A route that looks efficient on a map and a route that actually works on the ground are not the same thing. Off-the-shelf routing software optimizes the map. The gap between the two is where a lot of fleet operations quietly lose hours every day.
We haven't shipped a named logistics case study yet, and we won't claim one we don't have. What we can speak to honestly is where routing software actually breaks down, because it's a recurring conversation with logistics clients evaluating whether to build or keep buying.
Key takeaways
- The most common routing failure is not the software's math, it's the underlying address data: a mailing address points to a mailbox, not the actual loading dock, sometimes hundreds of meters away.
- Off-the-shelf tools handle single constraints (time windows, or capacity, or driver hours) reasonably well. Solving all of them simultaneously, the way real dispatch actually works, is where generic tools fall short.
- Gartner projects more than 80% of large enterprises will adopt AI-driven fleet optimization by 2026, a signal the market has moved past treating routing as a solved, off-the-shelf problem.
- Off-the-shelf routing runs $100-500 per vehicle monthly. Custom routing logic tends to pay for itself once a fleet passes roughly 50 vehicles, below that the math usually favors buying.
01The failure is usually the data, not the math
The most common routing failure has nothing to do with the optimization algorithm. A mailing address points to a mailbox, not a loading dock, and the actual delivery entrance can be hundreds of meters away, on a different street entirely. No routing engine, however sophisticated, optimizes around a destination it has the wrong coordinates for.
Off-the-shelf tools inherit whatever address data you feed them. Fixing that requires real-world ground-truthing of delivery points, which most generic routing products treat as the customer's problem, not something the software solves for you.
The market is moving fast enough that this gap matters more, not less. Gartner is reported, via fleet-technology research writeups we reviewed rather than a Gartner page we could open directly, as projecting more than 80% of large enterprises will adopt AI-driven fleet optimization by 2026, treat that specific figure as directional. The scale of enterprise investment in routing technology is the real signal, and it means a routing tool that quietly fails on bad address data will fail more visibly as the rest of the operation gets faster around it.
80%+
Share of large enterprises Gartner projects will adopt AI-driven fleet optimization by 2026
A route that looks efficient on a map and a route that actually works on the ground are not the same thing.
02Where multi-constraint routing breaks
Time windows, vehicle capacity, and driver hours are each solvable on their own. Real dispatch needs all three solved simultaneously, plus live traffic, and that's where a lot of off-the-shelf tools show their limits: multi-depot support, dynamic re-routing, and true enterprise-scale constraint handling are frequently the features that separate a basic routing app from something that survives real operational chaos.
The gap shows up as the same complaint from every fleet manager who's hit it: the plan looked right at 6am and fell apart by 9, because the tool solved yesterday's constraints, not the ones that changed after the first delivery ran late.
03What we build
Address and delivery-point verification built into onboarding, not assumed correct from a geocoding API alone. Multi-constraint solving, time windows, capacity, and driver hours, handled together, not as separate sequential passes.
Dynamic re-routing when conditions change mid-shift, not a static plan generated once at the start of the day and left to fail quietly as reality diverges from it.
04What not to do
Don't assume a routing problem is a software problem before checking the underlying address data. A more expensive routing tool solving on top of bad coordinates just produces confidently wrong routes faster.
We haven't built routing for fleets under roughly 50 vehicles, where off-the-shelf pricing at $100-500 per vehicle monthly usually still makes more sense than a custom build. If that's your scale, we'll say so plainly rather than sell past it.
05Getting started
Audit your actual delivery point data before evaluating any routing tool. This is usually the highest-leverage, lowest-cost fix available and most fleets skip it entirely.
Map your real constraints, all of them together, time windows, capacity, driver hours, before comparing software. A tool that handles one well and the others poorly will show up in daily plan failures, not in the demo.
Run the math on fleet size against the off-the-shelf cost curve. Below roughly 50 vehicles, buying is usually still the right call.
Frequently asked questions
Spot-check a sample of your delivery addresses against actual GPS coordinates of the real entrance or loading dock. If there's meaningful drift, that's your first fix, and it's usually cheaper and faster than switching routing tools.
Roughly 50 vehicles is the rough crossover point where custom routing logic tends to pay for itself against off-the-shelf per-vehicle pricing. Below that, an off-the-shelf tool is usually still the more sensible choice.
An honest audit of your delivery point data and your real operational constraints, not a feature comparison between routing platforms. The constraints and the data quality determine what actually needs building.