What manual routing is actually costing you
Most water treatment businesses with delivery routes — salt delivery, bottled water, filter exchanges — build their daily schedules the same way they did ten years ago. Someone comes in, looks at the list of stops, and figures out an order that seems reasonable. Maybe they use Google Maps to check a few legs. It takes one to three hours and the result is a route that works but isn't optimal.
The cost of that process is real and it shows up in three places:
A manually planned route covers more miles than necessary. Not dramatically — maybe 15-20% more than an optimized route. But across a five-day week, multiple trucks, that's hundreds of extra miles per month.
More miles means more hours. A driver who could complete 14 stops in a shift completes 11 because the routing wasted two hours of drive time. That's three missed revenue-generating stops per truck per day.
The two to four hours spent building routes every morning is administrative overhead that generates no revenue. It's also the kind of repetitive cognitive work that burns people out faster than anything else in an office job.
There's also a hidden cost that doesn't show up on any report: when a customer cancels or a driver finishes early, there's no mechanism to automatically fill the gap. The route just runs with a hole in it.
How AI route optimization actually works
The short version: AI routing engines evaluate millions of possible stop sequences simultaneously, applying a set of constraints you define, and return the sequence that minimizes total cost — usually measured as a combination of distance, time, and fuel.
The constraints that matter for water treatment businesses include:
- Time windows — customers who are only available between 8am and noon, or who need to be last because they take longer
- Vehicle capacity — how many bags of salt, how many bottles, how much equipment fits in each truck
- Driver availability — start and end locations, shift lengths, break requirements
- Priority stops — a customer whose tank is critically low gets scheduled before someone who just needs a routine top-off
- Traffic and road conditions — real-time data that adjusts the route as the day progresses
The difference between this and Google Maps is that Google Maps optimizes a single route for a single driver. AI routing engines optimize across your entire fleet simultaneously, accounting for all of the constraints above, in under ten seconds.
AI routing doesn't just find the shortest path. It finds the path that satisfies all your constraints while minimizing total operational cost. Those are different problems, and the second one is far more valuable for a business with multiple trucks, mixed cargo, and variable customer windows.
Why water treatment is a specific use case — not a generic delivery problem
Most route optimization content is written for e-commerce last-mile delivery: lots of small packages, standard vehicles, simple one-time stops. Water treatment is different in ways that matter for which tool you choose.
A single truck might carry salt bags, replacement filters, RO membranes, and a water softener unit for a new install. These have different weights, dimensions, and handling requirements. The routing system needs to account for load configuration, not just stop count.
A salt top-off takes ten minutes. A softener install takes three hours. A service call could be anywhere in between. Generic routing tools assume similar stop durations. Tools built for service businesses handle variable time windows correctly.
If you're using IoT brine sensors (covered in a separate guide), delivery demand is generated dynamically when a tank hits a threshold — not on a fixed calendar. Your routing system needs to incorporate same-day additions without rebuilding the entire schedule.
Bottled water and cooler businesses often have returns — empties to collect, equipment to swap. The routing system needs to account for pickup weight as well as delivery weight, which changes the vehicle capacity calculation mid-route.
Onfleet vs Timefold vs SimpliRoute
These three tools cover the realistic range for a small-to-midsize water treatment operation. There are larger enterprise platforms — Descartes, Oracle Transportation Management — but they're priced and scoped for national fleets, not a regional water treatment business with two to eight trucks.
- Clean dispatcher interface — low training curve for office staff
- Real-time driver tracking with automatic customer notifications ("your driver is 3 stops away")
- Route optimization handles time windows and capacity constraints reliably
- Proof of delivery — photo capture, signature, barcode scan
- API access on all plans for connecting to your FSM or CRM
- Good mobile app — drivers need minimal training
- Optimization algorithm is solid but not the most sophisticated — works well for standard routes, less so for highly complex constraint sets
- Pricing jumps significantly between tiers
- Multi-compartment vehicle loading not natively supported
$99/mo (Launch) · $349/mo (Scale) · Custom above
Launch covers up to 2,000 tasks/month. Most small operations fit here comfortably.
- Most sophisticated constraint solver of the three — handles complex multi-vehicle, mixed-cargo scenarios better than competitors
- Open source core with a commercial cloud layer — can self-host if you have the technical capability
- Handles multi-compartment loading constraints natively
- Good fit if you're building a custom integration with an existing FSM
- Scales to very large fleets without pricing becoming prohibitive
- More technical to set up — not a point-and-click tool
- Less polished driver-facing mobile experience
- Requires more configuration to get optimal results for your specific constraint set
Free open source · Cloud from ~$150/mo · Enterprise custom
The open source version is genuinely usable but requires technical setup. Cloud version simplifies this significantly.
- Fastest to get running — upload a spreadsheet of stops, get an optimized route
- Intuitive interface designed for non-technical users
- Good for businesses just starting with route optimization
- Handles basic constraints — time windows, vehicle capacity, driver availability
- Reasonable pricing at entry tier
- Less sophisticated optimization than Onfleet or Timefold for complex scenarios
- API access limited on lower tiers
- Real-time rerouting less robust than Onfleet
From ~$59/mo depending on region and fleet size
Pricing varies — get a direct quote for your fleet size.
| Feature | Onfleet | Timefold | SimpliRoute |
|---|---|---|---|
| Setup complexity | Low | High | Very low |
| Optimization quality | Strong | Best in class | Good |
| Mixed cargo / multi-compartment | Partial | Native | Basic |
| Real-time rerouting | Yes | Yes | Limited |
| Driver mobile app | Strong | Basic | Good |
| API / FSM integration | All plans | All plans | Higher tiers |
| Starting price | $99/mo | Free / $150/mo | ~$59/mo |
| Best for | 1–4 trucks, standard routes | Complex constraints, technical team | Getting started fast |
How to set this up without breaking your operation
The biggest mistake businesses make with route optimization is trying to switch everything at once. Run the AI route alongside your manual route for one week first. Compare the results. Build confidence before you hand the dispatcher's morning over to the software.
Every tool accepts a CSV or spreadsheet upload. Your FSM almost certainly has an export function. You need: customer name, address, delivery window if applicable, estimated stop duration, and any priority flags. Start with one route — your most predictable one.
Tell the system what matters: vehicle capacity in your units (bags, cases, pounds — whatever makes sense), driver start and end locations, shift length, and any time windows your customers have given you. Be conservative on capacity — leave a 15-20% buffer for the first few weeks.
Generate the AI route. Have your dispatcher review it and note any problems — a customer who's actually only available in the morning despite no window being set, a stop that takes longer than the default duration, a road that looks short on a map but isn't in practice. This is calibration, not failure.
Feed the week's observations back into the system as refined constraints. Then cut over. The dispatcher's role shifts from building routes to reviewing and approving routes — a 10-minute job instead of a 2-hour one.
Once the routing is working, connect it via API to your field service management software so stops flow in automatically without manual CSV exports. Onfleet and Timefold both have good API documentation. This is where the real time savings compound.
The ROI calculation for a small water treatment operation
Use the calculator below to estimate your potential return. The defaults are conservative estimates based on industry averages — adjust them to match your operation.
Start with the basics. How many trucks are you running, and how many days a month do they go out?
Route optimization typically reduces mileage by 10–20%. Enter your current fuel spend and a conservative reduction estimate.
Manual routing takes time — usually 20–30 minutes per truck each morning. Automation recovers that. What does your dispatcher cost per hour?
Tighter routes create room for additional service stops. Even one or two extra jobs per day across the fleet adds up significantly.
Bottom line
If you have more than one delivery truck and your dispatcher is still building routes manually every morning, this is the first AI tool worth implementing — not because it's the most exciting, but because the ROI is the most immediate and the setup is the least disruptive.
Start with Onfleet if you want something running fast. Start with SimpliRoute if you want to prove the concept with minimal commitment. Come back to Timefold if your operation has genuinely complex routing constraints that the other two can't handle cleanly.
The parallel-run approach eliminates implementation risk. One week of running both gives you data and confidence before you fully cut over.
Route optimization compounds when combined with IoT brine tank sensors — the sensors generate dynamic delivery demand, the routing engine incorporates it in real time, and your drivers never show up to a tank that doesn't need a refill. That combination is covered in the salt delivery sensors guide.
A real person will read your message and write back. No auto-responder.