A new direct ship to Mamuju, and the containers it still needs
Mamuju is the capital of West Sulawesi, a province on the west coast of the island of Sulawesi. Its shops and distributors sell packaged food, household goods and building supplies, and much of that stock is bought from suppliers in Java, around Surabaya and Jakarta. All of it has to cross the sea. West Sulawesi’s ports unloaded about 635,000 tonnes of domestic cargo in 2025, counting every kind of cargo.[11]
For years that stock came the long way. Ships unloaded it at Makassar, the big port city of the neighbouring province, and trucks carried it about 500 kilometres north.[1] Now Meratus Line, an Indonesian container shipping company, calls directly at Belang-Belang, a port that serves Mamuju; the service was formally launched in November 2026. Its ships stop about three times a month. Each call starts with 30 to 40 containers, and the target is 50 to 60. Meratus describes the service as still at a pioneering stage.[1]
This article follows Meratus. A call earns money for a shipping line only when the ship carries enough paying containers. The shops and distributors it hopes to carry for will switch only if the new route is cheaper and dependable once everything is counted. That means the freight, the unloading, the truck to the door, and the days spent waiting for the next sailing. The governor of West Sulawesi said the service would cut the region’s logistics costs. Nobody has yet measured whether it does.[1]
How one restocking order travels today
Most of the businesses receiving goods in Mamuju, from small shops to distributors (we call them all distributors here), are too small to fill a container. They go through a forwarder: a company that arranges transport without owning the ships. The forwarder gathers many customers’ small lots, packs them into shared containers and buys space on a ship. Here is one order of dry goods from Surabaya to a Mamuju distributor, as the research describes it.
- The distributor asks for prices. It needs stock by a certain date and asks forwarders, by website or WhatsApp, what it would cost. Published prices start at Rp5,000 (Indonesian rupiah) a kilo, about 30 US cents, with a 50 kg minimum, before tax, insurance and the unloading of heavy items.[2] Others advertise about Rp6,000 and Rp6,500 a kilo.[3] [4]
- The forwarder quotes and the distributor chooses. Each quote leaves out different things. One excludes unloading, another insurance, another says delivery to some districts must be confirmed.[2] [4] [5] One seller warns that not every load goes through Belang-Belang.[4]
- The goods wait in Surabaya. The supplier delivers them to the forwarder’s warehouse, where they are weighed, measured and packed with other customers’ goods. They wait there until the last loading date before a sailing.[6]
- Meratus carries the container. Its sales team, or its own logistics arm, sells space to forwarders and larger shippers. Its planners hand out space on each ship, and empty containers to fill, as bookings arrive.[7] [8] On the older route the container lands in Makassar instead, and a truck charter to Mamuju is advertised at Rp3.5 million to Rp9.5 million.[9]
- The port releases the goods. At Belang-Belang the terminal and authorised handlers unload the container and release the cargo. The head of the port unit promised a neutral, system-based service with no illegal fees.[1]
- A local truck delivers. A pickup can be hired in Mamuju for Rp320,000 a day without a driver.[10] The distributor checks the goods and pays, and any damage claim follows.
Two things stand out from Meratus’s side. It sees a customer’s demand only when a booking arrives, a short time before the sailing, and each customer chooses a carrier afresh for every order. And because every quote includes different things, neither Meratus nor its customers can easily tell whether the direct route is actually the cheaper delivery.
Our earlier investigation on What Could Work showed why that matters. On a road route in Java, a half-tonne factory lot paid roughly three times as much per tonne as goods filling a whole truck.[12] Full containers are where the savings are. Thin loads keep the price high.
The problem in one sentence: Meratus needs its new calls to carry enough paying cargo, but a shorter route alone neither fills its ships nor proves that its customers save money once waiting and extra handling are counted.
Improve: a quote that shows the delivered cost side by side
The step this targets is the price request and the choice in steps 1 and 2. Today the distributor collects offers that each include different things and works out for itself which one is cheaper. If Meratus is cheaper delivered, the customer may never see it. If Meratus is dearer, Meratus never learns which charge it lost on.
The change is small. When a willing customer asks Meratus’s sales team for a price, the customer can also share the other offers it has received. Meratus sends back its quote together with a side-by-side sheet: every offer on the same basis, with what each one leaves out and when the goods would arrive.
What the AI would do. Offers arrive as photographed price lists, PDFs and chat messages. AI reads the text and tables in them and pulls out the price per kilo, the minimum weight, and what is included or excluded.[A] It then lines the offers up on one list of items and flags what does not match, such as “this offer excludes unloading; that one includes it”. A charge an offer does not state stays blank and is never guessed. AI drafts the sheet, and a salesperson checks every flagged item before anything is sent.[B]
Why AI, and not a form? Because the documents are written by other companies, each in its own way, and customers will not retype them. Reading clean printed price lists is well established. Lining up terms that are worded differently is less reliable, which is why a person checks it. The cost per kilo itself is ordinary arithmetic and needs no AI. Arrival dates come from the published sailing calendar, not from a prediction.[B]
| Step | Today | With Improve |
|---|---|---|
| Comparing offers | The customer alone Reads each offer’s terms itself | Meratus prepares it AI reads the offers; sales checks the sheet |
| Spotting exclusions | Easy to miss Unloading, tax and insurance vary | Flagged line by line Unstated charges stay blank |
| Cost per kilo | Headline price Not on the same basis | Same-basis total Plain arithmetic, no AI |
| What Meratus learns | Little Lost quotes leave no reason | Where it loses Shared only with the customer’s consent |
| Choice, price, handling | Unchanged The customer chooses, sales negotiates, partners handle and deliver the goods | |
Who would use it: Meratus’s sales desk or its logistics arm. It could help a customer see a real saving where one exists, and show Meratus which charges cost it business. It cannot fill a ship by itself. If the direct route is not cheaper delivered, the sheet will say so. And customers who ship small lots through forwarders may never ask Meratus for a quote at all.
Reinvent: a standing restocking service that books itself
A larger change goes after the reason the ships sail part-empty. Today Meratus learns about demand only when someone books, and every order is a fresh sale. The Reinvent idea starts earlier, from the customer’s shelves.
Meratus would sell a recurring distributor a standing delivery service instead of one shipment at a time. The two sides agree once on the goods, the delivery address, how late stock may arrive, a spending limit and what the price includes. The distributor signs a delivery mandate it can limit or cancel. From then on, when its stock runs low and its Java supplier has the goods ready, Meratus proposes a shipment on a coming call. Meratus becomes the one company answerable for the delivery. It pays forwarders and local truckers to do the physical work.
| Step | Today | With Reinvent |
|---|---|---|
| What starts a shipment | A new freight request When the distributor decides to order | Stock running low Within limits agreed in advance |
| Choosing a carrier | Every order Price shopped each time | Once, at sign-up Reviewed at renewal |
| When Meratus sees demand | At booking Shortly before the cut-off | Earlier, from stock Distributor and supplier records |
| Filling each call | Whatever is booked In order of arrival | Planned ahead Committed and forecast loads |
| Who answers for delivery | Several firms Forwarder, carrier, trucker | Meratus alone Partners do the physical work |
| Weighing, port release, claims | Unchanged People weigh and pack, the port releases, the distributor checks the goods, staff decide claims | |
Where AI does the work
A standing contract like this could also be run with paper forms and simple rules. The AI parts are worth adding only if they do better than that version, which is most likely where records are messy and customers many.
- Reading the distributor’s records. Where small distributors keep stock records at all, they may be paper cards or photos sent by phone; the research does not say. AI would read those images and supplier notes, pulls out product, quantity and date, and flags anything unclear for a person. It would also suggest which supplier product names and stock items are the same thing, for a person to confirm.[C]
- Estimating when each distributor will need its next delivery. From past stock use, a forecasting model estimates each distributor’s next need and the total for each call. This is the least certain piece. The route is new and stock histories are likely short. So the forecast must beat simple reorder rules, such as “reorder when stock falls below a set level”, before anyone relies on it. A forecast can propose a shipment, but it is never counted as paying cargo.[D]
- Fitting shipments to calls. Deciding which goods go on which call, within each distributor’s limits and the ship’s space, is planning mathematics. So is booking warehouse and truck slots, and sending approved orders and invoices between systems. None of these needs AI.[E]
A Meratus planner approves every shipment outside a distributor’s agreed limits. The distributor still decides what it buys. If stock data are missing or out of date, shipments pause until a person checks with the distributor.
Why this is more than a set of fixes. Faster quotes, booking reminders or better forecasts from bookings all wait for the customer to order first. Only when the distributor’s stock, not a freight order, sets off a shipment can Meratus see and lock in loads before the customer shops around. That is what could fill the calls.
The limits worth stating. Forwarders sell Meratus’s space to small shippers today. If Meratus serves their customers directly, some forwarders may send the cargo they still handle via Makassar instead, cancelling out the gain. Distributors and their suppliers would take on new record-keeping. Meratus would carry the cost when a call is missed or a partner fails.
The research behind this article first suggested a different business: a new forwarder that pools small loads onto Meratus’s direct calls. If Meratus ran the standing service itself, that newcomer would face its own carrier competing for the same recurring customers. Newcomers face a hard sum on routes like this. When our earlier investigation on What Could Work looked at a specialist carrying rice to Papua for BULOG, Indonesia’s state food-logistics agency, existing forwarders’ prices left surprisingly little room for a newcomer’s profit.[13]
What each change could offer
Neither idea has been tried on this route, so the table judges potential: how well each could do if it were built and run sensibly, and why.
Favourable for the business Mixed Unfavourable
| Potential | Improve: side-by-side delivered-cost quote | Reinvent: standing restocking service |
|---|---|---|
| Technical feasibility | Moderate to highReading clean printed price lists and regular tables is established; lining up differently worded terms is conditional, so sales checks every flagged item. [A] [B] | ModerateContracts, planning and scheduling are ordinary software. Reading handwritten stock cards and forecasting from short histories on a new route are conditional and must beat simple rules. [C] [D] [E] |
| Economics | Low to moderate, unsizedPays only if clearer comparisons win profitable repeat cargo and the AI saves more work than a plain spreadsheet. No cost or benefit figure exists. [F] | Potentially high, unsizedPays if extra containers per call, over about 36 calls a year, earn more than the system, staff and reserved-slot costs, net of cargo merely moved from forwarders. Neither side of that sum is known. [1] [F] |
| Customer value | ModerateCustomers see what each offer leaves out and the real cost per kilo, but the freight price itself does not change. | Potentially highDistributors stop buying freight order by order, have one company answerable for delivery and may run out of stock less. The cost is sharing stock data and living with about three sailings a month. |
| Competitive positioning | Easy for incumbentsMeratus or any forwarder could adopt it quickly, which suits them. For the same reason it gives no lasting edge, and there is little room for a company built on selling it. | Moderate to highMeratus runs the calls and sets the calendar, and has the scale to spread the cost; a forwarder could copy the contract but not the capacity. Customers signed to standing service are harder to lure away. |
| Risks | ModerateA misread charge or arrival claim could cause disputes, so a person checks each sheet. Showing the true comparison may reveal that Meratus is dearer and push it into discounts. | HighForwarders may move cargo to the Makassar route, wrong or out-of-date stock data could send the wrong goods, Meratus carries the cost of missed calls across many customers, and the pioneering service may not continue. |
| How it would be tested | On real past offers for this route, compare today's quote, a staff-made spreadsheet and the AI-prepared sheet checked by sales, measuring minutes, wrong or missed charges, and later repeat paid cargo. | With a small group of recurring Mamuju distributors: a records check, then a background run over several calls measuring how much earlier demand is visible, then a supervised trial with cancellable mandates beside a no-AI group using reorder rules. |
Which is worth pursuing?
For each idea we ask: if the arrangement can be made to work and the AI part is practical, would it create enough value to justify the change?
Improve: potentially yes, as a modest step. The reading and checking are practical today, and a pilot is cheap and easy to stop. Its value for Meratus is indirect. First, seeing the delivered cost side by side has to win Meratus profitable repeat cargo. Second, the AI has to save more work than a plain spreadsheet would. It is best placed with an established company: Meratus’s own sales desk or logistics arm, or any forwarder. Because any rival could copy it quickly, a new company could not build a business on it.
It is worth doing on three conditions: customers are willing to share their other offers; Meratus’s delivered price wins on enough of them; and the AI beats a spreadsheet on effort without adding errors.
Reinvent: potentially yes, and it is the change aimed at the real problem. If distributors will plan their restocking with Meratus, the calls would carry loads that were known and committed in advance, not whatever happened to be booked. Most of it is contracts and ordinary planning software. The AI parts are kept only if they prove more accurate than simple reorder rules. It is best placed with Meratus itself, which already runs the calls, sets their calendar and has the scale to spread the cost. A new forwarder would lack that capacity. It is worth doing on five conditions:
- enough recurring distributors keep usable stock records and will share them;
- about three calls a month is often enough for their stock needs;
- forwarders agree to be paid by Meratus to pack and deliver the goods, or the cargo gained outweighs what they take away;
- the extra containers per call earn more than the service costs to run;
- the Meratus service lasts beyond its pioneering stage.
Worth pursuing: Reinvent. The side-by-side quote is a sensible small step, but it cannot fill a ship. The ships sail part-empty because Meratus learns about demand late and wins it one order at a time. Only the standing restocking service changes that. It should start small, with a check of a few distributors’ records and a trial run in which nothing is booked for real.
What would have to be tested
For Improve, a task pilot on Meratus’s sales desk. Take a set of real past offers on this route and compare three ways of handling them: today’s quote, a staff-made spreadsheet, and the AI-prepared sheet checked by sales. Measure minutes per comparison, missed or wrong charges, and blanks correctly left blank. Only then follow whether customers who got the sheet book more paid cargo.
For Reinvent, a staged trial with a small group of recurring Mamuju distributors. First, check whether they keep usable stock records. Then run the service in the background for several calls while they keep booking as usual, and measure how many days earlier their demand would have been visible. Only after that, a supervised trial with signed, cancellable mandates. A parallel group runs the same contract with simple reorder rules and no AI.
The idea fails if records are missing or demand is not visible much earlier than bookings. It also fails if forwarders take away more cargo than the service adds, or if distributors end up with more empty shelves or higher costs.
Before either, the missing evidence:
- The real delivered price on the direct route, with what it includes. No current all-in price was found.[5] [8] This decides whether either idea has a saving to show.
- Meratus’s loads, booking times, revenue and costs per call. None are published. This decides how many extra containers Reinvent must add to pay.
- Whether distributors keep usable stock records and would share them. This decides whether Reinvent can start at all.
- What waiting and holding stock cost the distributors. If three sailings a month cost them more in stock than the freight saves, Reinvent fails.
- How much cargo forwarders put on the calls, and how they would react. This decides whether Reinvent’s gain is real or just cargo moved from one channel to another.
- For Improve: how much work comparison takes today, and whether customers would share their offers. This decides whether the AI is worth more than a spreadsheet.
Research Notes
What this investigation is based on
What Could Work’s stored research on freight between Java and Mamuju (the sources below), and our own analysis of how the work is done and where AI could change it. No interviews or pilots were carried out, and no Meratus operating data were available.
Whose problem this article takes, and why. The research names several parties: the Mamuju distributors and commodity shippers who pay for the freight, the forwarders that pool their cargo, and Meratus. The distributors carry the cost, but the research did not show that they have the money, records, technical staff or shipment volume to change how the work is done. Forwarders such as LSJ Express came close but fell short on several of these. Meratus runs the calls and sets their calendar and terms, has the resources and scale, and its decisions reach every shipper on the route, so the article follows Meratus and treats the distributors as the customers it serves.
Sources
- ANTARA, Gubernur Sulbar: Peti kemas Meratus pangkas biaya logistik daerah (11 November 2026). The governor’s statements; about three calls a month with 30–40 containers initially, target 50–60; a first five-container cocoa shipment to Surabaya as a trial; Meratus Line’s Chief Trade Officer calls the service still at a pioneering stage; the earlier route via Makassar and about 500 km by road; the Belang-Belang port unit head’s pledge of a neutral, system-based service without illegal fees. Reported statements, not operating records; no measured saving.
- CNT, price list. Surabaya–Mamuju from Rp5,000/kg, 50 kg minimum, excluding tax, insurance and heavy unloading. Advertised floor; negotiated prices unknown.
- Makharya Cargo, Sidoarjo–Mamuju. About Rp6,000/kg, 50 kg minimum, door delivery, website and WhatsApp ordering. Vendor page; routing via Belang-Belang not verified.
- MUAT, Surabaya–Mamuju. Rp6,500/kg, 100 kg minimum, about 12 days; district delivery subject to access confirmation; does not promise every load uses Belang-Belang. Vendor page.
- LSJ Express, container service Surabaya–Mamuju. Full and shared containers, price on request, unloading and insurance excluded. Part of the page is copied from another route and is not relied on.
- Pujiwati Cargo, Surabaya–Mamuju. Weighing, agreed price, payment and warehouse preparation; onward shipment via Makassar. Vendor description of its own process.
- PT Matrans Lintas Buana, services. Full and shared containers and trucking; lists Mamuju among destinations and Meratus as a carrier partner. No rates.
- Meratus, support page. Search-index snippet only: the logistics arm advertises full and shared containers, warehousing, transport, a quotation portal and insurance limits. Its Mamuju small-load scope and terms are unverified.
- Boska Trans, charter price list. Makassar–Mamuju from Rp3.475 million (pickup) to Rp9.5 million (large truck). Terms not established.
- Balioh, Mamuju pickup rental. Rp320,000 a day without driver. Availability not verified.
- BPS, Statistical Yearbook of Indonesia 2026, table 10.1.6. West Sulawesi domestic cargo unloaded in 2025 (provisional): 634,770 t, all cargo types including bulk.
- Our earlier investigation on What Could Work: Small factory loads from West Java to Surabaya pay a premium per tonne; a pooled weekly truck could cut it (13 November 2026). A half-tonne lot paid about Rp1.5 million a tonne against about Rp500,000 a tonne in a full truck.
- Our earlier investigation on What Could Work: Could a specialist lower BULOG’s rice delivery costs to Papua? (11 November 2026). Existing forwarders advertised lower prices than assumed, leaving a newcomer little room.
Our explanations and assumptions
Where a note discusses a kind of AI, it draws on our AI Capability Atlas, What Could Work’s own reference on what today’s AI can do. For each kind of task it records what goes in and what comes out, how proven the approach is, its typical failures and how to check for them. Established means proven methods exist for a narrow, well-defined task. Conditional means it can work, but reliability depends heavily on the setting and has to be tested there. Exploratory means it is still unproven beyond experiments. The Atlas describes AI in general; nothing in this article has been tested in the business it describes.
- Reading competitors’ offers. Text recognition turns photographed or scanned price lists into text. Layout recognition keeps the rows, columns and footnotes of a rate table together. A further step interprets prices, minimum weights and units, and pulls out stated inclusions and exclusions with the sentence each came from. Our AI Capability Atlas rates text recognition as established for clean printed text and conditional for handwriting and poor images. It rates table reading as established for regular, consistent tables and conditional for complex ones, and warns that structure can be read correctly while the meaning is wrong (units, currencies, repeated totals). Pulling out stated terms is established for defined kinds of term and conditional for complex conditions, and extraction never shows that a term is enforceable or complete.
How this shaped the article: this is why a salesperson checks every flagged item, why unstated charges stay blank, and why the reading must be tested on real Mamuju offers, including cropped images and footnotes. Improve’s technical feasibility is rated moderate to high, not high. - Lining up the offers, drafting the sheet, and what needs no AI. Lining up differently worded terms on one list is rated conditional when a model proposes the matches; it is established only for known conventions. Flagging contradictions between sources is established for structured records and conditional for differently worded ones, and finding a difference never shows which source is right. Drafting a readable document from checked facts is established when the drafting is bounded and reviewed; the Atlas warns that fluent text can invent facts or commitments. The cost per kilo is ordinary calculation, established without AI. Predicting arrival times from history would need a long record of sailings, which this route does not have, so arrival dates come from the published calendar.
How this shaped the article: this is why the draft cannot promise a price or a saving, why arithmetic and arrival dates are described as plain software, and why the pilot compares the AI sheet with a staff-made spreadsheet. - Reading shops’ stock records and linking product names. Text recognition reads photographed stock cards and supplier notes; field extraction pulls out product, quantity, unit and date, with missing values flagged. Record linking proposes which supplier product name and shop stock item refer to the same thing. The Atlas rates text recognition established for clean print and conditional for handwriting; field extraction established for stable formats and conditional across mixed documents; record linking established with strong identifiers and conditional with sparse or ambiguous records, warning that a false match can corrupt a history.
How this shaped the article: this is why unclear quantities go to a person and are never filled in, why product links are confirmed by a person, and why stale or missing data pause shipments. - Forecasting each shop’s next need. A forecasting model estimates future stock use from past records. The Atlas rates it established for recurring, bounded patterns and conditional for sparse data and changing conditions. A new route and short shop histories are both. It warns that a single forecast figure hides its uncertainty, and that a forecast should first be tested against simple rules on past data.
How this shaped the article: this is why forecasting is called the least certain piece, why a forecast never counts as paying cargo, why Reinvent’s technical feasibility is moderate, and why the trial runs a parallel group with simple reorder rules and no AI. - Planning, scheduling and sending orders. Allocating goods to calls and containers within stated limits, scheduling warehouse and truck slots, computing each call’s earnings, and sending approved orders and invoices between systems are all established as ordinary software when the rules are clearly specified. The Atlas rates them conditional in fast-changing settings. It warns that forecast errors can outweigh planning gains, and that a system reporting success does not prove the goods moved.
How this shaped the article: this is why these steps are described as needing no AI, why a planner approves shipments outside agreed limits, and why each delivery is confirmed by the shop on arrival. - What each idea would cost and earn. No figure for either idea’s cost or benefit exists in the records, so none is given. For Improve, Meratus would pay for setting up one comparison sheet and calculator, a link to its sales system, testing and training, and then a year of reading service, secure storage, staff checking and upkeep. It pays only if extra profitable repeat cargo, plus staff time saved, exceeds what the AI adds over a plain spreadsheet, with set-up spread over three years. For Reinvent, Meratus would pay to build the system, sign up an assumed first group of ten distributors and their main Java suppliers, staff an account owner and planner time, and cover slots reserved but unused on about 36 calls a year. It pays if the extra containers per call, times the profit on each container, times 36 calls, exceeds that yearly cost, after removing cargo merely moved from forwarders. Put another way, the extra containers each call must carry equal the yearly cost divided by 36 times the profit per container. Neither the cost nor the profit per container is known. The rupiah amounts in the story are converted at roughly Rp16,500 to the US dollar, our approximation.
- Other ideas we considered. For Improve, a booking confirmation that also asks, a few days before the cut-off, whether the cargo is still ready. We did not choose it because no evidence shows that no-shows explain the part-empty calls, and a simple reminder form would capture most of its value without AI. For Reinvent, a model that sails a call only once a two-way load is agreed; it was dropped because Meratus’s freedom to make calls conditional is not established, return cargo rests on one five-container cocoa trial, and conditional sailing would undermine dependable service.
- Our summaries. The six-step order, both “today / with the change” tables and the description of Meratus’s sales and planning steps are our summaries of the process the research records. Mamuju’s geography and Surabaya’s role are general knowledge.
Evidence status of the main claims
| Claim | Status |
|---|---|
| Meratus calls directly at Belang-Belang about three times a month, 30–40 containers per call against a 50–60 target | Reported News report of company statements [1] |
| Forwarders sell per-kilo and container freight to Mamuju, with differing exclusions | Observed Published offers [2] [4] [5] |
| The direct route lowers the full delivered cost for distributors | Not established No matched price found |
| Meratus sees demand only at booking, order by order | Inferred From the recorded process |
| A side-by-side quote wins Meratus profitable repeat cargo | Hypothesis |
| A standing restocking service fills calls earlier and more fully | Hypothesis |
| Distributors keep usable stock records and would share them | Unknown |
Open questions
- What is Meratus’s all-in delivered price to a Mamuju address, and what does the Belang-Belang terminal charge? Meratus, its logistics arm and the port unit could answer.
- How many containers does each call actually carry, how early do bookings arrive, and what does each call earn? Only Meratus holds these figures.
- Do recurring Mamuju distributors keep stock records, and would they share them and accept a standing mandate? A few conversations with distributors would show.
- How much cargo do forwarders put on the calls, and would they accept paid work moving goods for a Meratus service?
- Will the direct calls continue beyond their pioneering stage, and do return cargoes such as cocoa have reliable seasons?
Leave a Reply