Cutting empty miles by 23% sounds like a headline pulled from a vendor's sales deck. So let's start with the mechanics — what empty miles are, why they accumulate even in well-run fleets, and what specifically changes when a Proprietary Algorithm replaces manual load matching. The 23% figure comes out the other end of that explanation on its own.
For a 90-truck bulk fleet starting at 41% loaded miles, a 23% reduction in deadhead translates to moving from 59% empty miles to roughly 49.6% empty miles — a loaded-mile percentage improvement from 41% to 50.4%. At average fuel and maintenance costs across a petroleum fleet, that differential is worth $847,000 per year. Same trucks. Same drivers. Same customer base.
Why empty miles accumulate even when dispatchers are good
Dispatchers don't create empty miles on purpose. The problem is structural, not personal. Bulk dispatch is a combinatorial optimization problem — and the human brain, as capable as it is at pattern recognition, has a hard ceiling on how many combinations it can evaluate under time pressure.
Think about what your dispatcher is holding in their head at 4pm when they're building tomorrow's board. Every driver's hours of service status. Every available load for the morning shift. Customer time windows, site restrictions, equipment requirements, commodity certifications. For 30 trucks, an experienced dispatcher can build a pretty good plan by feel. They know their drivers. They know the lanes. The mental model works.
For 90 trucks, the combinatorial space becomes genuinely vast. The number of possible driver-load matchings for a 90-truck fleet, factoring in HOS constraints and three customer time windows each, runs into the hundreds of millions. Your dispatcher doesn't evaluate hundreds of millions of options — they evaluate the ones that seem most plausible based on experience and get through the exercise in time for everyone to go home. That's not a failure of effort or skill. It's a ceiling that every human planner hits.
A skilled dispatcher can evaluate roughly 25–35 load combinations in a full planning session. Nivio Dispatch's Proprietary Algorithm evaluates millions of combinations against the same constraints — and finds load sequences a human planner would never reach.
Hiring a second dispatcher doesn't fix a planning problem. It just spreads the same manual process across two people, doubles your key-person risk, and adds payroll without expanding the combinatorial search space at all.
The before-and-after: a 90-truck petroleum fleet
Here's what the numbers actually look like when a fleet makes the switch from manual planning to algorithm-assisted dispatch. This analysis is drawn from a 90-truck petroleum fleet (refined fuel, multi-terminal operation) that ran manual dispatch for 6 years before adopting Nivio Dispatch.
| Metric | Manual planning | Algorithm-assisted | Change |
|---|---|---|---|
| Loaded-mile % | 41.0% | 50.4% | +9.4 pts |
| Deadhead miles | baseline | –23.0% | ↓ 23.0% |
| Fuel cost per total mile | $0.487/mi | $0.441/mi | –$0.046/mi |
| Maintenance cost per truck/yr | $18,340 | $15,290 | –$3,050/truck |
| Dispatcher planning time (daily) | 3.5 hrs/day | 31 min/day | –3 hrs/day |
| Annual savings (fuel + maintenance) | — | +$847,000 | +$847K/yr |
The $847,000 breaks down to roughly $573,000 in fuel savings and $274,500 in maintenance cost reduction across 90 trucks. (That math holds for petroleum fleets specifically — frac sand and water hauling have slightly different cost structures, but the directional improvement is consistent.)
Why the math works differently at different fleet sizes
Fleet size changes the ROI equation in a non-linear way — and this is worth understanding before you decide how much this matters for your operation.
On a 30-truck fleet, a 9-percentage-point improvement in loaded miles is worth roughly $235,000 annually (based on typical bulk fleet economics). Meaningful, but not life-changing for most operators.
On a 90-truck fleet — the same percentage-point improvement on three times the trucks and three times the total miles — the same calculation yields $847,000. The rate of improvement is identical. The dollar value scales with fleet size.
| Fleet size | Baseline loaded % | Optimized loaded % | Annual savings (est.) |
|---|---|---|---|
| 30 trucks | 41.0% | 50.2% | ~$235,000 |
| 60 trucks | 41.0% | 50.3% | ~$490,000 |
| 90 trucks | 41.0% | 50.4% | ~$847,000 |
| 120 trucks | 41.0% | 50.6% | ~$1,180,000 |
There's also a second effect at larger fleet sizes: the algorithm finds better plans at scale. With more trucks comes more combinatorial flexibility — more possible load combinations to evaluate. A 120-truck fleet running algorithm-assisted dispatch doesn't just scale linearly from a 30-truck result; it often reaches slightly higher loaded-mile percentages because there are more optimization levers available. This is why fleet operators who make the switch at 90+ trucks frequently see larger-than-expected improvements in the first 60 days.
What actually changes for your dispatcher
The common fear is that optimization software replaces the dispatcher. The actual experience is the opposite. Your dispatcher's expertise doesn't disappear — it gets redirected toward work that requires human judgment.
Before: your dispatcher spends 3–4 hours each afternoon matching drivers to loads, resolving HOS conflicts, and sequencing tomorrow's board. It works, but it's exhausting, it happens under time pressure, and it leaves little room for anything else.
After: the Proprietary Algorithm builds the optimal plan and presents it for review. Your dispatcher spends 25–35 minutes confirming assignments, handling exceptions (the driver who called out sick, the customer who moved a window), and managing the things that require knowing your operation. The afternoon is now available for customer calls, driver conversations, and the relationship work that actually builds a fleet.
Given your available loads, driver HOS clocks, customer windows, equipment types, and commodity certifications, the Proprietary Algorithm evaluates all feasible plan combinations and identifies the one that maximizes loaded miles. It runs each night for the following day's board. Your dispatcher reviews, adjusts for known exceptions, and releases. The plan is better than what manual planning produces — not occasionally, but every day, consistently.
Empty miles are a planning problem, not a lane problem
Most fleet operators, when they think about empty miles, think about lane structure — where the loads are relative to where the drivers are. That's a real factor. But it's not where the biggest gains come from.
The biggest gains come from sequencing. Which driver picks up which load, in what order, after dropping off what previous load, in what geographic cluster. A dispatcher evaluating 30 options won't find the combination where driver 47 repositions 14 miles east before picking up a back-to-back pair that saves 31 miles of deadhead versus the obvious assignment. The algorithm finds it because it has the time and processing capacity to look.
This is why fleets that add lanes or customers sometimes see their loaded-mile percentage go up slightly — more load density in a corridor creates more sequencing opportunities. The algorithm's gains scale with complexity, not against it.
See your fleet's number — free
Send one day of completed loads. You'll get your current loaded-mile %, your deadhead rate, and what an optimized plan for that same day would have recovered — back in 24 hours. No sales call required.
See your number free →The takeaway
For a 90-truck bulk fleet running 41% loaded, cutting deadhead miles by 23% is a $847,000 annual improvement — and it comes entirely from a better planning process, not from adding trucks, drivers, or dispatchers. The constraint isn't effort; it's the number of combinations a human brain can evaluate under afternoon time pressure. A Proprietary Algorithm removes that ceiling. Your dispatcher's job gets better, your loaded miles go up, and the savings compound every day the fleet runs.