Buy a real-time transportation visibility platform when most of your freight moves on truckload carriers you cannot integrate one by one, or when customers expect live location data you have no other way to get. Extend the TMS or ERP you already run when a handful of parcel and LTL carriers already send status by API or EDI 214 and the gap is screens and alerts. Build your own tracking layer when the hard part is joining carrier events to purchase orders, ship notices, dock appointments and receipts, and writing the result back to ERP and WMS.
Those three answers are not exclusive. You can combine a platform for carrier connectivity with your own layer for milestones, rules and write-back. The rest of this field note gives the criteria, the data sources by mode, a worked example and an evaluation plan you can run in about eight weeks.
Buy, build or extend: what each option means
Each option buys a different thing, so define them before you compare prices. The words “visibility” and “tracking” hide three distinct jobs: collecting carrier events, turning them into milestones and exceptions, and putting the results where people work.
| Option | What you get | What you still own |
|---|---|---|
| Buy a visibility platform | A vendor-run network of carrier connections (APIs, EDI, telematics and driver apps), predicted ETAs, a tracking screen and an outbound data feed | Carrier onboarding follow-up, the link between platform shipments and your orders, and any write-back into ERP or WMS |
| Extend the TMS or ERP | A tracking module, carrier status interfaces or add-on screens from the vendor whose system already holds your loads | The gaps the module does not cover, usually truckload location data and inbound supplier freight |
| Build a tracking layer | Software you specify: your milestone model, exception rules, joins to POs and ASNs, and write-back | Every carrier interface you choose to build, plus its upkeep when carriers change their APIs |
Disclosure first. GenFrontier is a Kansas City firm that builds shipment-tracking and inventory software around clients’ ERP, WMS and TMS, so we have an interest in the build option. We resell no software and have no vendor partnerships, so we have no stake in which platform or TMS module you pick. Where buying wins, this note says so.
A useful test is to ask which of the three jobs is actually failing today. If you cannot get events from carriers at all, that is a connectivity problem, and building your own connections to dozens of small truckload fleets is rarely the best use of an integration team. If events arrive but nobody acts on them, that is a rules and ownership problem, and no feed will fix it. If people act on events but the ERP still shows the wrong date, that is a write-back problem.
What a real-time transportation visibility platform does
A real-time transportation visibility platform is hosted software that collects shipment status and location from many carriers, normalizes it, predicts arrival times and shares the result with shippers and their customers. Gartner tracks the category under that name. Its Peer Insights market definition describes platforms that get data through API or EDI integration with carrier systems, through telematics feeds such as in-cab or trailer devices, and through mobile devices. The same definition says the market mostly addresses domestic road freight. (The live page blocks automated readers, so we read that wording from an archived 2022 copy.)
What the platform holds is the carrier side of the shipment: the load or tracking number, the carrier, stops, pings, status events and an ETA. What it does not hold, unless you send it, is your business context. It does not know which purchase order lines are on the trailer or which dock appointment the load is booked into. It does not know whether a late arrival matters, or which customer order is waiting.
What the platform cannot fix
Three problems stay with you whichever option you choose:
- Reference data. A tracking event is only useful if it lands on the right shipment. That needs a shared key on the tender, the bill of lading and the carrier’s record: a load number, a BOL number, a PRO number or a PO number. If your shipping team keys the PO into the wrong field, no platform will match it.
- Ownership. An alert that goes to a shared inbox is a notification, not an exception. Someone has to own each exception type, with a response time and a closing reason. Our note on shipment exception management for B2B freight covers that catalogue.
- Write-back. The receiving appointment in your WMS, the promised date on the customer order in your ERP and the in-transit quantity on a PO do not update themselves. Someone has to specify and build each update, and decide which system wins when two disagree.
These three items are where most of the work sits in any of the options. Price them in.
What a predicted ETA can and cannot know
Platforms sell predicted ETAs, and they are useful, but every ETA is bounded by the driver’s hours. Under 49 CFR 395.3, a property-carrying driver may drive 11 hours within a 14-hour window that starts after 10 consecutive hours off duty. The driver also needs a 30-minute break after 8 hours of driving.
So an ETA can jump by ten hours or more in one update, when a driver who was “two hours out” runs out of available hours. A good ETA model knows the driver’s remaining hours; one built only on distance and speed does not. Ask each vendor whether its ETA uses hours-of-service status, and how it shows the reason for a jump.
Keep three dates apart in your data model: the carrier’s estimate, the platform’s prediction and the confirmed appointment at the receiving dock. They answer different questions, and mixing them in one field makes it impossible to tell later which one was wrong.
Where tracking data comes from, mode by mode
The data you can get depends far more on the mode and the carrier than on the software you buy. Before comparing vendors, list where each kind of event can physically come from.
Electronic logging devices and telematics
Under 49 CFR 395.8, motor carriers subject to the hours-of-service rules must have their drivers record duty status on electronic logging devices (ELDs), with narrow exceptions such as trucks built before model year 2000. The regulation’s date for installing them is December 18, 2017. Those devices record location, but read what the rule requires before treating that data as a tracking feed. Under 49 CFR 395.26, an ELD records date, time and location at each duty status change. While the truck is moving, it adds an intermediate record when no other record has been made in the previous hour.
The technical appendix to the same subpart puts reported position accuracy on the order of plus or minus 1 mile. During authorized personal use, location is recorded at a coarser resolution of roughly a 10-mile radius. The rule’s data transfer duty runs to authorized safety officials on request (49 CFR 395.24(d)). Subpart B does not mention shippers or brokers at all. A shipper gets that location data only with the carrier’s consent, typically when the carrier connects its ELD or telematics account to a platform.
That is why truckload visibility is mostly a network question. A platform whose network already includes your carriers, or that gets their drivers to run a tracking app, can show you a ping trail. Building that consent network yourself, carrier by carrier, is slow.
Carrier APIs
Parcel and LTL carriers such as FedEx and Old Dominion publish tracking APIs. These return status events keyed to a tracking or PRO number, and they come with usage rules. The FedEx Track API documentation caps a single track transaction at 30 packages. It also asks developers to limit tracking calls to what the business needs. Old Dominion’s Shipping API Integrations page lists a REST tracking API that “provides data for a single shipment”, and says all of its SOAP endpoints retire on March 1, 2027.
Those two details matter for the build option. Polling hundreds of open shipments every few minutes, within each carrier’s rules, is real engineering. Carrier platform changes arrive on the carrier’s schedule, not yours, and each one is a project for whoever owns the link.
EDI 214 status messages
The X12 214 is the long-standing EDI route. X12 describes it as a message “used by a transportation carrier to provide shippers, consignees, and their agents with the status of shipments in terms of dates, times, locations, route, identifying numbers, and conveyance” (X12, 214). Each status sits in an AT7 segment. That segment carries a status code, an optional reason code, appointment status and reason codes, and a date, time and time zone code.
In the 004010 release, as published in Stedi’s X12 reference (a secondary source), status code AF means “Carrier Departed Pick-up Location with Shipment” and D1 means “Completed Unloading at Delivery Location”. Reason code NS means “Normal Status”. Carriers differ in which codes they send and how promptly, so a 214 feed from one carrier is not interchangeable with another’s. The choice between 214, APIs and telematics, and how to normalize them, is the subject of our note on EDI 214 versus carrier APIs.
LTL terminal scans and the NMFTA standards
LTL freight moves through terminals, so its events are scans at pickup, at origin and destination terminals and at delivery. There may be long quiet stretches between them; our note on why LTL tracking goes quiet between terminals explains why. The NMFTA’s Digital Standards Development Council publishes common LTL API standards, including an electronic bill of lading (eBOL) standard at version 2.1 and a Pickup Request and Visibility standard. Its API page also lists an In-Transit Visibility standard at version 1.0.0, while the council’s LTL page still labels that one “In Development” (both checked November 2026).
| Mode | Typical sources | What you usually get | Who controls access |
|---|---|---|---|
| Parcel | Carrier tracking API, carrier email or web notices | Scan events per package, carrier estimated delivery date | Carrier developer terms and your account |
| LTL | Carrier tracking API, EDI 214, eBOL response with PRO number | Pickup, terminal arrivals and departures, out for delivery, delivered, appointment events | Carrier, through your account or EDI setup |
| Truckload, asset carrier | EDI 214, carrier API, ELD or telematics shared with consent | Status events plus location pings, if the carrier shares them | Carrier consent, often through a platform |
| Truckload, small fleet or broker | Driver tracking app, broker portal or API, phone and email check calls | Pings only while the app runs; otherwise manual check calls | Driver and broker |
| Inbound from suppliers | Supplier's carrier data, EDI 856 ship notice, supplier portal | Ship notice and whatever the supplier's carrier shares | Supplier, under your routing guide |
Six questions that decide the choice
Six questions do most of the deciding. Answer them with your own shipment data, not a vendor’s demo, and the right option is usually clear.
1. What is your mode and carrier mix?
Pull 12 months of loads from your TMS or freight bills. Count loads by mode and by carrier, then see how many carriers cover most of your volume. A shipper with three national LTL carriers and two parcel carriers can integrate them directly. A shipper using 150 truckload carriers, most of them small fleets, cannot do so in any reasonable time.
2. Who consumes the data, and how fresh must it be?
List every consumer: the transportation desk, the dock scheduler, customer service, buyers and customers themselves. For each, write the decision the data supports and how late it can arrive before that decision fails. A dock scheduler rebooking doors needs an ETA that is hours old at most. A buyer checking whether a supplier shipped needs a daily answer.
3. Where must tracking results be written back?
Write down each update and its target system: the WMS receiving appointment, the ERP order promise date, the in-transit quantity on the PO, the proof-of-delivery image on the invoice record. If the list is short and lands in the TMS, extending the TMS is attractive. If it is long and spans ERP, WMS and customer channels, someone has to build those interfaces whichever option you buy. The ERP and WMS integration work is often larger than the tracking feed itself.
4. Who carries the carrier onboarding burden?
Every option needs carriers to connect. A platform vendor may handle the outreach, but the carrier still has to consent, and some never will. Ask each vendor who chases a non-responsive carrier, how that is reported, and what happens to loads tendered to carriers who are not connected. For the build option, onboarding means credentials, test shipments and a code mapping for every carrier.
5. Can you afford the upkeep of point-to-point links?
Every direct carrier link is a small product with its own maintenance schedule. Carriers retire endpoints, change authentication and adjust rate limits. Count the links, then estimate the yearly hours to keep each one working. Budget that before choosing to build.
6. Who owns the rules and the raw events?
Find out whether you can export every raw event with both timestamps: when it happened and when it arrived. Check whether you can change exception thresholds yourself and see why an ETA moved. If the answer is no, you are renting conclusions, not data, and leaving later will be expensive.
| Question | Points to buy a platform | Points to extend TMS or ERP | Points to build a layer |
|---|---|---|---|
| Mode and carrier mix | Many truckload carriers, many small fleets | Few carriers, mostly parcel and LTL with APIs or 214 | Mixed modes plus heavy inbound supplier freight |
| Consumers | Customers want live maps and ETAs | Mostly the transportation desk | Dock, buyers, customer service and customers each need different views |
| Write-back | Little: people read the platform screen | Into the TMS only | Into ERP, WMS and customer channels |
| Carrier onboarding | You cannot staff carrier outreach | The TMS vendor already supports your carriers | Carriers are few and stable |
| Upkeep capacity | No integration team | The TMS vendor maintains the links | An integration team, or a partner, with time for upkeep |
| Rules and data | Platform defaults fit your exceptions | Module rules fit, with small changes | Rules depend on your POs, appointments and customer terms |
When buying a platform is the better choice
Buy when the problem is carrier connectivity at scale and you do not want to run an integration program to solve it. Three situations make that case clearly.
Truckload with a long tail of carriers. If much of your freight moves on small fleets or through brokers, location data depends on ELD consent or a driver app. A platform with an existing carrier network and an onboarding team reaches those carriers faster than you can. No in-house build changes that.
Customers who want live tracking now. If a large customer requires a shared tracking link or a live data feed as a condition of business, a platform can usually provide one sooner than a custom build. That is a sound reason to buy even if you later add your own layer.
No integration team. A tracking layer you build has to be maintained. If you have no developers who know EDI and carrier APIs and no budget for a partner to support the software, buying a hosted service is the responsible choice.
Buying still leaves work. You will map your load and order references into the platform, decide who acts on its alerts, and build any write-back the platform does not provide. Ask for a written list of the carriers already connected that match your own carrier list, and ask how many of your loads they cover. That number matters more than the total network size in the sales deck.
Questions to put to every platform vendor
Before you sign, get written answers to these:
- Which of our top carriers are connected today, and by what method: API, EDI, ELD, telematics or driver app?
- Who contacts carriers that are not connected, and how is progress reported to us?
- Can we export all raw events with event time and received time, through an API, at no extra charge?
- Can we set our own exception thresholds per lane, customer or appointment type?
- How does the platform receive our PO numbers, order numbers and dock appointments?
- What does the platform send back, and in what format: webhook, API, EDI 214 or file?
- What happens to our data and history if we leave?
When extending your TMS or ERP is enough
Extend the system you already run when your carriers are few, already integrated with it, and the missing piece is visibility on screen rather than new data. It starts from software you already license and support, and the vendor maintains the carrier links.
A TMS that tenders loads by EDI uses the 204 Motor Carrier Load Tender, receives the 990 Response to a Load Tender, and takes status back in the 214 (X12 transaction set list). If those messages already flow, the data is there. The work is configuring milestones, alerts and views, and adding a customer-facing status page if needed.
Extension falls short in three common places:
- Truckload location. A TMS that receives 214 status messages may still lack position data between pickup and delivery.
- Inbound supplier freight. Loads arranged by suppliers often never enter your TMS. Buyers learn about them from emails and the ship notice, if one is sent.
- Cross-system rules. A TMS knows the load but often not the receiving appointment in the WMS or the customer’s must-arrive-by date in the ERP. Exceptions that need both stay manual.
Before extending, ask the vendor for the list of carriers and message types supported in your version, and for the cost and timeline of each missing one. Then compare that list with the carrier counts from your own data.
Questions for your TMS or ERP vendor
Put these to the vendor in writing before you commit to extension:
- Which carriers and message types (204, 990, 214, carrier APIs) does our version support today, and which need custom work?
- Can the module receive location pings, or only status events?
- Can it hold inbound loads that suppliers arrange, without a tender from us?
- Can exception rules read data from our WMS appointment schedule or ERP orders?
- What does an upgrade do to any custom screens or rules we add?
When building your own tracking layer makes sense
Build when your competitive problem is not “where is the truck” but “what does this delay mean for this order, this dock and this customer”. A tracking layer you own sits between event sources and your business systems. It is the place where carrier events meet POs, ship notices, appointments and receipts.
“Build” rarely means building a carrier network. In practice it means four components:
- A milestone model. One agreed list of milestones per shipment type, each with the event that proves it, its source and its owner. You can draft one with our tracking milestone map.
- Source interfaces. Direct links to the carriers that carry most of your volume, plus a feed from a visibility platform if you subscribe to one. A platform becomes one source among several.
- Rules. ETA logic and exception rules written in plain language, with thresholds your team can change.
- Write-back. Interfaces that update the WMS appointment, the ERP order and the customer’s view, with a defined winner when two sources disagree.
The build option wins when inbound and outbound freight both matter, when rules depend on data only you hold, and when several internal teams need different views of the same shipment. It also wins when you want to avoid locking your history into one vendor’s format. Our shipment tracking software page describes what such a layer contains.
It loses when carrier connectivity is the main gap, when you have no one to maintain it, or when the team cannot agree on milestones and owners. Software built before those agreements exist encodes the argument instead of settling it.
The people a build needs
Software you own needs owners. Plan for three roles, even if one person covers two of them in a smaller company:
- A product owner in transportation or customer service, who decides milestones, thresholds and who sees what, and who signs off changes.
- An integration developer or partner, who maintains carrier links, mappings and write-back interfaces, and handles carrier API changes.
- A support contact on each shift, who works the unmatched-events queue and knows how to escalate a broken feed.
If none of these roles can be filled, the build option is not available yet, whatever its other merits. Documentation and training are part of the deliverable for the same reason: the layer has to outlast the people who specified it.
The hybrid option
A hybrid combines the three. A platform supplies truckload location and ETAs, direct APIs or 214s cover the main LTL and parcel carriers, and your own layer joins everything to orders and writes back. This is not indecision. Each piece does the job it is best placed to do.
If you choose the hybrid, decide early which component owns the ETA shown to customers. Two ETAs for the same load, one from the platform and one from your rules, will confuse customers and staff alike. Write the rule down: for example, the platform ETA until the carrier confirms a delivery appointment, then the appointment time.
How tracking results get back into ERP and WMS
Write-back is the part of a visibility project that changes what people see in the systems they already use. Each update needs a target field, a trigger and a rule for what happens when two sources disagree. Without those three, the tracking screen and the ERP drift apart and people stop trusting both.
| Update | Target system | Typical trigger | Conflict rule to decide |
|---|---|---|---|
| Expected arrival on the receiving appointment | WMS or dock scheduling | New ETA differs from the appointment by more than your threshold | Does a carrier-confirmed appointment outrank a predicted ETA? |
| In-transit quantity on the PO line | ERP | Ship notice received, or pickup event on an inbound load | Ship notice quantity until receipt, then the receipt |
| Promise date on the customer order | ERP or order management | ETA moves past the promised date | Who approves the change before the customer is told? |
| Delivered status and proof of delivery | ERP order and invoice record | Delivered event, delivery receipt image | Can a delivered status ever revert without manual review? |
| Carrier on-time record | TMS or reporting database | Delivered event | Your dock or gate time, when present, beats the carrier’s |
Matching events to the right shipment
Every event must land on the right record, and carriers do not all key shipments the same way. XPO’s LTL shipment tracking API guide (release 1.0, 2018) treats the PRO number as the first part of a shipment’s unique identifier and the pickup date as the second. It accepts either a PRO or a customer reference number as the lookup. Old Dominion’s trace page offers logged-in searches by PRO, BOL, PO, load and pickup numbers.
The practical rule is to store every reference you send and receive, and to match in a fixed order. For example: carrier PRO plus pickup date first, then your BOL number, then the PO. Log every event that matches nothing, and review that queue daily, because unmatched events often point to a data entry problem upstream.
Timestamps, duplicates and late arrivals
Store two times for each event: when it happened and when you received it. The XPO guide states that all its response timestamps are in UTC. In a 214, the AT7 segment carries a time code that qualifies the time, so check which zone each carrier uses before you compare a carrier’s delivered time with your dock scan.
Events also arrive twice, late and out of order. Never let an older event overwrite a later milestone; flag the conflict instead. Keep the raw message, because the day a customer disputes a delivery time is the day you need it.
What each option costs to own
Compare the options on what they cost to own over several years, not on the first year’s license or build quote. The cost categories differ more than the totals.
| Cost category | Buy | Extend | Build |
|---|---|---|---|
| License or subscription | Recurring; ask how it scales with loads, carriers and users | Module fee on the existing contract | None for software you own; hosting costs |
| Initial integration | Reference mapping and write-back interfaces | Configuration and any missing carrier links | Every interface, the rules engine and the screens |
| Carrier onboarding | Vendor outreach; your follow-up on stragglers | Vendor-supported carriers only | Credentials, testing and code mapping per carrier |
| Ongoing upkeep | Vendor maintains carrier links; you maintain write-back | Vendor maintains; upgrades may break changes | You maintain everything, including carrier API changes |
| Exit cost | Data export and re-onboarding carriers elsewhere | Tied to the TMS or ERP contract | Low, if code and documentation were delivered |
| Staff time | Exception handling and vendor management | Configuration and exception handling | Product ownership, exception handling and support |
Two items are easy to underestimate in every option. The first is exception handling: better visibility creates more alerts, and someone has to work them. The second is reference data cleanup: wrong PO numbers on bills of lading and missing appointment IDs break every option equally.
Example (fictional). A distributor ships about 1,500 outbound loads a month from two warehouses: 900 LTL loads on four national carriers, 450 truckload loads on 70 carriers (most of them small fleets) and 150 parcel shipments on two carriers. It also receives about 600 inbound supplier loads a month, most arranged by suppliers.
The team scores the six questions. Truckload location points to buying, because 70 carriers is too many to integrate directly. LTL and parcel point to direct APIs, because six carriers cover all that volume. Inbound freight and the write-back list (WMS appointments, ERP promise dates, PO in-transit quantities) point to building.
They choose a hybrid: a platform subscription for truckload only, direct links to the four LTL and two parcel carriers, and a small tracking layer that joins everything to orders and appointments. The build is scoped in writing first, so the platform contract can require a raw-event API that the layer can read.
How to run the evaluation in eight weeks
The plan below fits in about eight weeks and answers the question with your own data rather than a vendor’s demo data.
- Weeks 1 and 2: measure your freight. Count loads by mode, carrier, lane and inbound or outbound for the last 12 months. Find the carriers that cover most of your volume. List the data each one already sends you today, using a sheet such as the integration inventory worksheet.
- Weeks 2 and 3: write the milestone model and the consumer list. Agree on milestones, the event that proves each one, owners and exception thresholds. List each consumer and how fresh its data must be. This becomes the requirement every option is tested against.
- Week 3: write the write-back list. For each update, name the source, target system, field and rule for conflicts. Our requirements analysis approach starts here because it decides how much integration any option needs.
- Weeks 4 to 7: pilot with real loads. Ask one or two platforms, your TMS vendor, or both, to track a sample of live loads across your main modes for at least three weeks. If you are considering a build, pull the same period’s 214s and API responses from your top carriers.
- Weeks 4 to 7: measure three numbers per source. Measure coverage, latency and accuracy, defined in the next section.
- Week 8: score and decide. Apply the six questions, add the cost categories and pick an option or hybrid. Write the reasons down, so the decision can be revisited when the carrier mix changes.
Three measurements that settle arguments
Measure each option on the same loads, with these definitions:
- Coverage: the share of pilot loads with each milestone present, such as pickup, at least one in-transit update, arrival and delivery.
- Latency: for each event, the received time minus the event time. Look at the slowest tenth of events, not just the average, because late events cause the missed appointments.
- Accuracy: for delivered loads, compare the source’s delivered time with your own dock or gate record. For ETAs, compare the ETA held a set number of hours before arrival with the actual arrival.
Example (fictional). A pilot tracks 400 outbound truckload loads over four weeks through one platform. Pickup events exist for 388 loads and delivered events for 371, so delivered coverage is 371 of 400, or about 93 percent. At least one in-transit ping exists for 352 loads.
Latency looks fine on average, but the slowest tenth of delivered events arrived more than six hours after delivery. All of those came from nine carriers whose drivers confirm delivery in an app at the end of the day. The team adds those nine carriers to a list for contract follow-up and, for the rest, compares delivered times with the receiving confirmations its customers send.
Expect the numbers to differ by carrier as much as by vendor. That is useful: it tells you which carriers to fix with direct links or contract terms, whichever software you choose.
Mistakes that sink visibility projects
Each of these errors can undo any of the three options. Check your plan against them:
- Buying before agreeing on milestones. If transportation and customer service disagree on what “shipped” means, a new screen will show the disagreement in color.
- Judging network size instead of your coverage. A platform’s total carrier count says little. The share of your own loads it tracked in the pilot says a lot.
- Leaving inbound freight out. Supplier-arranged loads are the ones least likely to be in your TMS, and they are the ones receiving has to plan doors and labor around.
- Showing two ETAs. Pick one owner for the customer-facing ETA, and show the source next to it.
- Alerts without owners. An alert nobody owns trains people to ignore alerts. Assign each exception type to a role with a response time before go-live.
- Discarding raw events. Normalized statuses lose detail. Keep the original messages for disputes, audits and later rule changes.
What to do next
Start with your own data, not with demos. A one-page answer to each of the six questions will narrow the field faster than any vendor comparison chart.
- Pull 12 months of loads by mode and carrier, and identify the carriers that carry most of your volume.
- Draft your milestones, owners and exception thresholds in the tracking milestone map.
- List every write-back: system, field, trigger and conflict rule.
- Ask each candidate vendor the platform questions above, in writing.
- Run a pilot on live loads and measure coverage, latency and accuracy per carrier.
- Decide which component owns the customer-facing ETA before go-live.
If the evaluation points to building some or all of the tracking layer, or to integration work around a platform you buy, the first deliverable is the written milestone model and write-back list. You can also describe your carriers and systems to us and we will tell you which parts we would scope first.
