | Per event count | 0.02 gal/event |
| Route detour | Unplanned miles ÷ baseline MPG | Gallons for unplanned distance |
| Engine over-rev | Minutes over RPM threshold | 0.08 gal/minute |
See It Before It Happens
- Triggers when burn rate projects empty in under 2 hours, or level drops below 10%
- Estimated hours and miles remaining, plus a clear recommendation
- Confidence 0.80 to 0.95, higher when recent burn data is dense
- Weighs candidate stations against vehicle position and route
- Accounts for current burn rate and regional fuel pricing
- Top 3 options ranked by total cost (price + detour time)
- 30, 60, and 90-day fleet fuel cost projections, refreshed hourly
- Per-vehicle burn rates and regional EIA fuel pricing
- Risk factors flagged: price volatility, driver behavior trends
Fuel Card & IFTA
Fuel Card Reconciliation
Mismatches generate a card_mismatch event with the card transaction ID and nearest gauge event for human review.
IFTA Quarterly Reporting
Exported in the format needed for IFTA filing, from the dashboard or the API.
One Dashboard. Every Tank.
Got vehicles with more than one tank? Fleet Command Center handles primary diesel, auxiliary saddle tanks, and DEF systems. Each is calibrated on its own and rolls into one total consumption record.
An Immutable Record of Everything
Every fuel event is written to an append-only, tamper-evident ledger.
Immutable
Written with append-only access controls. No event is ever modified or deleted.
Idempotent
Each event carries a unique identifier. Duplicate writes are rejected without error.
Versioned
Every event records which version of the River pipeline emitted it, enabling replay with updated logic.
Feedback LoopWhen you dismiss a false alert, that feedback (outcome, reason, attribution) becomes training data, so accuracy keeps improving for your specific fleet.
Every Metric Includes Provenance
Every fuel metric from the Fleet Command Center API carries its provenance: how confident the system is and where the number came from. Most tools never tell you that.
{
"value": 7.8,
"confidence": 0.85,
"source": "lake_projection"
}- value
- The metric itself: gallons, MPG, burn rate, cost, and so on.
- confidence
- How sure the system is, from 0.55 (generic default) to 0.97 (card-validated calibration).
- source
- Where it came from: river_event (real-time), lake_projection (batch), or epa_estimate (baseline).
Ready to see it live?
Try the demo with a mock fleet before buying any hardware. No signup required.