America has decided to reindustrialize. New factories, semiconductor fabs, battery plants, energy projects and defense capacity are going into the ground across the country, backed by hundreds of billions of dollars in public and private investment.
\nI spent the first part of my career building a manufacturer, so I am glad to see it. I also know what those new plants will run into on their first day of operation.
\nA new plant arrives with modern machines and modern software. Then it plugs into the same coordination model the rest of industry has used for decades. Suppliers confirm by email. Carriers receive load tenders as EDI messages from systems built in the 1980s. Bills of lading and rate confirmations move as PDFs. Inventory lives in one system and the production schedule in another, and partners share what they must, late and by hand.
\nThe results are familiar. Buffer inventory to cover uncertainty. Lines waiting on parts that exist somewhere else. Trucks leaving the dock at a third to half of their capacity, as they do across American freight, with one mile in six driven empty.1 People spending their days moving information between systems that cannot see each other.
\nBuilding new plants on that foundation means pouring new concrete onto an old coordination model, and inheriting its costs from the first shipment.
\nThe more important shift is happening on the machine side.
\nDriverless trucks began hauling commercial freight between Dallas and Houston in May 2025.2 Humanoid robot makers are building factories sized for mass production. Robot makers, autonomous trucking companies and factory automation vendors are moving at a pace the industrial economy has not seen before.
\nThese machines are capable, and they will need work. A driverless truck needs to know what to carry, where to go and what the trip is worth. A fleet of humanoid robots needs orders, inventory and schedules. A highly automated factory needs to know what to build before anyone asks.
\nThe market they have to get that work from was not built for them. Much of American freight is still tendered through EDI over value-added networks, often generated by mainframe and AS400 systems. Purchase orders arrive as PDFs. Status updates arrive by email. Capacity is negotiated by phone. A driverless truck cannot read an X12 204 sitting in a VAN mailbox. A robot cannot reply to a rate confirmation attached to an email.
\nThe companies best positioned for this shift are the ones that own the whole stack. When one company builds the robots, the vehicles and the factories they work in, on its own systems, its machines can work together because it controls every system they touch.
\nAlmost no one else is in that position. Without a shared translation layer, autonomous machines will be limited to the few customers who can rebuild their operations around them, and the rest of the industrial economy will keep running on people relaying messages. That gap is one of the biggest risks to reindustrialization, and one of the least discussed.
\nIOI is built to close it. It makes the kind of integrated coordination only the largest vertically integrated operators have today open and easy to set up for every manufacturer, carrier and distributor, on top of the systems they already run.
\nThe countries that lead this century will all have capable machines. What will separate them is whether their physical economies are legible enough for those machines to coordinate with everything else.
\nSome countries will get there from the top down. China, with heavy government backing, pushed freight-matching platforms onto millions of drivers' phones to drive down empty running, and its logistics costs fell from 18% of GDP in 2012 to 14.4% in 2023.3 America's better path is a market: a neutral layer every company can plug into, with full control over what it shares. America also starts with an advantage it rarely gets credit for. Thanks to the ELD rule, nearly every truck in the country already reports where it is.
\nOway.
IOI is built to be that layer. Each company runs its own node, connected to the systems it already has, legacy and autonomous alike. The node keeps an IOI data lake built on an open-source schema, and that schema covers actions as well as state: read, write, produce, quote, book, track, tender and capacity, among others.
\nOnce a new plant is on IOI, it can see what its suppliers can deliver and when, which carriers have open space heading its way and where its inputs and outputs are at any moment. Its own agents can plan production against live conditions inside limits the plant sets. Juno can price, schedule and route its freight in seconds.
\nThe same holds for the machines. A driverless truck on IOI can claim a load that was tendered from an AS400, because the node has already translated the tender into the open schema. A robot fleet can read the orders and inventory that live in an ERP. An automated line can accept a request to produce from a customer's planning system. The legacy market keeps running as it does today, and the autonomous market gains a way to work with it immediately.
\nOway started with the empty space in trucks because it was the clearest place to prove the idea, down to the pallet. Today more than 10,000 vehicles are on our network, shippers move freight at up to 50% below market and carriers earn more on lanes they already drive.
\nIOI extends that approach to production, warehousing, robotics, energy and every operation that makes and moves things. Juno brings industrial AI that acts inside clear permissions. Oway OS is where the work runs.
\nThis is the last note in our series on indexing the physical economy, and the work is just beginning. The challenge of our generation is to make the physical world legible, coordinated and trustworthy enough for intelligence to act on. It is very winnable, and we would love to build it with you. Join the team or talk to us about your operation.
\nAmerican Transportation Research Institute (2025), An Analysis of the Operational Costs of Trucking, for 16.7% empty miles in 2024; American Trucking Associations, Trucking Trends, for average Class 8 payloads of about 29,000 lbs against rated capacities of 60,000 to 90,000 lbs. See The Half-Empty Truck. ↩
\nAurora Innovation (May 1, 2025). \"Aurora Begins Commercial Driverless Trucking in Texas.\" Business Wire. ↩
\nPeople's Daily (February 2025). \"Reduced logistics costs help release China's economic vitality.\" See also Xu, X. et al. (2026), Nature Communications 17, 4714. ↩
\nJuno is Oway's industrial AI. It reads the live state of an operation through IOI and decides what should happen next: what a piece of work should cost, where it fits, and who or what should do it. Juno's first and largest deployment prices and routes freight onto trucks already on the road, across a network of more than 10,000 vehicles.
\n| Signal | \nWhat Juno reads | \n
|---|---|
| Capacity | \nOpen space, position and route of every vehicle and asset on the network. | \n
| Identity | \nOperating authority, insurance and equipment for every carrier and vehicle. | \n
| Demand | \nRequests, orders and documents: origin, destination, quantity, weight, class and handling requirements. | \n
| Market | \nReal quotes, bookings and acceptances across the network. | \n
| Operation | \nSchedules, facility hours, dock windows and constraints from the customer's IOI node, with permission. | \n
| Output | \nWhat Juno produces | \n
|---|---|
| Price | \nA price for the work, with the cost of the detour or change it requires. | \n
| Plan | \nWhere the work fits: the vehicle, machine or crew, the stop order and the timing. | \n
| Action | \nA booking, dispatch or reroute, taken directly or proposed for approval. | \n
Benchmark reports are shared with enterprise customers and research partners. Request access.
\nJuno is built to price, schedule, route and dispatch physical work for companies that run on Oway, and for AI agents acting on their behalf through MCP.
" }, { "id": "https://www.shipoway.com/research/letting-ai-act-in-the-physical-world/", "url": "https://www.shipoway.com/research/letting-ai-act-in-the-physical-world/", "title": "Letting AI Act in the Physical World", "summary": "How Juno and other AI agents act on real operations through IOI and MCP, with read, propose and act scopes, approvals and a full audit trail.", "date_published": "2026-09-21T12:00:00Z", "authors": [ { "name": "Oway Research" } ], "tags": [ "Industrial AI", "MCP", "Security", "Juno" ], "content_html": "Software that reads an operation is useful. Software that acts on one is far more useful, and it has to be held to a much higher standard. A wrong answer from a chatbot costs a reread. A wrong action in the physical world costs a missed delivery, an idle line or money committed to the wrong carrier.
\nHere is how we think about letting AI act on the physical world.
\nJuno, Oway's industrial AI, and any MCP-compatible agent a company chooses reach an IOI node the same way: through the Model Context Protocol, with permissions the node's owner sets. Our own AI uses the same door, under the same rules, as every other agent.
\nThrough MCP, an agent works with the same IOI schema operations as any other connected system: reading inventory, checking capacity, quoting, booking, tendering a load or asking a plant to produce. What it is allowed to do with them depends on its scope.
\nEvery agent gets a scope:
\nEvery read, proposal and action is logged with its inputs and who approved it.
Oway. Scopes can be narrowed or revoked at any time.
Scopes can be narrowed to specific data, facilities or actions, and revoked at any time. Most companies start an agent on read, review its proposals for a while, and then grant it act on specific, well-understood operations.
\nSome decisions should always have a person behind them. Commitments that move money or change a plan can be set to require approval, and Juno confirms before anything is charged. A proposed fix to a dock conflict, for example, shows up for the site lead with the reasoning and the alternatives, and goes ahead only when they say so.
\nJuno follows rules that come from the physical world and from regulation. It adds a stop only when every existing delivery window still holds. It verifies carrier authority, insurance and ELD connection before freight moves. It respects the weight and space left in the trailer and the hours a driver has left. Every action runs inside the scopes the owner grants.
\nIndustrial AI reads a lot of content it did not write: EDI messages, emails, purchase orders, nightly files, PDFs and messages from partners. Any of them could contain text written to manipulate an AI. We treat all outside content as untrusted data. It can inform a decision. It can never change the rules an agent runs under or the permissions it holds.
\nEvery decision Juno or another agent makes is logged with its inputs, its outcome and who approved it. When a customer asks why a load moved or a window changed, the answer is on the record along with the data behind it.
\nAI is quickly getting good enough to run much of the coordination work people do by hand today. The companies that benefit will be the ones that can let it act, confidently, inside clear limits. IOI and MCP are built to give them exactly that.
\nFor developers, the docs cover MCP, the API and scopes. For security details, see our Security page. In the last note of this series, we step back and look at what all of this means for American industry.
" }, { "id": "https://www.shipoway.com/research/ioi-search/", "url": "https://www.shipoway.com/research/ioi-search/", "title": "IOI Search: Ask Your Operation Anything", "summary": "IOI Search lets anyone ask an industrial operation a question in plain English and get the answer, the numbers and the sources from its live data lake.", "date_published": "2026-09-07T12:00:00Z", "authors": [ { "name": "Oway Research" } ], "tags": [ "IOI", "IOI Search", "Industrial AI" ], "content_html": "Most operational questions take ten seconds to ask and two days to answer. Which sites are nearly full this week? Which suppliers keep missing their dock windows? Where is product sitting the longest?
\nToday, answering any of those means pulling exports from several systems, joining them in a spreadsheet and emailing it around, by which point the data has often changed. Analysts across industry spend a large share of their week on exactly this work.
\nIOI Search answers the question directly.
\nBecause an IOI node keeps a cleaned, current data lake of the operation, a question can reach across every connected system at once, whether the underlying data came from an ERP, EDI messages, nightly files or vehicle telemetry. You type it the way you would ask the most informed person in the building:
\nIOI Search answers in plain English, shows the numbers behind the answer and lists the sources it used, so anyone can check the work before acting on it.
\nReno DC is at capacity and inbound dwell is up for the third week. Fresno is worth watching. Dallas has room.
Illustrative example.
Ask how your facilities are doing and you get each site with how full it is, how long things are sitting, the trend over recent weeks and a status: healthy, worth watching or at capacity. Ask which suppliers miss their windows and you get them ranked by on-window performance, with the actual shipments behind each number.
\nEvery answer links back to the records it came from, whether that is an order in the ERP, an EDI 214 status message or a position from a truck's ELD.
\nTraditional reporting describes how last month went. IOI Search reads the data lake as it stands right now, so \"what needs attention today\" reflects this morning's data across every connected system and permitted partner.
\nPeople use IOI Search inside Oway OS. Software uses the same capability through the API, for example to ask a node a question from an internal tool. AI agents, including Juno, use it through MCP to understand the state of the operation before they propose or take an action.
\nSearch respects the permissions on the node. A partner or agent sees only what its grant allows, and every query is logged.
\nAn answer is most useful when it leads somewhere. If IOI Search shows two inbound loads about to collide at a site that is already nearly full, Juno can propose a fix, such as moving one delivery to an open window, and the site lead can approve it.
\nThat handoff from seeing to doing is the subject of the next note: how we let AI act in the physical world safely. You can try IOI Search on the IOI page.
" }, { "id": "https://www.shipoway.com/research/inside-an-ioi-node/", "url": "https://www.shipoway.com/research/inside-an-ioi-node/", "title": "Inside an IOI Node", "summary": "What an IOI node connects to, how its data lake cleans legacy and machine data into an open-source schema, and how permissions, audit and ownership work.", "date_published": "2026-08-17T12:00:00Z", "authors": [ { "name": "Oway Research" } ], "tags": [ "IOI", "Security", "Industrial AI", "Context layer" ], "content_html": "An IOI node is the part of IOI a company actually runs. It sits beside the systems the company already has, keeps the company's IOI data lake current and controls who and what can use it.
\nA default node is live in days. It starts as a monitoring surface for the company's freight quotes, and IOI optimizes physical operations from the secured data lake from there. Everything below is optional and can be connected at any time afterwards, without replacing anything the company already runs. The optional connections fall into two groups.
\n| Group | \nExamples | \n
|---|---|
| ERP | \nSAP ECC, Oracle EBS | \n
| EDI over a VAN | \nX12 204 load tenders, 214 status updates, 990 responses | \n
| Mainframe and AS400 | \nGreen screens and COBOL batch jobs | \n
| Flat files over SFTP | \nNightly CSV drops | \n
| Email and PDF | \nBills of lading, rate confirmations, invoices | \n
| Spreadsheets | \nExcel on a shared drive | \n
| Custom stacks | \nWhatever was built in house | \n
| Vehicle telemetry | \nPosition, hours of service, empty space | \n
| Self-driving trucks | \nVehicles that take a load and route themselves | \n
| Automated factories | \nLines, docks and throughput | \n
| Deployed robotics | \nArms and movers on the floor | \n
The first seven are the legacy surfaces where most industrial information lives today. The last four are the physical assets that increasingly do the work. A node connects to both, along with suppliers, carriers and customers who choose to connect.
\nEach of those systems describes the world in its own format. The same load can arrive as an EDI 204 from a mainframe, a rate confirmation PDF attached to an email and a row in a dispatcher's spreadsheet. The node reads all three, reconciles them and writes one clean record into the company's IOI data lake.
\nFour formats, one load
ST*204*0001~B2**OWAY**SH77412~S5*1*LD~Load SH77412 · 9 pallets · Coppell to LaredoSH77412 | 9 plt | 7,850 lb | Thu AMSH77412,COPPELL TX,LAREDO TX,9,7850IOI data lake
tender {\n load: \"SH77412\",\n origin: \"Coppell, TX\",\n dest: \"Laredo, TX\",\n pallets: 9,\n weight: 7850,\n pickup: \"Thu 08:00\"\n}
Anything that speaks IOI
The same load arrives as an EDI 204 from an AS400, a rate confirmation PDF, a spreadsheet row and a nightly file. The IOI node reconciles them into one tender record that a carrier truck, a self-driving truck or Juno can act on.
Illustrative example.
The data lake is a cleaned, queryable set of the company's industrial data, IOI-compatible and built on an open-source schema. It stays current as the operation moves. When a telematics feed shows a truck's open space change, or a nightly file updates inventory, the data lake updates, and so does every system and agent that depends on it.
\nBecause the schema is open, anything that speaks it can work with the data lake directly: the company's own software, Juno, a partner's planning system, a robot fleet or a self-driving truck. And because the schema covers actions as well as state, a node can do more than report. It can quote, book, tender, write a status update to another factory or ask a plant to produce, within the permissions the owner sets.
\nThe node and its data lake belong to the company. Nothing on it is pooled with another customer, and Oway uses only the access a company grants to its own node to run and optimize that company's work.
\nSharing is where most of the value of an index comes from, and also where most of the risk is. IOI treats every connection the same way:
\nThat lets two companies share exactly what one purpose needs, such as dock windows for a shared shipment, and nothing else. A carrier can see when a dock is open without seeing the shipper's margins.
\nA node is designed for people and for machines. People use IOI Search and Oway OS. Software uses the API to query the data lake and subscribe to changes. AI agents, including Juno, reach the node through MCP with read, propose or act scopes the owner sets.
\nIn the next note, we show what it looks like to ask a node a question in plain English. For the full security model, visit our Security page.
" }, { "id": "https://www.shipoway.com/research/introducing-ioi/", "url": "https://www.shipoway.com/research/introducing-ioi/", "title": "Introducing IOI, the Internet of Industrials", "summary": "IOI connects legacy industrial systems and autonomous machines through an open-source schema, so companies and AI can read, quote, book and produce together.", "date_published": "2026-07-27T12:00:00Z", "authors": [ { "name": "Phillip Nadjafov" } ], "tags": [ "IOI", "Industrial AI", "Physical economy", "Context layer" ], "content_html": "In The Physical Economy Has No Index, I argued that the physical economy needs a neutral, live index of what capacity exists, where it is, what it costs and what is moving through it. Today I want to introduce the system we are building to make that index possible. We call it IOI, the Internet of Industrials.
\nThe physical economy runs on two very different sets of systems.
\nThe first is the legacy surface that most of industry still runs on. ERPs such as SAP ECC and Oracle EBS. EDI over a value-added network, where a load tender is an X12 204, a status update is a 214 and an acceptance is a 990. Mainframes and AS400 green screens running COBOL jobs. Flat files dropped over SFTP every night. Bills of lading, rate confirmations and invoices sent as PDFs by email. Spreadsheets on a shared drive. And whatever custom stack each company built in house over the past thirty years.
\nThe second is the new physical layer. Telematics that report a vehicle's position, hours and empty space. Self-driving trucks that can take a load and route themselves. Automated factories with lines, docks and throughput that software can schedule. Robots that pick, move and load on the floor.
\nThese two worlds do not share a language. A self-driving truck has no way to read an EDI 204 from an AS400. An automated factory cannot answer a rate confirmation that arrives as a PDF. So people translate between them, the same way they have translated between legacy systems for decades, and the autonomous machines end up waiting on email.
\nLegacy surfaces
Internet of Industrials
IOI node and data lakeA cleaned, queryable set of your industrial data, IOI-compatible and built on an open-source schema.
Schema, for example
READhow many units of a SKU sit in a warehouse right nowWRITEsend a status update to another factoryPRODUCEask a plant to run more mild steelQUOTEprice nine pallets against trucks already rollingBOOKcall a truck for 8 units of mild steel by 8am tomorrowTRACKposition and status without a phone callTENDERoffer the load straight to the carrierCAPACITYwhat is empty tonight, and where it is headedPhysical assets
Oway. The schema is open source; these operations are examples.
IOI is the layer in the middle.
\nA few companies have already solved this problem for themselves. The largest vertically integrated operators built their own coordination layers, where factories, warehouses, fleets, robots and planning systems share one internal model of the operation. A robot, a truck and a production plan can act on the same information in real time. That capability is a large part of how those companies move as fast as they do.
\nIt also took them years and enormous engineering teams to build, and it only works inside their own walls. A mid-sized manufacturer, a regional carrier or a distributor running on an AS400 has no realistic way to build the same thing.
\nIOI makes that capability open and easy to set up. Connecting a node works much like adding a carrier, with no multi-year integration project, and every company that connects can coordinate with the others through the same open schema. The coordination advantage that used to require owning the whole stack becomes available to the entire market.
\nEach company runs its own IOI node. The node connects to the systems the company already has, on both sides, and nothing gets ripped out. Inside the node is the company's IOI data lake: a cleaned, queryable set of its industrial data, IOI-compatible and built on an open-source schema.
\nA default setup takes days. Every node starts as a monitoring surface for the company's freight quotes, and IOI optimizes physical operations from the secured data lake from there. ERP, EDI and every other surface are optional, and can be added at any time afterwards.
\nAs a node grows, it follows three steps:
\nFour formats, one load
ST*204*0001~B2**OWAY**SH77412~S5*1*LD~Load SH77412 · 9 pallets · Coppell to LaredoSH77412 | 9 plt | 7,850 lb | Thu AMSH77412,COPPELL TX,LAREDO TX,9,7850IOI data lake
tender {\n load: \"SH77412\",\n origin: \"Coppell, TX\",\n dest: \"Laredo, TX\",\n pallets: 9,\n weight: 7850,\n pickup: \"Thu 08:00\"\n}
Anything that speaks IOI
The same load arrives as an EDI 204 from an AS400, a rate confirmation PDF, a spreadsheet row and a nightly file. The IOI node reconciles them into one tender record that a carrier truck, a self-driving truck or Juno can act on.
Illustrative example.
IOI's schema covers taking action as well as reading state, between companies and between machines. A few examples of what it covers:
\n| Operation | \nExample | \n
|---|---|
| Read | \nHow many units of a SKU sit in a warehouse right now | \n
| Write | \nSend a status update to another factory | \n
| Produce | \nAsk a plant to run more mild steel | \n
| Quote | \nPrice nine pallets against trucks already rolling | \n
| Book | \nCall a truck for 8 units of mild steel by 8am tomorrow | \n
| Track | \nPosition and status without a phone call | \n
| Tender | \nOffer the load straight to the carrier | \n
| Capacity | \nWhat is empty tonight, and where it is headed | \n
Because every connected company speaks the same schema, these operations work across company lines. A manufacturer's planning agent can check a supplier's capacity, ask its plant to produce, book a truck with open space heading that way and track the result, without anyone re-keying an EDI message or forwarding a PDF. A self-driving truck can claim a tender that originated on an AS400. The legacy system keeps running exactly as it does today. It simply gains a counterpart that can act on what it says.
\nThe schema is open source because a common language only works if anyone can speak it. Fleets, factories, robot makers and software companies can build to it directly, and every company that connects makes the network more useful for every other.
\nThe node and its data lake belong to the company that runs them. Its data is never pooled with another customer's. Connecting a partner, an agent or a machine takes the owner's explicit permission. Every grant has a defined scope, every grant can be narrowed or revoked at any time, and every access is logged. AI agents get one of three scopes: read, propose or act.
\nWe learned the importance of this in freight. Carriers will not share where their open space is if it helps a competitor undercut them, and shippers will not share their volumes if they end up in someone else's rate sheet. An index only works when the people feeding it trust it. What becomes shared knowledge is the aggregate shape of the economy. Every individual position stays private.
\nIOI is the foundation for everything Oway builds:
\nFreight was the first surface because the data already existed, thanks to the ELD rule, and the waste was easy to measure. IOI is how the same approach reaches production, warehousing, robotics, energy and every operation that makes and moves things.
\nOver the next few notes we will open up an IOI node, show what IOI Search can answer, and explain how we let AI act in the physical world safely. To see it on your own operation, request a demo or read the docs.
" }, { "id": "https://www.shipoway.com/research/the-physical-economy-has-no-index/", "url": "https://www.shipoway.com/research/the-physical-economy-has-no-index/", "title": "The Physical Economy Has No Index", "summary": "The internet became useful once it was indexed. The physical economy still has no live index of its capacity. Why that costs trillions, and how to build one.", "date_published": "2026-07-06T12:00:00Z", "authors": [ { "name": "Phillip Nadjafov" } ], "tags": [ "IOI", "Physical economy", "Industrial AI", "Reindustrialization" ], "content_html": "One of the largest industrial problems of our time is slowly killing our economy and most have no idea it exists. It touches every physical thing around you and charges a tax you don’t get a benefit for.
\nThe American industrial sector is operating at around half of its productive efficiency, irrespective of economic conditions.1 That might sound like a shock, but it’s by design of the vehicles moving everything around: the trucks. This trillion-dollar industry acts as an artery of the physical economy and doesn’t have a mechanism to know how truly inefficient it is. Because of the lack of a reliable big nervous system, trucks depart warehouses with only one-third to half of their capacity filled and cannot utilize their empty space even when there is high demand.2 Warehouses sit idle, machines wait for work that is already needed three states away, robots cannot fix it, and you the consumer pay the cost.
\nAmerica is flying blind through the heart of its own economy, paying a tax it doesn’t even know exists. Fixing it would unlock hundreds of billions of dollars in productivity in America alone.
\nAnd the pieces have now fallen into place to do it. In fact, we have already solved this problem once.
\nThe internet indexed humanity's collective data, search made it query-able. Limitless, effectively free information enabled us to find and match novel wasted supply to demand instantly at almost no cost.3 It took historically binary decisions and split them into countless new outcomes. The payoff was trillions of dollars in value and a generation of asset-light companies that we now use as utilities, all built on one simple idea: if you can see everything and match it for the right use case instantly, historical waste turns into value.
\nThis same pattern is beginning to fall into place again. Over decades, companies have started connecting their factories, vehicles, and robots to the internet, and the sensors on these devices are transmitting new, novel information we’ve never seen before. Unfortunately, the physical economy is still stuck in behavioral gridlock. The industrial economy, which is the overwhelming majority of what humans actually produce and the entirety of our survival (food, medicine, materials, every object sitting around you), still gets orchestrated roughly the way it did in the 1980s. Binary, simple decisions because we physically cannot do better relative to the time and information available. And so far, we have accepted that as the best it can be. The bill for that acceptance is not abstract: moving goods through the US economy now costs $2.6 trillion a year, 8.7 percent of GDP.4
\nWe are missing a layer.** **
\nThe physical economy needs its own index. A live matrix of what productive capacity exists, where it is, what it costs, and what is actually moving through it. This index has to be neutral, akin to how the index of the internet belongs to no single website. Neutral rails that everyone can use without handing their business to a competitor. Once that index exists, three things happen almost instantly. Matching becomes instant. Waste becomes visible. Prices will equilibrate. All because there finally exists a physical source of truth.5
\nI started working on this problem with my team, beginning where the physical economy is most broken and most legible at the same time: the supply chain. Specifically, we believe the empty space in trucks can create a structurally new category of moving goods. We are building an index of live, empty carrier capacity, and treating it as a real-time measure of how inefficient the economy actually is. We operate this network with AI that reads the index and decides what should go where, filling capacity that would otherwise have driven empty.
\nA wave of startups tried to fix our supply chain. Most of them died trying. It is worth being precise about why. They digitized the broker, not the market: apps layered over phone calls, marketplaces competing on price. And behind every load? A human still doing the actual coordination. The cost of coordinating never fell, so they inherited the same thin margins as the industry they set out to replace.6 None of them built the index; matching without an index is just guessing faster. And the coordination itself could not be automated, because software could not yet read context. That constraint has expired. We do not need more logistics companies doing the old thing slightly faster. We need the neutral coordination layer the physical economy was never given, and indexing trucks is only the first surface.
\nThis is also why the index has to be built as infrastructure from day one, not retrofitted later. Every load coordinated through it becomes structured data: what moved, where, when, at what price, and what capacity was left behind. Compounded over years, that becomes the ground truth of the physical economy, a data backend designed to be read not just by people but by machines. An autonomous truck does not need an app. It needs an API to the economy: what to carry, where to go, what the trip is worth. A dark factory needs to know what to build before anyone asks. Autonomy without context is a fleet of brilliant machines with no nervous system. We are designing the index as that backend now, so that when the machines arrive, the world they need to read already exists. We will not build the trucks or the factories or the robots. We are building the thing they will all have to plug into.
\nAn index this powerful only works if it is trusted, because the heart of the industrial economy also houses its most sensitive information on private business behavior. Who ships what, to whom, when, and at what cost is some of the most sacred information companies have. Markets with hidden information do not merely leak value; they unravel entirely when participants cannot trust what they cannot verify.7 A neutral coordination layer must abstract specifics away by design, not as a feature bolted on later.
\nThe index has to surface the match without exposing the participants. It has to give every player its full benefit without forcing them to surrender position or leverage to access it. The index itself should ultimately be open, a reference anyone can build on. And what compounds into that reference is the aggregate, never the participant: the shape of the economy becomes shared knowledge while every individual position stays private. That is the deal, and it only works one way. Neutrality and privacy are not constraints on this idea. They are what make the rails usable by everyone at once. You earn the right to coordinate the economy only by being the party that everyone can plug into without fear.
\nThe part that matters beyond efficiency is what this index does for everyone globally. Coordination is deflationary in the best possible sense of the word. Every idle machine, every empty mile, every shipment that sits and waits is a cost, and that cost does not vanish. It is passed on to you.8 The waste of the physical economy is a hidden variable in your rent, your groceries, your medicine, and every object around you. You pay for the disorder of a system you cannot even see.
\nMinimize the waste in coordination and you lower the cost of physical existence itself. Then add the machines. When they plug into a layer that already knows what the world needs and where, the cost of moving and making things falls toward the cost of the energy and intelligence required to do it. That is the difference between a century of material abundance and a century of stagnation: affordability not as a slogan but as a direct consequence of letting intelligence act on a world it can finally read.
\nThere is a larger clock running. America has decided to reindustrialize: new factories, new fabs, new energy, new defense capacity, hundreds of billions in steel and silicon going into the ground. But we are attempting to rebuild an industrial base on top of a coordination layer from the 1980s. Every new plant inherits the same darkness: the phone calls, the guesswork, the waste. Reindustrialization without an index is pouring concrete into a blind system. The nations that win this century will not be the ones with the most machines. They will be the ones whose physical economies are legible enough for machines to run. Some will coordinate their industry by mandate. We will do it by market. That is the challenge of our generation, and it is winnable.
\nNow is the time for founders who look at the physical world the way the last generation looked at information and simply refuse to accept that it has to stay dark, fragmented, and dumb. We need coordination layers for energy, materials, manufacturing, and verticals that do not even exist yet. The people who successfully build them will be among the most important builders of the century, not because they optimized a niche, but because they are giving the physical economy the nervous system it was never born with.
\nThe future everyone keeps promising, the self-driving everything and the robotics that build and the abundance that follows, does not arrive on the back of better hardware alone. It arrives the moment the physical world becomes legible, coordinated, and trustworthy enough for intelligence to act on it.
\nSomeone has to build that layer.
\nWe are building a part of it. Come build the rest.
\nIn manufacturing, Overall Equipment Effectiveness (OEE), the standard measure of realized output against productive potential (Nakajima, S., 1988, Introduction to TPM, Productivity Press), averages roughly 60 percent across industry benchmark surveys, with 40 to 60 percent typical of most operations and 85 percent considered world class. Peer-reviewed studies of road freight measure the same shape: average vehicle load factors at or below half of capacity (e.g., Transportation Research Part E, 2021; European Environment Agency load factor methodology, 2001). ↩
In US trucking, empty (deadhead) miles averaged 16.7 percent of all miles driven in 2024, per the American Transportation Research Institute (2025), An Analysis of the Operational Costs of Trucking; ATRI is a 501(c)(3) nonprofit research organization. American Trucking Associations fleet data has shown average Class 8 payloads of roughly 29,000 pounds against rated capacities of 60,000 to 90,000 pounds (Trucking Trends), a pattern that has persisted across decades and cycles. ↩
Rochet, J.-C., and Tirole, J. (2003). “Platform Competition in Two-Sided Markets.” Journal of the European Economic Association, 1(4), 990-1029. ↩
Council of Supply Chain Management Professionals and Kearney (2025). Annual State of Logistics Report. US business logistics costs totaled $2.6 trillion, 8.7 percent of GDP. ↩
Hayek, F. A. (1945). “The Use of Knowledge in Society.” American Economic Review, 35(4), 519-530. ↩
Coase, R. H. (1937). “The Nature of the Firm.” Economica, 4(16), 386-405. ↩
Akerlof, G. A. (1970). “The Market for ‘Lemons’: Quality Uncertainty and the Market Mechanism.” Quarterly Journal of Economics, 84(3), 488-500. ↩
The American Transportation Research Institute (2025) puts the average marginal cost of operating a truck at $2.26 per mile in 2024. Every one of the industry’s empty miles incurs this cost while generating no value, an expense ultimately embedded in consumer prices. ↩
Today more than 10,000 vehicles are on the Oway network. People assume the hard part of running it is finding trucks.
\nAmerica has plenty of trucks, and plenty of open space on them. The hard part is knowing, in time to act, which truck has room for a given shipment and is heading the right way.
\nOn any given morning there is an enormous amount of open space on American highways. As we covered earlier in this series, one mile in six is driven completely empty, and the average loaded truck runs at a third to half of its rated capacity.1 That is roughly 32 billion empty miles a year, plus the unused space on every loaded trip.
\nThe space is spread across thousands of trucks, changes by the minute and is known only to the drivers and dispatchers inside them. By the time that information travels through a phone call, an email and a load board to a shipper, the truck has usually left. Capacity information has a shelf life measured in minutes.
\nELDs solved part of this. Once carriers connect to Oway through them, Juno can see where open space is now and where it will be in an hour. On active lanes, a truck can be at a shipper's dock in an average of 30 minutes.
\nFinding a truck with space is the first of six steps. Turning a match into a delivered shipment also means:
\nOway.
Across most of the industry, people still do each of those steps by hand, moving information between EDI messages, PDF rate confirmations, emails and spreadsheets. That labor is a large part of why moving a few pallets costs what it does. It is also why so many freight startups that digitized the broker never changed the economics. The interface improved, while a person still did the coordination behind it.
\nWe automated those steps one at a time. Every one depended on the same thing: a reliable, current picture of what exists, where it is, what it can do and what it costs.
\nWhen that picture was stale, every automated decision got worse. A price built on yesterday's capacity is confidently wrong. When the picture was accurate, decisions that used to take a day took seconds.
\nThe lesson extends well beyond freight. The physical economy is limited by coordination, and coordination is limited by information that nobody has been able to share live, safely and without giving up their position to a competitor. China showed in freight how quickly load factors rise once capacity becomes visible. The same holds for factory lines, warehouses and every other asset with time to sell.
\nThe internet had this problem before search. The information existed, scattered across millions of computers, and most of it was effectively unfindable. Indexing it made it usable, and a large share of the modern economy grew on top of that index.
\nThe physical economy never got that layer. I wrote about what it would take in the next piece in this series, The Physical Economy Has No Index. After that, we introduce IOI, the system we are building to give it one.
\nAmerican Transportation Research Institute (2025), An Analysis of the Operational Costs of Trucking, for 16.7% empty miles in 2024; American Trucking Associations, Trucking Trends, for average Class 8 payloads of about 29,000 lbs against rated capacities of 60,000 to 90,000 lbs; Federal Highway Administration, Highway Statistics 2024, Table VM-1, for 192.5 billion combination-truck miles. See The Half-Empty Truck. ↩
\nFor most of trucking history, a driver's record of where he had been and how long he had driven was a paper logbook. The industry nicknamed them comic books, because paper logs were easy to fill in creatively.
\nThat detail matters because some of the most important data infrastructure in American freight came from a federal rulemaking about those logbooks. It did not come from a startup or a technology company.
\nIn December 2015, the Federal Motor Carrier Safety Administration published the electronic logging device rule.1 The goal was simple and sensible: enforce hours-of-service limits so tired drivers stop driving. Paper logs were easy to fudge. An electronic device wired to the engine is not.
\nThe rule phased in over four years. Carriers had to install compliant ELDs by December 18, 2017, and trucks running older automatic recorders were given until December 16, 2019.2 By the end of that decade, roughly 3 million drivers were logging their hours on a connected device, and FMCSA expected the change to save the industry more than a billion dollars a year, mostly in paperwork and crashes avoided.3
\nSources: FMCSA ELD rule and implementation timeline; DAT Freight and Analytics; Oway.
The side effect is what matters for the physical economy. To log hours, an ELD has to know when the engine is running, how far the truck has moved and where it is. So the rule put a GPS-connected computer into nearly every interstate truck in the country, recording position, miles and duty status around the clock.
\nA rule written to enforce driver rest also produced something nobody had before: a live, machine-readable map of where American freight capacity is.
\nFor a few years, that data mostly served compliance. Then the pandemic arrived and the entire country took a crash course in supply chains, from port congestion to empty shelves.
\nTrucking had one of the strangest years in its history. Demand swung from collapse to a buying frenzy in a matter of months, and by late 2020 the national dry van spot rate had climbed to about $2.35 a mile, its highest in five years.4 Every shipper wanted to know exactly where its freight was, and a phone call to the driver stopped being an acceptable answer.
\nSo carriers started sharing what their ELDs knew. Location feeds that used to stay inside a fleet's office got piped, with permission, to shippers, brokers and tracking tools. Live tracking went from a premium feature to the price of admission.
\nWhen the market swung the other way and rates fell, the lesson flipped too. Carriers who had spent 2020 turning freight away suddenly cared about every empty pallet position on every trip. Same data, new question: where is the space I am not getting paid for?
\nPut those two shocks together and something new exists. For the first time, a large share of America's freight capacity is reporting where it is, how long its driver can keep driving and where it is headed, live, in a format software can read.
\nThat is the beginning of what I call a physical context layer: a machine-readable description of the real world that software and AI can reason about. China built its matching layer by putting an app in every driver's pocket. America got a sensor wired into the engine of nearly every truck, by regulation, and then a pandemic taught everyone to share it.
\nWhat coordination needs Added by Oway and IOI
What an ELD knows In nearly every cab since 2019
Oway.
It is only the beginning, though. An ELD knows where the truck is and how long the driver can drive. It has no idea what is inside the trailer, what the truck is allowed to carry, how much room is left or what the plant down the road needs moved today. The load factor, the thing that actually matters, is invisible to it.
\nThat is the layer we went after. When carriers connect to Oway through their ELDs, Juno combines encrypted location and route data with what it knows about the freight already on board. It can see where open space is right now and where it will be in an hour. On active lanes, that means a truck at a shipper's dock in about 30 minutes on average.
\nA regulation meant to enforce driver rest became one of the most important freight data projects in American history, without anyone designing it to be one.
\nIt also taught us how much is still missing. Location is one signal. Coordination needs all of them: capacity, constraints, timing, cost and trust, shared across companies that do not want to give up their position. That is the problem the next note is about.
\nFederal Motor Carrier Safety Administration (2015). Electronic Logging Devices and Hours of Service Supporting Documents, Final Rule, 80 Fed. Reg. 78292, published December 16, 2015. ↩
\nFMCSA ELD implementation timeline. Compliance date December 18, 2017; carriers using grandfathered automatic onboard recording devices (AOBRDs) had until December 16, 2019. ↩
\nFMCSA regulatory evaluation of the ELD rule. Roughly 3 million drivers affected; estimated net benefits of more than $1 billion a year, primarily from reduced paperwork and crash reductions. ↩
\nDAT Freight and Analytics, 2020 truckload market reports. National average dry van spot rate of about $2.35 per mile in late 2020, a five-year high. ↩
\nFreight pricing was built for a world where a truck is either yours or someone else's. You look up the lane, check the rate table, add a fuel surcharge and send a quote.
\nThat model has nothing to say about the truck that is already driving to Houston with half its trailer empty. The truck is going anyway. The diesel is already burning. The real question is what it costs to add your freight to that trip, right now, without putting any delivery already on board at risk. That is the question Juno answers.
\nWhen Juno prices a partial load, it is answering four questions at the same time:
\nA broker answers those by calling around, which takes hours. Juno answers them in seconds, from live data.
\nLive signals
Juno
Outputs
Oway. Signals are read with each operation's permission.
Juno works from five groups of live signals:
\n| Signal | \nWhat Juno reads | \n
|---|---|
| Capacity | \nOpen space, position and route of every vehicle on the network | \n
| Identity | \nOperating authority, insurance and equipment for every carrier and vehicle | \n
| Demand | \nOrigin, destination, pallets, weight, class and handling requirements | \n
| Market | \nReal quotes, bookings and acceptances across the network | \n
| Operation | \nSchedules, facility hours and dock windows, with permission | \n
Out the other side come three things. A price that includes the cost of the detour. A plan that says which truck takes the freight and in what stop order. And an action: the booking itself, made directly or sent to a person for approval.
\nTake a 53-foot dry van rolling south from Dallas to Houston with 13 of its 26 pallet positions open. By pallet positions, that is a load factor of 50%, close to the American average for a loaded truck.
\nA plant near the route needs 13 pallets moved the same direction. Juno checks the carrier's authority, insurance and ELD connection, confirms that the weight and the space fit, and works out the detour: about six miles and eleven minutes off the route, with the Houston delivery still on time.
\nBefore: 13 of 26 positions full
After: 26 of 26
A 53-foot dry van from Dallas to Houston with 13 of 26 pallet positions full picks up 13 more pallets for a 6.2 mile, 11 minute detour. The Houston delivery stays on time, the shipper pays 33 percent below market and the carrier earns 20 percent more on the lane.
Illustrative example based on a typical Oway Rideshare move.
Because the truck was making the trip anyway, the shipper pays well below market and the carrier earns revenue on space that would otherwise have run empty. In our illustrative version of this trip, that comes out to a shipper rate 33% below market and 20% more revenue on the lane for the carrier. Juno prices, checks and books it in seconds, and the truck's load factor goes from 50% to 100%.
\nA fast price is only useful if it is safe to act on. Juno adds a stop only when the existing delivery windows hold. Every action runs inside the scopes the operation's owner grants. Anything that commits money or changes a plan can require a person's approval, and every decision is logged with what Juno saw and what it did. We would rather lose a quote than cause a missed dock appointment.
\nWe hold Juno to four measures:
\nBuilding this taught us that a good price depends most on knowing where the trucks are and how much room they have, minute by minute. For most of trucking history nobody had that information. Then a federal regulation put a connected device in nearly every cab in America. That story is next. You can read more about Juno on its platform page.
" }, { "id": "https://www.shipoway.com/research/the-half-empty-truck/", "url": "https://www.shipoway.com/research/the-half-empty-truck/", "title": "The Half-Empty Truck", "summary": "US trucks drive one mile in six empty and run loaded at a third to half of capacity. What that costs, and how China cut empty running with freight matching.", "date_published": "2026-04-27T12:00:00Z", "authors": [ { "name": "Oway Research" } ], "tags": [ "Freight", "Load factor", "Oway Rideshare", "Physical economy" ], "content_html": "Stand at the exit of any American distribution center for an hour and count the trucks. A large share of them are hauling air, at highway speed, with a professional driver, at about $2.26 a mile.1
\nThe measure for this is load factor: the share of a vehicle's capacity that is actually carrying freight. It is one of the most important numbers in the physical economy, and one of the least discussed outside of freight.
\nA truck can waste capacity in two ways.
\nThe first is driving with nothing in it. These are deadhead miles. The American Transportation Research Institute found that 16.7% of all US truck miles in 2024 were driven completely empty, exactly one mile in six.1
\nIn 2024, 16.7 percent of US truck miles were driven empty, one mile in six. Across 192.5 billion combination-truck miles, that is about 32 billion empty miles costing roughly 70 billion dollars a year.
Sources: ATRI (2025); FHWA Highway Statistics 2024, VM-1. Cost is an Oway estimate at $2.26 per mile.
Scale that to the country. US combination trucks drove about 192.5 billion miles in 2024.2 At a 16.7% empty rate, that is roughly 32 billion empty miles. At ATRI's marginal operating cost of $2.26 a mile, those miles cost on the order of $70 billion a year, spent moving nothing.3
\nThe second kind of waste is quieter and larger: a truck that is loaded but far from full. American Trucking Associations fleet data has shown average Class 8 payloads of about 29,000 pounds against rated capacities of 60,000 to 90,000 pounds.4 That puts the average loaded American truck at a load factor of roughly 32% to 48% by weight.
\nThe average US Class 8 truck carries about 29,000 pounds. Against rated capacities of 60,000 pounds that is a 48 percent load factor, and against 90,000 pounds it is 32 percent.
Source: American Trucking Associations, Trucking Trends. Load factor = average payload / rated capacity.
Put together, America runs a fleet that drives a sixth of its miles empty and most of the rest a third to half full. Every one of those miles still pays for diesel, tires, insurance and a driver. Carriers price the waste into their rates, shippers pass it into their prices, and consumers pay it at the shelf. Moving goods through the US economy cost $2.6 trillion in 2024, about 8.7% of GDP.5
\nA shipper with six pallets has two standard options, and neither is efficient.
\nLTL picks the freight up, takes it to a terminal, unloads it, sorts it, loads it onto another truck and often handles it again at a second terminal before delivery. Every touch adds time, cost and a chance of damage, and accessorial and reclassification charges tend to arrive after the fact.
\nA full truckload moves the freight directly, at the price of 26 pallet positions to carry six. The shipper pays for 20 positions of empty space.
\nBoth options exist because the empty space on trucks already heading the same direction is invisible to the shipper at the moment it matters.
\nChina faced the same problem at a larger scale and with a far more fragmented industry. Its long-haul trucking runs on millions of owner-operators, and for years the commonly cited estimate was that around 40% of truck trips ran empty. A study by China's Ministry of Transport used 45% as its baseline.6 Chinese heavy trucks averaged about 36,000 km a year against 90,000 km in the US, largely because drivers waited days between loads.7
\nChina's response was to build a matching layer on the driver's phone, and it worked. Freight-matching platforms let shippers post loads and let nearby drivers claim them. By 2025, China had more than 3,000 digital freight platforms coordinating over 8 million trucks, and the time drivers spent waiting for a load fell from two to three days to eight to ten hours.8 The largest platform alone fulfilled 197 million orders in 2024.9
\nThe results show up in the data. GPS records from 2021 put empty running at about 27% for Chinese tractor-trailers, well below the old 40% figure. Trucks matched through these platforms run at load factors approaching 90%.10 Across the whole economy, China's logistics costs fell from 18% of GDP in 2012 to 14.4% in 2023.8
\nEmpty running, Chinese trucks
Load factor
Chinese tractor-trailer empty running fell from a commonly cited 40 percent to 27 percent in 2021. Platform-matched Chinese trucks run near 90 percent load factor, against 32 to 48 percent for the average loaded US truck. Waiting time fell from 2 to 3 days to 8 to 10 hours, and China logistics costs fell from 18 to 14.4 percent of GDP.
Sources: Xu et al., Nature Communications (2026); People’s Daily (2025); American Trucking Associations. Matched-truck load factor comes from a platform sample.
China's work is far from finished. Platform matching still covers only about a fifth of its road freight, and its overall empty rate remains higher than America's.10 But the direction is clear, and so is the lesson: once spare capacity becomes visible and matchable, load factors rise quickly. China had less trucking data than America and built the matching layer anyway. America has something better than a phone app in nearly every truck cab, which is the subject of part four.
\nFor most small shipments there is a better fit: put them in the open space of a truck already going the same way. We call it Oway Rideshare.
\nA shipper gives us four things: pickup ZIP, delivery ZIP, pallet count and total weight. Juno finds trucks on the network with open positions on a compatible route, checks that the pickup fits without breaking anyone's delivery window, and prices the move in seconds. The freight is loaded once and rides straight through.
\nShippers move LTL and partials from 1 to 18 pallets at up to 50% below market, with an average pickup time of 30 minutes on active lanes. Carriers get paid for space they were going to haul anyway and have earned up to 30% more revenue per lane. Every match raises the load factor of a truck that was already on the road.
\nFilling empty trucks works. The harder question is why the space stays empty when the truck, the freight and the road are all right there. The answer is information that arrives too late to act on: where the space is, what it is worth and whether a stop fits the schedule.
\nThat is a pricing problem and a coordination problem, the subjects of the next notes in this series.
\nAmerican Transportation Research Institute (2025). An Analysis of the Operational Costs of Trucking: 2025 Update. Deadhead miles were 16.7% of all miles in 2024; average marginal operating cost was $2.26 per mile. ↩↩
\nFederal Highway Administration, Highway Statistics 2024, Table VM-1. Combination trucks traveled 192,520 million vehicle-miles in 2024. ↩
\nOway estimate: 192.5 billion combination-truck miles × 16.7% empty × $2.26 per mile ≈ $72.7 billion. ATRI's empty-mile share and cost come from its fleet survey and are applied here to the national fleet as an approximation. ↩
\nAmerican Trucking Associations, Trucking Trends. Average Class 8 payloads of about 29,000 lbs against rated capacities of 60,000 to 90,000 lbs. Load factor here is payload divided by rated capacity. ↩
\nCouncil of Supply Chain Management Professionals and Kearney (2025). Annual State of Logistics Report. ↩
\nXinhua, reporting research by China's Ministry of Transport with a national freight-matching platform. Lowering the empty-load rate from 45% to 25% could cut CO2 emissions by 69.51 million tonnes a year. ↩
\nInternational Council on Clean Transportation. Barriers to energy efficiency in China's long-haul freight. Heavy trucks average about 36,000 km a year in China against 90,000 km in the US. ↩
\nPeople's Daily (February 2025). \"Reduced logistics costs help release China's economic vitality.\" 3,286 digital freight platforms coordinating over 8 million trucks; waiting times cut from 2 to 3 days to 8 to 10 hours; social logistics costs fell from 18% of GDP in 2012 to 14.4% in 2023. ↩↩
\nChina's largest digital freight-matching platform, fiscal year 2024 results: 197.2 million fulfilled orders and 2.64 million average monthly active shippers. ↩
\nXu, X. et al. (2026). \"The potential role of truck-hailing and operational efficiency improvement in China's road freight decarbonization.\" Nature Communications 17, 4714. 2021 GPS data: 27% empty running for tractor-trailers, 34% to 36% for straight and dump trucks; load factors near 90% in the truck-hailing sample; truck-hailing penetration of about 22%. ↩↩
\nA factory makes decisions all day. Which supplier ships first. Which line runs which order. Whether the part in one building is the part a team three states away has been waiting on since Tuesday. Whether a truck can take one more pallet today, or whether that pallet waits until Thursday.
\nEvery one of those decisions depended on information that lived somewhere else. The ERP knew what we ordered. The floor knew what we built. Suppliers knew what they shipped, carriers knew where their trucks were, and none of those systems could see each other. So people carried the information between them, by phone, by email and in spreadsheets that only one person fully understood.
\nWe were good at it. We grew because our people coordinated faster than our competitors did. But I always knew what we were paying for that speed: buffer inventory, idle lines, half-full trucks and engineers spending their afternoons chasing status updates.
\nWhen I compared notes with other operators, I heard the same story in every industry. Different products, identical blind spots. Every company had spent heavily on machines and software. Almost none had a way for those systems to share what they knew with each other, or with partners, in real time and on their own terms.
\nThe cost is large. Moving goods through the US economy cost $2.6 trillion in 2024, about 8.7% of GDP.1 A meaningful share of that pays for coordination failures: trucks driving empty, trucks driving half full, freight waiting in terminals and inventory parked as insurance against not knowing.
\nFor a long time there was a good reason nobody fixed it. Coordination depends on context, and context is messy: an EDI message from a system built in the 1980s, a PDF purchase order, an email that says \"can we push to Thursday,\" a driver running late because a dock in Ohio is backed up. Software could store that context. It could never read it.
\nThat changed. AI can now read the messy, real-world context coordination depends on and act on it inside clear limits. We founded Oway in 2023 to work on problems that simply were not solvable before.
\nWe needed a first problem where the waste was enormous, the data existed and the result could be counted to the dollar. American freight checked all three boxes.
\nOne in every six miles a US truck drives is driven completely empty, and the average loaded truck carries a third to half of what it is rated to haul.2 A few miles away from that empty space, a shipper is paying premium LTL rates to send six pallets through a chain of terminals where they will be unloaded, sorted and reloaded. The capacity and the demand are often on the same highway, at the same time, going the same direction. They cannot see each other.
\nSo we built the thing that lets them. Oway OS lets a shipper move 1 to 18 pallets in space a truck is already carrying, at up to 50% below market. Carriers earn more on routes they already drive. Juno, our industrial AI, prices and books each move in seconds. Today more than 10,000 vehicles are on the Oway network.
\nTrucks are the first surface. The same pattern shows up across industry: production lines, warehouses, robots and energy. Capacity that exists, demand that exists, and no shared, trusted way to match the two. That is why we are building IOI, the Internet of Industrials, so the systems the physical economy already runs on, and the autonomous machines arriving now, can share what they know and act on it together, with permission.
\nThis series is the story of what we learned along the way: what America's half-empty trucks cost, how China raised its own load factors with freight matching, how a regulation and a pandemic gave America the beginnings of a physical context layer, and what it takes to make the physical economy legible to software and AI.
\nIf that sounds like work you want to do, come build it with us.
\nCouncil of Supply Chain Management Professionals and Kearney (2025). Annual State of Logistics Report. ↩
\nAmerican Transportation Research Institute (2025), An Analysis of the Operational Costs of Trucking, for 16.7% empty miles in 2024; American Trucking Associations, Trucking Trends, for average Class 8 payloads of about 29,000 lbs against rated capacities of 60,000 to 90,000 lbs. ↩
\nThe transportation industry is currently facing a perfect storm. As fuel prices climb to historic highs, the traditional model of freight, characterized by half-empty trucks and fragmented routes, is no longer just inefficient. It's unsustainable.
\nFor carriers, high fuel costs eat directly into margins. For shippers, those costs are passed down as surcharges that hurt the bottom line. But what if there was a way to bypass these rising costs while actually speeding up delivery?
\nEnter Oway, the next-generation logistics platform designed to optimize every mile driven.
\nHere is why the current economic climate makes joining the Oway ecosystem a strategic necessity for both carriers and shippers.
\nIn an era of expensive diesel, \"deadhead\" miles (empty miles) and LTL (Less-than-Truckload) inefficiency are the enemies of profit.
\nFor Carriers: Oway's intelligent matching system ensures that your trucks aren't just moving; they're moving full. By consolidating shipments and optimizing route density, Oway helps carriers maximize revenue per gallon. You stop paying to haul air and start getting paid for every square inch of trailer space.
\nFor Shippers: By leveraging Oway's network of shared capacity, shippers can avoid the \"fuel surcharge trap.\" Instead of paying for a whole truck when you only need half, Oway's platform finds the most cost-efficient slot in a vehicle already heading your way.
\nTraditional logistics often trades speed for cost. Oway's platform uses advanced algorithms to prove you can have both.
\nBy utilizing a dynamic network of localized carriers and relay-style logistics, Oway minimizes the time goods spend sitting in massive, bloated distribution centers. Because the platform identifies the most direct and efficient path, often utilizing carriers already on that specific corridor, delivery times are slashed. In a world of next-day expectations, Oway gives you a head start without the premium price tag.
\nThe most effective way to reduce carbon emissions in logistics is simple: drive fewer miles.
\nThe current fuel crisis has aligned economic incentives with environmental ones. Every gallon of fuel saved is a reduction in overhead and a reduction in carbon footprint. Oway's platform is built on the principle of logistics synchronization: by filling existing trucks and optimizing routes to prevent idling and backtracking, Oway helps partners meet their ESG (Environmental, Social, and Governance) goals naturally.
\nWorking with Oway doesn't just save money. It makes your supply chain part of the solution to a greener planet.
\nThe Oway platform aligns closely with the N.O.W. Philosophy, particularly as an example of an \"Autonomous Orchestrator\" that addresses the inefficiencies of the Great Reconfiguration. By utilizing AI to monetize unused truck space (a $100B+ inefficiency), Oway moves beyond traditional, linear logistics into the realm of real-time, networked execution.
\nOway moves operations from siloed to connected by acting as a Cognitive Ecosystem.
\nAutonomous Execution: Instead of manual tasks like calling brokers, Oway uses AI to automatically identify and match available pallet space with shipper needs in real-time.
\nMulti-Tier Connectivity: By integrating directly with carrier networks and telematics, it normalizes data across a broad partner network, ensuring intelligence is based on live capacity rather than historical averages.
\nDecision Velocity: The platform transforms simple visibility into rapid action, allowing shippers to book shipments in seconds, drastically increasing supply chain speed.
\nOway provides the elastic infrastructure necessary to pivot operations in hours, not months.
\nEvent Orchestration: The platform moves beyond alert fatigue by triggering automated workflows: when a truck has extra space, the system proactively offers that capacity to a shipper who needs it.
\nDynamic Optimization: It replaces static routing guides with continuous scenario simulation, adjusting assets instantly to fill unused space in the network.
\nGoverned Autonomy: While the AI handles complex matching and pricing, it operates within the guardrails of shipper requirements, ensuring high-speed execution remains controlled and reliable.
\nOway treats transparency as a new currency by eradicating the blind spots inherent in traditional LTL shipping.
\nEradicating Blind Spots: By tracking the exact location and available volume of vehicles in the network, Oway provides visibility that extends from the carrier's current route to the shipper's final mile.
\nReal-Time Intelligence: The platform uses high-fidelity data to provide accurate predictive alerts, moving away from average lead times that lead to stockouts or excess inventory.
\nMulti-Dimensional Analytics: Because the system maps existing routes, it can provide an accurate real-time carbon footprint for every SKU, calculating the emissions saved through optimized ridesharing.
\nOway's architecture directly counters the \"Not-NOW\" trap of legacy logistics thinking:
\nFrom Deterministic to Probabilistic: Instead of relying on fixed lead times, Oway's AI uses probabilistic reasoning to suggest detours and batching based on current truck movements.
\nFrom Monolith to Elastic: As an API-first marketplace, Oway represents the modular architecture required to scale without a costly rip-and-replace cycle.
\nFrom Scribe to Steward: The platform automates documentation (like Bills of Lading) and communication, freeing logistics professionals to focus on strategic trade-offs rather than manual data entry.
\nIn an environment where agility is the only currency, Oway provides a documented ROI:
\nFinancial Viability: Shippers can see cost reductions of up to 50%, while carriers increase annual revenue by up to 30%.
\nSustainability: By filling empty trucks, the platform directly improves the carbon footprint of individual SKUs, a key requirement for 2026 ESG targets.
\nOway fits the Level 4 (Autonomous) status in the N.O.W. Maturity Model because it utilizes Agentic AI for self-correcting workflows and offers holistic value-chain visibility through its rideshare protocol. By adopting Oway, an organization moves from Level 1 (Reactive) firefighting to Level 4 (Autonomous) orchestration.
\nThe old way of shipping relied on finding the best rates, but inadvertently perpetuated systemic inefficiency. Those days are over.
\nFor Carriers: Stop leaving money on the table. Join the Oway ecosystem to access a steady stream of high-intent shippers and fill your empty capacity with high-margin freight.
\nFor Shippers: Stop overpaying for under-optimized shipping. Complement your transportation services with Oway for a faster, cheaper, and greener way to move your products.
\nIn an ever-more complex world of freight, lowering costs while improving quality of service and having a positive impact on the environment isn't just a goal. It's finally within reach.
" }, { "id": "https://www.shipoway.com/research/case-study-euri-lighting/", "url": "https://www.shipoway.com/research/case-study-euri-lighting/", "title": "Euri Lighting", "summary": "How Euri Lighting cut freight costs by up to 53%, saved hours of quoting every week and shipped about 200 loads with zero damage claims.", "date_published": "2026-02-21T12:00:00Z", "authors": [ { "name": "Will Honda" } ], "tags": [ "Case study", "Distribution", "Oway Rideshare", "Fragile freight" ], "content_html": "Euri Lighting, founded in 2013, makes innovative, high-quality lighting, EV chargers and water conservation products. It ships a high volume of freight every day.
\nIn a competitive industry, a single damaged or late shipment can send a customer to a competitor. Euri's daily volume made shipping complex and costly, with frequent damage claims and slow quoting and scheduling.
\n\n\n\"Every month we had at least one or two insurance claims with traditional carriers like FedEx, XPO, Custom, and Central Transport due to damaged shipments.\" Madhara, Euri Lighting
\n
Traditional carriers move freight through warehouse hubs, and every extra touch raises the risk of damage to fragile products made of glass and fine components.
\nEuri moved its freight to Oway Rideshare, which uses machine learning to route shipments into unused space on trucks already on the road. Each truck works like a mini-warehouse: freight is loaded once and goes directly from pickup to delivery.
\nLower freight costs. Shipping costs fell by up to 53% compared with Euri's previous contracted LTL carriers, often saving hundreds of dollars per shipment, with no contracts or subscriptions.
\nHours saved every week. Traditional carriers asked for freight class and dimensions. Oway quotes from four inputs: pickup ZIP, delivery ZIP, number of pallets and total weight. Quotes take seconds, which at Euri's volume adds up to hours each week.
\nZero claims. Across more than 4,000 Oway shipments, including about 200 for Euri, there have been no insurance claims.
\n\n\n\"Since switching to Oway, we haven't had a single claim for damaged or missing products.\" Madhara, Euri Lighting
\n
Same-day delivery. When Euri needed urgent same-day deliveries of close to 100 miles for its biggest customers, Oway delivered within a few hours at about 20% of what other carriers quoted.
\nResponsive service. Support answers in minutes, from the same dedicated team every time. Some of Euri's customers now ask for Oway by name.
\n\n\n\"Oway now runs 100% of our shipments in California.\" Madhara, Euri Lighting
\n
Shippers sign up for free, with no contracts. Get an instant quote or see how Oway OS works.
" }, { "id": "https://www.shipoway.com/research/case-study-vybes/", "url": "https://www.shipoway.com/research/case-study-vybes/", "title": "Vybes", "summary": "How a fast-growing beverage brand cut freight costs by up to 63% across California and Nevada by shipping on trucks already making the trip.", "date_published": "2026-02-21T12:00:00Z", "authors": [ { "name": "Will Honda" } ], "tags": [ "Case study", "Food and beverage", "Distribution", "Oway Rideshare" ], "content_html": "Vybes is a fast-growing beverage company known for its magnesium-based energy drinks, a healthier option for people looking for an afternoon boost. Its drinks are sold widely in grocery and convenience stores across California.
\nAs Vybes grew, high freight costs and unreliable shipping started to threaten that growth. Getting product to distributors meant paying national carrier rates and absorbing surprise fees for liftgates and freight reclassifications.
\n\n\n\"One of the biggest cost drivers in my business is actually getting my product to the distributor. Because I'm shipping all over the country, it can be very expensive.\" Jonathan, CEO, Vybes LA
\n
First-mile shippers like distributors and manufacturers carry the heaviest freight costs in a supply chain, and traditional LTL pricing reflects that.
\nVybes moved its pallets to Oway Rideshare. Oway uses machine learning to find unused space on trucks already driving the same direction, so freight rides straight from pickup to delivery with no warehouse transfers. Oway absorbs the accessorial fees that usually show up after the fact, and prices are up to 50% below the large national carriers.
\nLower freight costs. Vybes saved as much as 63% on shipments.
\n\n\n\"You guys have kind of come in and found this really great niche for Vybes. When I first started working with you guys, I wasn't sure if that meant the quality would suffer… But overall, it's been just as good as Echo [our traditional carrier], and you're actually saving me money.\" Jonathan, CEO, Vybes LA
\n
Faster, more dependable delivery. Skipping warehouses removed the handoffs where shipments usually slow down.
\nEasy to run. Ordering and tracking fit into the team's day without new work.
\n\n\n\"Your software is super easy to use. It's very intuitive… And it reminds the customer you're always saving them money, which is really important.\" Jonathan, CEO, Vybes LA
\n
Oway's mid-mile and long-distance lanes across Southern California and Nevada have helped Vybes strengthen its regional position while it builds toward national expansion.
\nShippers sign up for free, with no contracts. Get an instant quote or see how Oway OS works.
" }, { "id": "https://www.shipoway.com/research/rtd-delivery-systems/", "url": "https://www.shipoway.com/research/rtd-delivery-systems/", "title": "RTD Delivery Systems", "summary": "How a regional carrier filled the empty space on its existing lanes and earned $1,520 in extra revenue from five shipments in its first week on Oway.", "date_published": "2026-02-21T12:00:00Z", "authors": [ { "name": "Will Honda" } ], "tags": [ "Carriers", "Case study", "Oway Rideshare" ], "content_html": "RTD Delivery Systems has been a trusted regional carrier since 2005. As the freight market tightened, broker rates stalled, capacity went unused and operating pressure grew. RTD wanted more profit without adding work.
\nRTD worked with brokers whose rates had not kept pace with the market. Its most popular lanes ran at 75 to 80% capacity, and some routes ran half full. Filling that space meant cold outreach and manual broker coordination, which pulled people away from growing the business.
\nOway fills unused space on routes a carrier already runs, with no detours and no extra effort. For RTD, that meant loads matched to its existing routes and fully automated booking: click the email, review the load and accept it in seconds. No calls, no negotiation and no surprises at pickup or delivery.
\nIn its first week, RTD earned an additional $1,520 from five shipments, all in space that would otherwise have driven empty. Loads were secured in seconds.
\n95% of pickups and drop-offs were completed in under 10 minutes, and every shipment went smoothly. RTD called Oway faster and easier than any platform it had used before.
\n\n\n\"It's crazy out there. We don't know how some carriers are surviving. Oway helped us fill empty space and earn extra income with virtually no extra effort. Just click and go.\" RTD Delivery Systems Inc.
\n
RTD wanted a better way to fill the space it was already driving, without a new dispatch team or a complicated tech stack. Oway delivered profit from underused capacity and automated booking with no back-and-forth. In a market defined by volatility and thin margins, RTD now runs leaner, moves faster and earns more without changing how it operates.
\nCarriers join free, with no minimums. See Oway for Carriers.
" } ] }