---
title: "Route Optimisation for Southeast Asia: Why Address Quality Caps Your Routing Gains in 2026"
id: "26456"
type: "post"
slug: "route-optimisation-southeast-asia-geocoding-2026"
published_at: "2026-09-09T13:00:00+00:00"
modified_at: "2026-09-10T06:06:53+00:00"
url: "https://locus.sh/blogs/route-optimisation-southeast-asia-geocoding-2026/"
markdown_url: "https://locus.sh/blogs/route-optimisation-southeast-asia-geocoding-2026.md"
excerpt: "In Southeast Asia the solver is rarely the constraint. Geocode error relative to drop spacing decides how much routing optimisation can return at all."
taxonomy_category:
  - "General"
---

#### [General](https://locus.sh/blogs/category/general/)

# Route Optimisation for Southeast Asia: Why Address Quality Caps Your Routing Gains in 2026

[Aseem Sinha](/author/aseem_locus/)

Sep 9, 2026

15 mins read

Route optimisation is the process of converting a set of orders into a sequenced, capacity-feasible plan for each vehicle. It takes coordinates as an input and treats them as true. In Southeast Asia that assumption is the weakest part of the system, because a large share of delivery addresses are described by landmark, alley and relation rather than by a numbered position on a mapped street, and the coordinate the planner receives is an estimate with an error attached.

That matters more than it sounds. A solver given noisy coordinates still returns a mathematically optimal answer. It is simply optimal for a set of points that does not exist. The route looks correct on the screen, the sequence looks sensible, and the rider discovers the discrepancy one stop at a time. The useful question for an operator in Jakarta, Manila, Ho Chi Minh City or Bangkok is not which solver is strongest. It is whether coordinate quality is good enough for sequencing to carry any information at all.

## Key Takeaways

- Compare geocode error against mean drop spacing, not against a map. The ratio between them decides whether sequencing carries information.
- Published research on residential geocoding in well-mapped markets finds typical positional errors of 26 to 201 metres, and Southeast Asian addressing conditions are structurally harder than that baseline.
- Higher drop density makes geocode error worse, not better, because density shrinks the spacing the error is measured against.
- Geocode error surfaces as lost drop capacity. On a 40-stop route, 15% of stops needing a six-minute search costs about 13% of the route, or five drops.
- Where the error-to-spacing ratio exceeds roughly one, solver improvement returns almost nothing until coordinates are fixed.

## Why Southeast Asia is a different routing problem

The commercial stakes are large and growing. The tenth edition of the Google, Temasek and Bain [e-Conomy SEA report](https://www.temasek.com.sg/en/news-and-resources/news-room/news/2025/e-conomy-sea-2025-report-aseans-digital-economy-poised-to-surpass-300-billion)
 put the region’s digital economy on track to pass 305 billion dollars in gross merchandise value in 2025 against 135 billion in revenue, both growing around 15% year on year, with coverage expanded from six markets to ten. Physical delivery is the settlement layer for most of that, and it runs on addresses.

Addressing is a recognised infrastructure gap rather than a data-hygiene problem. The Universal Postal Union runs a dedicated [addressing programme](https://www.upu.int/en/Postal-Solutions/Programmes-Services/Addressing-Solutions)
 precisely because many member countries lack a countrywide address infrastructure, and maintains S42, an international addressing standard, because national formats are not interoperable. An enterprise operating across six or ten Southeast Asian markets is therefore not handling one address format with local quirks. It is handling several addressing systems with different underlying completeness.

The measurement problem is well documented even in easier conditions. Research comparing [geocoding solutions for residential addresses](https://www.tandfonline.com/doi/full/10.1080/19475683.2015.1085437)
 reports typical positional errors ranging from roughly 26 to 201 metres, with urban medians commonly between 30 and 170 metres and rural medians substantially worse, one dataset showing 173 metres rural against 25 metres non-rural. Work on [error propagation in geocoded datasets](https://pmc.ncbi.nlm.nih.gov/articles/PMC3319659/)
 shows that positional error does not stay contained: it propagates into every downstream calculation built on the coordinate. Those figures come from well-mapped residential markets, which makes them a floor rather than an estimate for landmark-addressed alley networks.

And the prize that Southeast Asian operators are chasing is exactly the thing that amplifies the problem. McKinsey’s work on [out-of-home delivery](https://www.mckinsey.com/de/publikationen/2024-10-28-ooh-delivery)
 finds that raising drops per stop from one to five cuts labour and vehicle cost by more than 50%. Density is the dominant lever in the region’s dense urban cores. Density also shrinks the distance between consecutive stops, which is the yardstick geocode error is measured against.

One regional feature multiplies the cost of every one of those errors. Where cash on delivery remains a significant share of orders, a stop the rider cannot locate is not simply a redelivery. It is an uncollected payment, a cash position that does not reconcile at end of day, and a settlement exception that someone has to clear manually. The same coordinate error that costs six minutes in a card-paid market costs six minutes plus a finance workflow in a cash-paid one, which is why address quality carries more weight in Southeast Asian unit economics than the equivalent metric in North America or Western Europe.

| Also Read: Route Optimization: The Complete 2026 Guide |
| --- |

## How to diagnose whether your coordinates support optimisation

### 1. Calculate mean drop spacing for the route, not for the city

Take the service area and the stop count. For stops distributed across an area, mean nearest-neighbour spacing is approximately half the square root of area divided by stops. A 40-stop route across five square kilometres gives roughly 177 metres between neighbouring stops. That number, not the city’s size, is the scale at which sequencing decisions are made.

### 2. Estimate your geocode error honestly

Use the difference between the planned coordinate and the position where proof of delivery was actually captured, aggregated across thousands of completed stops. That distribution is the real error, it is already sitting in the execution data, and it is far more useful than a vendor’s stated match rate. Match rate says a coordinate was returned. It says nothing about whether the coordinate was right.

### 3. Divide one by the other

The ratio of geocode error to mean drop spacing is the diagnostic. Below roughly 0.3 the planned sequence largely survives. Around 0.5 to 1.0 sequencing information is substantially degraded. Above 1.0 the sequence is dominated by noise, because the error is larger than the distance between the stops being sequenced.

| Stops in 5 sq km | Mean spacing | Error 20 m | Error 50 m | Error 150 m | Error 500 m |
| --- | --- | --- | --- | --- | --- |
| 10 | 354 m | 0.06 | 0.14 | 0.42 | 1.41 |
| 40 | 177 m | 0.11 | 0.28 | 0.85 | 2.83 |
| 80 | 125 m | 0.16 | 0.40 | 1.20 | 4.00 |
| 150 | 91 m | 0.22 | 0.55 | 1.64 | 5.48 |
| 250 | 71 m | 0.28 | 0.71 | 2.12 | 7.07 |

### 4. Note that density works against you here

This is the counterintuitive part, and it is the reason the problem is worst in exactly the markets that look most attractive. Holding the service area and the error constant, moving from 10 stops to 150 stops raises the ratio from 0.42 to 1.64, a 3.9 times increase. The density that cuts cost per drop is the same density that destroys the sequencing signal, so in a dense Southeast Asian core the two effects arrive together and only one of them is usually measured.

### 5. Account for the fact that reachability depends on the vehicle

A coordinate is not equally reachable by every asset. A two-wheeler can enter alley networks and gated lanes that a van cannot, so the effective delivery position for the same address differs by vehicle type: the motorcycle reaches the door, the van reaches the mouth of the lane a few hundred metres away. This means geocode error and fleet mix interact. A plan that assigns a van to a cluster of alley addresses inherits a systematic positional offset that has nothing to do with geocoding quality and everything to do with which vehicle was chosen, and it will present in the data as poor address accuracy for that route. Any error measurement worth acting on has to be segmented by vehicle type before it is compared against spacing.

### 6. Convert error into the currency that gets funded

Geocode error does not appear in reporting as a data problem. It appears as service time. Take a 40-stop route with three minutes of service and four minutes of riding between stops, a 280-minute baseline.

| Share of stops needing a search | Search time | Added to route | Share of route | Drop capacity lost |
| --- | --- | --- | --- | --- |
| 5% | 4 min | 8 min | 2.9% | 1.1 stops |
| 10% | 6 min | 24 min | 8.6% | 3.4 stops |
| 15% | 6 min | 36 min | 12.9% | 5.1 stops |
| 15% | 10 min | 60 min | 21.4% | 8.6 stops |
| 25% | 8 min | 80 min | 28.6% | 11.4 stops |

Five lost drops on a 40-stop route is a 12.5% capacity reduction that no routing report attributes to addressing, because the time was booked as service time.

### 7. Sequence the investment accordingly

Treat the usable sequencing signal as roughly one minus the ratio, capped at zero. It is a rough model rather than a published result, but it makes the ordering clear. At 40 stops with 150-metre error the ratio is 0.85, so roughly 15% of sequencing information survives and a 10% better solver returns around 1.5%. Fix the error to 30 metres and the ratio falls to 0.17, roughly 83% of the signal survives, and the same solver improvement returns around 8.3%. At 150 stops with 150-metre error there is effectively no signal left to improve.

The conclusion is uncomfortable for anyone mid-procurement: in a high-density, low-coordinate-quality network, solver capability is not the binding constraint and buying more of it will not show up in the numbers.

| Also Read: Understanding the Last Mile Problem and Key Solutions |
| --- |

## Where the constraint actually sits by market type

| Dimension | Well-addressed grid market | Southeast Asian dense urban core | Southeast Asian peri-urban and rural |
| --- | --- | --- | --- |
| Address basis | Numbered street position | Landmark, alley and relational description | Landmark and village-level description |
| Typical error against spacing | Low ratio, sequence survives | Ratio often near or above 1 | High error but wide spacing, ratio moderate |
| Binding constraint | Solver quality and constraint fidelity | Coordinate quality | Coverage and access, then coordinates |
| Where lost time appears | Travel time | Service time, booked as dwell | Travel time and failed attempts |
| Highest-return investment | Better optimisation | Address resolution and coordinate learning | Territory design and rider local knowledge |
| Rider dependency | Low, plan is followed | High, plan is corrected in the field | High |

The middle column is where most Southeast Asian volume sits and where the standard playbook misfires. The right-hand column is the one operators expect to be hardest, and on coordinates it often is not, because wide spacing tolerates error that a dense core cannot. The practical consequence is that a single regional routing configuration will be wrong in two directions at once, over-investing in solver capability in the cities and under-investing in coverage outside them.

| Also Read: Route Optimization for Enterprise Logistics: 2026 Guide |
| --- |

## Five criteria for evaluating routing in a low-address-quality market

**1. Does it learn coordinates from execution?** The single highest-value capability is using captured delivery positions to correct the stored coordinate for the next order to that address. Ask specifically whether the platform writes back a learned position, and after how many visits.

**2. Does it report positional error, not just match rate?** A match rate of 98% tells you a coordinate was returned for 98% of addresses. Ask for the distribution of distance between planned coordinate and captured proof-of-delivery position.

**3. Can it handle multiple national address formats natively?** Operating across several markets means several addressing systems. Ask which are supported as structured formats rather than parsed as free text.

**4. Does it model rider local knowledge as an asset?** In alley networks the rider who has served a neighbourhood knows things the map does not. A platform that reassigns territories freely each day destroys that knowledge; one that balances familiarity against efficiency preserves it.

**5. Does it separate service time into components?** If dwell time is one number, addressing cost is invisible. Search time, waiting time and handover time need to be distinguishable, or the business case for fixing coordinates can never be built.

| Also Read: Routing Efficiency: Definition, Benefits and Tips |
| --- |

## What this looks like in regional deployments

A global FMCG manufacturer distributing across ten Asian countries through more than 1,000 distributors and 5,000 riders reached [3X ROI](https://locus.sh/case-studies/global-fmcg-logistics-automation/)
 while saving more than 12,000 trips a month, optimising over 4 billion dollars of orders and reaching 1.8 million retail outlets. At that outlet count, address resolution is not a preprocessing step. It is the difference between a plan a rider can execute and a plan a rider has to rebuild, and trips saved is the metric that captures it, because a trip that has to be repeated is the most expensive form of geocode error.

A leading ASEAN apparel retailer took the parcel-side version of the same problem, consolidating [multi-carrier parcel management](https://locus.sh/case-studies/apparel-multi-carrier-parcel-management/)
 across a fragmented regional carrier base. Where multiple carriers each hold their own view of an address, the same delivery point can carry several different coordinates, and normalising them is a prerequisite for comparing carrier performance at all. Otherwise a carrier looks slow when it was simply given a worse coordinate.

## Four mistakes in Southeast Asian route optimisation

**Buying solver capability to fix a coordinate problem.** It is the most common procurement error in the region, and the reported gains disappoint in a way nobody can explain afterwards.

**Reading match rate as accuracy.** They are different measurements. One says a coordinate exists, the other says it is in the right place.

**Reassigning territories daily for theoretical efficiency.** In alley networks, rider familiarity is a real productivity asset. Churn destroys it faster than optimisation rebuilds it.

**Booking search time as dwell.** It hides the entire cost of the problem inside a metric nobody questions, which is why addressing rarely gets funded.

| Also Read: Dynamic Route Optimization: 2026 Guide for Last-Mile Teams |
| --- |

## How Locus approaches routing where addresses are hard

Locus, the world’s first Decision-Intelligent, Agentic TMS, treats address resolution as part of the routing problem rather than as an upstream data task, which is the design choice that matters in this region. The [route planning system](https://locus.sh/route-planning-system/)
 combines geocoding and address conversion with constraint-based sequencing, so the coordinate and the plan improve together rather than the plan inheriting whatever the coordinate happened to be.

Geocoding and carrier integrations are built for the address infrastructure of each market rather than applied as a single global model, which is the practical requirement when one deployment spans several national addressing systems. The Fireworks routing engine then plans against more than 250 real-world operating constraints, covering two-wheeler and mixed-fleet capacity, rider shift hours, time windows and access restrictions, and turns raw orders into dispatch-ready routes in roughly two minutes so that corrections made in the field can be absorbed rather than deferred to the next planning run.

Across the region’s deployments, route planning delivers up to 34% fewer miles, 25% higher drop density and 28% fewer trips through order consolidation, with fleet utilisation up to 90%. In a market where the coordinate is the weak link, the trip reduction is the number to watch, because it counts journeys that did not have to be repeated.

Locus supports more than 360 enterprise customers across 30-plus countries, with over 1.5 billion deliveries optimised, more than 320 million dollars in documented client logistics savings and 99.99% uptime. It has been [recognised by Gartner for seven consecutive years](https://locus.sh/analyst-recognition/)
, featured in the 2026 Hype Cycle for Supply Chain Execution and Logistics Technologies, named a Leader in TMS by QKS Group (SPARK Matrix), and ranked #1 in Route Planning on G2’s 2026 Best Software Awards.

In October 2025, Ingka Investments, the investment arm of Ingka Group, the world’s largest IKEA retailer, acquired Locus. Locus continues to operate independently.

So what limits route optimisation in Southeast Asia? Not the solver. The binding constraint is geocode error measured against mean drop spacing, and because density shrinks that spacing, the ratio worsens in exactly the dense urban cores where drop density should be paying off. Diagnose it by dividing your positional error by mean spacing: below 0.3 the sequence survives, above 1.0 it is noise, and no amount of solver improvement recovers information that was never in the input. Locus addresses the constraint where it actually sits, resolving and learning coordinates as part of route planning rather than treating addressing as someone else’s preprocessing step, then sequencing against more than 250 operating constraints across mixed two-wheeler and van fleets. For enterprises running high-density delivery across Southeast Asian markets, that combination of address resolution and constraint-based planning is why Locus is the reference platform in the region. [Request a Locus route planning assessment](https://locus.sh/schedule-demo/)
 to measure your own error-to-spacing ratio.

## Frequently Asked Questions

**Why does route optimisation underperform in Southeast Asia?** Usually because coordinate quality, not solver quality, is the binding constraint. Many addresses are described by landmark, alley and relation rather than a numbered street position, so the coordinate given to the planner carries error. The solver returns an optimal sequence for points that are not where it thinks they are, and the rider absorbs the difference as search time.

**How do I know whether my geocodes are good enough to optimise?** Divide your typical positional error by the mean distance between consecutive stops on a route. Mean spacing is roughly half the square root of service area divided by stop count. Below about 0.3 the planned sequence largely holds. Above about 1.0 the error exceeds the gap between stops and the sequence is dominated by noise.

**Does higher drop density help or hurt in low-address-quality markets?** Both, and they are usually measured separately. Density cuts cost per drop, which is why it is pursued. It also shrinks the spacing that geocode error is compared against, so the same error does more damage. Moving from 10 to 150 stops in one service area raises the error-to-spacing ratio by roughly four times.

**What is the difference between geocode match rate and geocode accuracy?** Match rate is the share of addresses for which a coordinate was returned. Accuracy is how far that coordinate sits from the real delivery point. A platform can report a 98% match rate while placing a meaningful share of those coordinates over a hundred metres away, which is why positional error against captured proof-of-delivery positions is the measurement that matters.

**How much does poor address data actually cost a delivery operation?** It appears as service time rather than as a data cost. On a 40-stop route with a 280-minute baseline, 15% of stops needing a six-minute search adds about 13% to the route, equivalent to losing five drops of capacity. Because that time is usually recorded as dwell, the cost is real and the cause is invisible.

**Should I fix addressing before buying route optimisation software?** They are not sequential if the platform learns coordinates from execution, which is the capability to insist on. What does not work is buying solver capability while leaving coordinates untouched: where the error-to-spacing ratio is near or above one, a 10% better solver returns a fraction of a percent, because there is little sequencing information left for it to exploit.

MEET THE AUTHOR

Aseem Sinha

Vice President - Marketing

Aseem, leads Marketing at Locus. He has more than two decades of experience in executing global brand, product, and growth marketing strategies across the US, Europe, SEA, MEA, and India.

### Related Tags:

[https://locus.sh/blogs/best-tms-grocery-food-delivery-2026/](https://locus.sh/blogs/best-tms-grocery-food-delivery-2026/)
#### [General](https://locus.sh/blogs/category/general/)

## [Best TMS for Grocery and Food Delivery: The 2026 Enterprise Guide](https://locus.sh/blogs/best-tms-grocery-food-delivery-2026/)

[Aseem Sinha](https://locus.sh/blogs/author/aseem_locus/)

Sep 7, 2026

Grocery and food delivery break generic TMS platforms on time windows, temperature and mixed fleets. How to evaluate, and why Locus is built for these constraints.

[Read more](https://locus.sh/blogs/best-tms-grocery-food-delivery-2026/)

[https://locus.sh/blogs/route-optimisation-europe-urban-access-regulations-2026/](https://locus.sh/blogs/route-optimisation-europe-urban-access-regulations-2026/)
#### [General](https://locus.sh/blogs/category/general/)

## [Route Optimisation Across European Urban Access Zones: Why the Strictest Stop Sets Your Vehicle in 2026](https://locus.sh/blogs/route-optimisation-europe-urban-access-regulations-2026/)

[Ishan Bhattacharya](https://locus.sh/blogs/author/ishan_locus/)

Sep 9, 2026

Europe has over 500 low-emission zones, each with its own rules. Vehicle-to-route assignment is now a compliance decision, and the strictest stop governs the route.

[Read more](https://locus.sh/blogs/route-optimisation-europe-urban-access-regulations-2026/)

## Route Optimisation for Southeast Asia: Why Address Quality Caps Your Routing Gains in 2026

- Share
- [Print](javascript:window.print())
- [Download](#)
- [Schedule a Demo](https://locus.sh/schedule-demo/)

### Is your team spending more time on fixing logistics plan than running the operation?

- Agentic transportation management from order intake to freight settlement
- Route optimization built on 250+ real-world constraints
- AI-driven dispatch with automatic execution handling

20%Cost Reduction

66%Faster Planning Cycles

[Schedule a demo](/schedule-demo/)

Insights Worth Your Time

#### [General](https://locus.sh/blogs/category/general/)

## [Locus 2026 US Consumer Survey: Generative AI isn’t Just Changing How Consumers Shop, it’s Breaking the Demand Patterns US Retail Was Built On](https://locus.sh/blogs/generative-ai-shopping-effect-retail-fulfillment-operations-locus-q2-2026-consumer-survey/)

[Ishan Bhattacharya](https://locus.sh/blogs/author/ishan_locus/)

May 29, 2026

#### [General](https://locus.sh/blogs/category/general/)

## [Embedded vs Bolted-On AI: The Architecture Question European Logistics Buyers Are Asking](https://locus.sh/blogs/embedded-vs-bolted-on-ai-european-logistics-platform-architecture-business-benefits/)

[Aseem Sinha](https://locus.sh/blogs/author/aseem_locus/)

May 21, 2026

#### [General](https://locus.sh/blogs/category/general/)

## [Hybrid Fleet Management: How Owned, 3PL, Gig, ICE, and EV Capacity Actually Operate at Most Enterprises](https://locus.sh/blogs/three-workforce-fleet-reality-owned-3pl-gig-drivers/)

[Aseem Sinha](https://locus.sh/blogs/author/aseem_locus/)

May 7, 2026

#### [General](https://locus.sh/blogs/category/general/)

## [US Returns Hit $850 Billion in 2025: Why US Retailers Are Restructuring Reverse Logistics in 2026](https://locus.sh/blogs/850-billion-us-returns-ai-routing-reverse-logistics-2026/)

[Ishan Bhattacharya](https://locus.sh/blogs/author/ishan_locus/)

May 7, 2026