Take parcel status from each carrier’s tracking API. Take LTL status from the EDI 214 or the carrier’s API, whichever that carrier supports better, and add the other only to fill a gap you can name. Take truckload status from ELD or telematics location shared by the carrier, with a 214 or API milestone feed underneath it. Then map every feed into one status model you own, with your own event codes, reason codes, UTC timestamps, a source on every event, and a timer that turns silence into an event of its own.
The mode decides where the data comes from. The model decides whether anyone downstream can use it. This field note covers both halves: what each feed carries according to the published guides, a decision table by mode, a design for normalizing 214 status codes and API statuses, fallback rules for quiet feeds, and the upkeep each connection needs after go-live.
Two neighboring questions have their own field notes. Why LTL updates stall for a day or more is covered in why LTL tracking goes quiet. The UPS, FedEx and USPS platform changes are in carrier API migrations. Whether to buy a visibility platform at all is a separate decision, and this note assumes you are wiring carrier feeds into systems you already run.
Three kinds of status feed, defined
There are three ways carrier status reaches your systems: an X12 214 message, a carrier tracking API, and location data from the truck. Each one answers a different question, so the first design step is knowing which question each feed can answer.
- EDI 214. X12 names transaction set 214 the Transportation Carrier Shipment Status Message. Its published purpose is to let a carrier give “shippers, consignees, and their agents” the status of shipments “in terms of dates, times, locations, route, identifying numbers, and conveyance.” It is a batch document. The carrier decides when to generate it and sends it through a VAN, AS2 or SFTP.
- Carrier tracking API. A web service the carrier publishes. You call it with a PRO, tracking number or reference, and it returns the current status and often an event history. Some carriers also push events to an endpoint you host (a webhook).
- ELD and telematics location. Position data from the device in the tractor. It tells you where the truck is. It does not, by itself, tell you that a milestone happened, such as freight loaded or delivery signed.
A milestone is a business event in the life of a shipment: picked up, departed terminal, out for delivery, delivered. A status feed is only useful if it reports milestones you can act on, with a time and place you can trust.
What an EDI 214 carries, segment by segment
An EDI 214 is a structured list of status events for one shipment, keyed by the carrier’s PRO number and the carrier’s SCAC. The details you need to map it are in each carrier’s implementation guide, not in the standard alone, because every carrier sends a subset.
Two published LTL guides show the usual shape. The ArcBest 214 specification and the Estes Express Lines 214 guide, version 4010 both use this sequence:
- ST: transaction set header, with 214 as the identifier.
- B10: the beginning segment. Estes’s guide says B1001 “is the carrier’s PRO (invoice number) that identifies the shipment.” B1002 is the shipper’s shipment identification number and B1003 is the SCAC, which the guide marks mandatory.
- L11: reference numbers with a qualifier, such as a purchase order number.
- N1, N3, N4: names and addresses of parties such as shipper, consignee and bill-to.
- LX: an assigned number that opens a status loop.
- AT7: the shipment status details. This is the segment that matters most.
- MS1 and MS2: where the event happened (city, state, country) and the equipment owner and number. ArcBest sends these; Estes marks them “Not Used.”
- AT8: weight and handling-unit quantities, used by Estes.
- SE: trailer with a segment count.
The AT7 segment is where status lives
The AT7 segment carries a status code, a reason code, an appointment code, a date, a time and a time code. Estes’s guide lists its syntax rules, and they matter for parsing:
- Only one of AT701 (shipment status code) or AT703 (appointment status code) may be present.
- If AT701 is present, AT702 (the reason code) is required.
- If AT703 is present, AT704 (the appointment reason) is required.
- If AT707 (time code) is not present, “AT706 represents local time of the status.”
That last rule is the one most mappings get wrong. A 214 event time is local time at the event location unless the carrier sends a time code. If you store it as if it were your server’s time zone, every cross-country shipment gets events in the wrong order.
ArcBest’s guide includes a sample status line: AT7*AF*NS***20230627*1058*LT~. Read left to right, that is status AF, reason NS, no appointment fields, a date, a time and a time code. Estes’s guide defines NS as “Normal Status.” The same ArcBest sample carries MS2*SCAC*UNKNOWN~, and the guide notes the equipment number is sent as “UNKNOWN” when the data is not available. Plan for placeholder values like that in your parser.
Carriers send a subset of the code list
Each guide publishes only the codes that carrier sends. ArcBest calls its list “base status code offerings” and says more codes “are available upon request for most service lines.” Here are the codes and meanings exactly as the two guides publish them:
| Code | Element | Meaning as published | Listed by |
|---|---|---|---|
| X3 | AT701 | Arrived at pick-up location | ArcBest |
| AF | AT701 | Carrier departed pick-up location with shipment | ArcBest, Estes |
| X6 | AT701 | En-route to delivery location | ArcBest |
| X1 | AT701 | Arrived at delivery location | ArcBest, Estes |
| D1 | AT701 | Completed unloading at delivery location | ArcBest, Estes |
| AG | AT701 | Estimated delivery | Estes |
| AI | AT701 | Shipment has been reconsigned | Estes |
| AV | AT701 | Available for delivery | Estes |
| SD | AT701 | Shipment delayed | Estes |
| NS | AT702 | Normal status | Estes (ArcBest sample uses it) |
| BS | AT702 | Refused by customer | Estes |
| AA | AT703 | Pick up appointment date and/or time | ArcBest |
| AB | AT703 | Delivery appointment date and/or time | ArcBest, Estes |
| X9 | AT703 | Delivery appointment secured on this date and/or time | Estes |
| NA | AT704 | Normal appointment | Estes |
Two things follow from this table. First, the two carriers overlap on only three status codes, so a mapping built from one carrier’s guide will drop events from the next. Second, some codes are not events at all. AG is an estimate, and AB or X9 describe an appointment, so they belong in different fields of your model than “something happened.”
ArcBest’s guide shows one more trap: the company uses different SCACs by business line (ABFS for LTL, and others for truckload, managed transportation, dedicated, expedite and intermodal). The guide lists eight SCACs across those lines, so if your carrier master keys on SCAC, one carrier can appear as eight. Map SCACs to a carrier entity, and keep the SCAC on each event.
What the 214 does well, and where it is thin
The 214 is strongest where you already run EDI. It rides the same VAN or AS2 connection as your 204 load tenders and 210 freight invoices, your translator already handles envelopes and the 997 Functional Acknowledgment, and LTL carriers publish 214 guides, such as the two cited here. Estes lists “SFTP, AS2, or VAN” as communication options.
It is thin in three places:
- Timing. Events arrive when the carrier’s system batches and sends them, and when your translator picks them up. Neither guide states a send frequency, so ask each carrier and write the answer into your partner profile.
- Coverage. You get the codes in that carrier’s guide and nothing more. Exceptions often arrive as a reason code on a normal status, or not at all.
- Location. MS1 gives a city and state, not a coordinate. That is enough for “at the Little Rock terminal” and not enough for “two miles from our dock.”
What carrier tracking APIs return
A carrier tracking API returns the carrier’s own view of the shipment on request: current status, event history, references, and often estimates and exceptions in more detail than a 214. The cost is that every carrier’s API is different, and you build and maintain each connection separately.
An LTL example: XPO’s published tracking API
XPO’s Shipment Tracking API implementation guide (release 1.0, dated November 2018; check the carrier’s developer site for the current version) describes three operations:
- getShipmentStatus retrieves status “based on PRO and/or reference numbers.”
- listShipmentReferenceNumbers returns the references tied to a PRO.
- listShipmentTrackingEvents returns the shipment history, and the guide says to use it “when investigating a possible exception to a shipment.”
Several details in that guide change how you design the intake:
- “All response timestamps are in UTC.” That is the opposite default from a 214, where an AT706 time without a time code is local.
- Status is a numeric code with a description. The appendix lists codes such as 1 (Picked up), 8 (Out for delivery), 19 (Attempted delivery), 20 (Delivered part short), 23 (Delivered by) and 33 (Possible Delay Notification).
- Exceptions are their own object, with a type, a reason, a resolution code and a count of over, short and damaged pieces.
- The sample event history includes entries such as “PRO Inquiry” and “Record updated.” Those are system events, not freight movement. If you load the history without a filter, your timeline fills with noise.
A parcel example: FedEx Track API business rules
Parcel carriers publish tracking APIs built for high volume, and their business rules tell you how to call them. The FedEx Track API documentation states:
- Limit the number of tracking numbers in a single track request to 30, which the page says reduces reply size and response time.
- “Limit the number of times a package/order is tracked to what is necessary for business needs.” For batch tracking, it says to remove packages that no longer need tracking from the batch.
- It “takes up to 24 hours for a new package/order to show up in the system.”
- “Tracking information is available for 90 days after delivery” for FedEx Express, FedEx Ground and FedEx Custom Critical.
- A delay detail on scan events returns ON_TIME, EARLY or DELAYED.
Those rules translate directly into design: batch in groups no larger than the limit, stop polling delivered packages, do not raise “no tracking” alerts in the first day, and copy any tracking history you need for claims or audits into your own store before the carrier’s retention window closes.
Parcel carriers have also been changing authentication and retiring older interfaces. The dated status is in carrier API migrations, so this note does not repeat it.
Poll or push
A poll is a call you make on a schedule. A push is a call the carrier makes to you when something happens, usually a webhook with a signed payload. Polling is simple to build and wasteful at scale; pushing is efficient and adds an endpoint you must secure, monitor and keep available.
Example (fictional). A distributor polls an LTL carrier’s API for 1,200 open shipments every 30 minutes. That is 2,400 calls an hour and 57,600 a day. Most shipments change status a handful of times between pickup and delivery, so nearly every call returns nothing new. The team changes the schedule by milestone: every 4 hours while a shipment is at or between terminals, every 30 minutes once it is out for delivery, and never after delivered. On a typical day 1,000 shipments are at or between terminals (6 calls each), 150 are out for delivery (48 calls each) and 50 are delivered (none). That is 13,200 calls a day, under a quarter of the original, and the out-for-delivery shipments, the ones customers ask about, are checked as often as before.
Shared API standards for LTL and truckload
The NMFTA’s Digital Standards Development Council (DSDC) publishes API standards so that shippers do not have to build a different interface for every carrier. Where your carriers implement them, they reduce the per-carrier mapping work; where they do not, they are still a good template for your own model.
The DSDC API standards page lists, for LTL, an eBOL standard at version 2.1 and Pickup Request and Visibility, Preliminary Freight Charges and In-Transit Visibility standards at version 1.0.0. For full truckload it lists Appointment Scheduling, eBOL, Invoice and Documentation, In-Transit Visibility, Rate/Quote and Book and Tender.
The LTL In-Transit Visibility standard is the one that competes with the 214. As described on DSDC’s page:
- Updates are “delivered via push notification where possible,” and parties can also pull the latest status by PRO, BOL, PO or account number.
- Milestones include Picked-up, Arrived at Terminal, Unloaded at Terminal, Loaded at Terminal, Departed Terminal, In Transit, Interline, Out for Delivery and Delivered.
- Exceptions include Noted Late, Noted Damage, Noted Short or Lost, Noted Reweigh, Noted Accessorial Added, Reconsignment Notification, Appointment Scheduled, Missed, or Rescheduled, Missed Delivery, and Customs Information Updated.
Compare that list with the 214 table above. The DSDC milestones include terminal movements that neither published 214 guide lists, and they name exceptions such as reweigh and accessorials, which neither 214 guide lists.
The full truckload specifications on DSDC’s GitHub site are labeled public preview. The In-Transit Visibility specification there (version 1.0.0-public-preview) supports pulling events, webhook delivery signed with HMAC, and subscription management. It defines what it calls a “normalized event taxonomy across modes,” with types such as PICKED_UP, ARRIVED_AT_TERMINAL, IN_TRANSIT, DELAY, OUT_FOR_DELIVERY, DELIVERY_ATTEMPTED, DELIVERED and PROOF_OF_DELIVERY, and locations typed as FACILITY, CITY or GEOPOINT with latitude and longitude.
A standard helps only when your carriers implement it. Ask each carrier which DSDC standards and versions it supports, and treat “planned” as “not available” until you have tested it.
Truckload: location pings are not milestones
For full truckload, the richest data is the truck’s position, which comes from the ELD or a telematics unit. Position is useful for estimates and for proving arrival at a facility, but it only becomes a milestone after you apply rules to it.
What the ELD rule requires a device to record
The federal ELD rule exists for hours-of-service records, and its location requirements reflect that. From 49 CFR 395.26 and the ELD functional specifications in appendix A to subpart B:
- The ELD records date, time, location, engine hours, vehicle miles and identifiers when the driver changes duty status.
- When the vehicle “is in motion and there has not been a duty status change or another intermediate recording in the previous 1 hour,” the ELD records an intermediate recording.
- Position is recorded to “hundreds of a degree precision,” and the appendix says position reporting accuracy “will be on the order of ±1mile.”
- During authorized personal use, location is recorded at a single decimal place, “approximately within a 10-mile radius.”
- The device must be able to get a valid position at least once every 5 miles, but it “records CMV location information only during ELD events.”
- The rule directs ELD record production to authorized safety officials, and section 395.22 requires carriers to retain ELD records “so as to protect a driver’s privacy in a manner consistent with sound business practices.”
The practical reading: the regulated ELD record is an hourly, roughly one-mile position log built for compliance. More frequent position feeds come from telematics services rather than from the regulated record, and getting them is a commercial arrangement between you, the carrier and its provider. Expect to ask each carrier for consent, often load by load, and expect some to decline.
Turning pings into milestones with geofences
A geofence is a boundary drawn around a facility. When pings move from outside to inside the boundary, you create an arrival event; when they move out, a departure event. This is how location becomes a milestone.
Write down three rules for each facility before trusting the result:
- The boundary. Draw it around the gate or the yard, not the whole industrial park, or trucks staging next door will “arrive” early.
- The dwell. Require two or more pings inside, or a minimum time inside, before declaring arrival. A truck passing on the frontage road should not count.
- The source of truth. Decide whether a geofence arrival overrides the carrier’s reported arrival, or only fills the gap when the carrier is silent. At your own docks, your gate or yard check-in is usually better than either.
Even with good geofences, location cannot tell you that freight was loaded, that the count matched, or that the consignee signed. Those still come from the carrier’s milestone feed, the bill of lading and proof of delivery.
Choosing a source by mode
Choose the primary feed by mode, then add a second feed only where it fills a specific gap. The table summarizes the trade-offs, based on the guides and rules cited above.
| Mode | Primary feed | Add when | What you get | What you will not get |
|---|---|---|---|---|
| Parcel | Carrier tracking API, polled or pushed | A 214 only if a partner or customer requires it | Package scans, delivery estimates, delay detail, proof of delivery documents | Status for the first day before the package appears in the carrier's system |
| LTL | EDI 214 where you already run EDI; carrier API where the carrier's API exposes more events | The API for exception detail and on-demand checks; DSDC In-Transit Visibility where the carrier supports it | Pickup, terminal, out for delivery and delivery milestones; appointments; reason codes | Truck position between terminals |
| Truckload | Carrier milestones by 214 or API | ELD or telematics location shared by consent, turned into arrivals by geofence | Arrival and departure at stops, position for estimates | Loading, counts and signatures from location alone |
| Your own docks | Gate, yard and dock events in your own systems | Carrier arrival events as a cross-check | Check-in, door assignment, unload start and finish | Anything that happens before the truck reaches your gate |
Four decision rules sit behind that table:
- Use the feed the carrier maintains best. If a carrier’s 214 lists five status codes and its API exposes exceptions and terminal events, the API is the primary feed for that carrier, whatever your default is.
- Do not build a second feed for completeness. Every connection costs upkeep (see the cost section below). Add one only for a gap you can name, such as exceptions, position or proof of delivery.
- Let your own systems win at your own facilities. Your yard and dock records are first-hand. Carrier arrival events at your dock are second-hand.
- Decide by carrier, not only by mode. Two LTL carriers can need different primary feeds. Record the choice and the reason in your integration inventory.
Normalizing every feed into one status model
Normalizing means translating every carrier’s codes into one set of events and reasons that your business defines, while keeping the original message. Do it in a mapping table that people can edit, not in code that only developers can change.
The canonical event record
Each event in your model should carry enough to explain itself months later. A workable minimum:
| Field | What it holds | Why it matters |
|---|---|---|
| Shipment key | Your shipment or load ID, plus carrier PRO or tracking number | Joins carrier data to your orders and invoices |
| Stop | Which pickup or delivery stop the event belongs to | Multi-stop loads need events per stop |
| Event code | Your code, such as PICKED_UP or DELIVERED | One vocabulary for every carrier |
| Event class | Actual, estimate or appointment | Keeps AG and AB out of the actuals |
| Reason code | Your reason, such as REFUSED or WEATHER | Drives exception handling |
| Event time (UTC) | The time the event happened, converted | Correct ordering across time zones |
| Original time and zone | As sent, with the zone you inferred | Lets you audit the conversion |
| Location | Facility, city and state, or coordinates | Supports geofences and terminal views |
| Source | 214, API, webhook, geofence, manual | Tells people how much to trust it |
| Carrier code and SCAC | The carrier’s own code and SCAC | Lets you fix a mapping later |
| Received time | When your system got it | Measures feed delay |
| Raw message reference | Pointer to the stored original | Lets you replay after a mapping fix |
A crosswalk from carrier codes to your events
The crosswalk is the mapping table: one row per carrier code, pointing to your event code and class. The figure shows how the published codes from the guides above line up with a simple internal model.
| Internal event | Class | 214 code (guide) | XPO API status code | DSDC LTL milestone or exception |
|---|---|---|---|---|
| ARRIVED_PICKUP | Actual | X3 (ArcBest) | None listed | None listed |
| PICKED_UP | Actual | AF (ArcBest, Estes) | 1 Picked up | Picked-up |
| AT_TERMINAL | Actual | None listed | 3 Arrived at interim, 6 Arrived at destination | Arrived at Terminal |
| IN_TRANSIT | Actual | X6 (ArcBest) | 2 and 5 En route | In Transit |
| OUT_FOR_DELIVERY | Actual | None listed | 8 Out for delivery | Out for Delivery |
| ARRIVED_DELIVERY | Actual | X1 (ArcBest, Estes) | None listed | None listed |
| DELIVERED | Actual | D1 (ArcBest, Estes) | 23 Delivered by | Delivered |
| DELIVERY_ATTEMPTED | Actual, exception | None listed | 19 Attempted delivery | Missed Delivery (exception) |
| DELAYED | Exception | SD (Estes) | 4 and 7 Delayed en route | Noted Late (exception) |
| RECONSIGNED | Exception | AI (Estes) | None listed | Reconsignment Notification |
| REFUSED | Reason | BS in AT702 (Estes) | 21 Refused delivery, 22 Refused for damage | None listed |
| DELIVERY_ESTIMATE | Estimate | AG (Estes) | Estimated delivery date field | None listed |
| DELIVERY_APPOINTMENT | Appointment | AB, X9 in AT703 (Estes) | 13 Appointment required at destination | Appointment Scheduled, Missed, or Rescheduled |
Notice that the pairings are not one to one. XPO’s code 6 means arrived at the destination terminal, while the 214 code X1 means arrived at the delivery location, the consignee’s dock. Mapping both to “arrived” would tell a customer their freight is at their door when it is still on a terminal floor. Write the meaning of each internal event in one sentence, and test every row against that sentence.
Rules that keep the model honest
These rules come from the guides above. Each one prevents a specific error you would otherwise find in production.
- Store the raw message first, then map. If a mapping row is wrong, you fix the row and replay the stored messages. Without the raw copy, the history stays wrong.
- Route unknown codes to a queue, never to a default. A new code from a carrier should create a work item for the mapping owner, not silently become IN_TRANSIT.
- Separate actuals, estimates and appointments. AG, AB and X9 in a 214 and estimated delivery fields in an API change your plan; they did not happen. Keep them in different fields so that “last actual event” stays true.
- Convert time deliberately. For 214 events without a time code, infer the zone from the event location in MS1 or the stop address, then convert to UTC. For APIs that state UTC, as XPO’s guide does, store them as sent.
- Order by event time, not arrival time. Batched 214s and retried webhooks arrive out of order. Never let an older event overwrite a newer status.
- Deduplicate on a key. Carrier, PRO, event code and event time make a reasonable key. A carrier that resends a batch should not create twice the events.
- Filter system events. API histories can include inquiries and record updates. Map them to an ignored class rather than deleting them.
- Map SCACs to carriers. One carrier can use several SCACs, as ArcBest’s guide shows. Keep both on every event.
- Keep reason codes separate from events. “Delivered” with reason “refused for damage” is a different situation from “delivered” with “normal status,” and your shipment exception rules need both fields.
Example (fictional). Shipment EX-40417 moves LTL from Kansas City to Denver. The carrier sends a 214 with AT7*X1*NS***20261103*0915~ and no time code, and its API later returns status 23, delivered, at 16:58 UTC on the same day. The intake stores both raw messages. The 214 event has no time code, so the parser infers Mountain Time from the Denver stop and converts 09:15 to 16:15 UTC: arrived at the consignee. The API event maps to DELIVERED at 16:58 UTC. The model now shows a 43-minute unload, in the right order, from two sources. Had the parser treated 09:15 as Central Time, arrival would have been stored as 15:15 UTC and the unload as 103 minutes, which is the kind of error that starts a detention dispute.
The pipeline below puts these rules in order. Sources enter on the left of the flow, are logged raw, pass through the mapping table, and land in one event store that every consumer reads.
When a feed goes silent
Silence is the failure you will see most, and it has two causes that need different responses. Either the freight has not generated an event, or the pipe that carries events has broken. Your model should detect both and tell them apart.
Freight silence and pipe silence
Look at the pattern across shipments. If one shipment from a carrier has gone quiet while that carrier’s other shipments keep updating, the freight is probably sitting somewhere, and the right response is an operational check. If every shipment from a carrier stops at the same time, the connection is probably broken, and the right response is an IT check.
Pipe silence has its own signals. For a 214 connection, look at the translator: did files arrive, did they fail translation, are 997 acknowledgments overdue in either direction? For an API, look at error rates, authentication failures and rate-limit responses. For webhooks, look at whether the endpoint was reachable and whether the carrier reports failed deliveries.
Timers per milestone
A silence timer is the time you expect between one milestone and the next. When it runs out without a new event, the model records a NO_UPDATE event against the shipment. That way silence shows up in the same views as everything else.
Set timers per mode and per milestone, from your own history once you have it. Before you do, start with conservative assumptions and adjust. For example, suppose an LTL shipment sits at a terminal: a gap of several hours between terminal events is normal, so a short timer there only creates noise. Out for delivery is different: a gap of a few hours after that milestone usually deserves a look. These figures are hypothetical starting points, not benchmarks. LTL behavior is covered in detail in why LTL tracking goes quiet.
The tracking milestone map is a working template for writing these down: each milestone, its source, the message or event, an owner, when it is expected and the exception rule.
The fallback ladder
When a timer fires, work down this list and stop at the first step that answers the question:
- Check the pipe. Is the carrier’s feed silent for all shipments? If yes, raise it with IT or the carrier’s EDI or API support and stop here for individual shipments.
- Pull on demand. If the primary feed is a 214, call the carrier’s API for this PRO. XPO’s guide, for example, points to its event history call for investigating a possible exception.
- Check location. For truckload, look at the last position and its age. A fresh ping near the consignee changes the question.
- Ask a person. Use the carrier’s terminal or dispatcher contact, and record the answer as a manual event with source MANUAL, so the history shows where it came from.
- Tell the customer what you know. A NO_UPDATE event with a time is more honest than a status that has not changed since yesterday.
What each connection costs to keep running
Every feed has a build cost and a running cost, and the running cost is the one teams forget to budget. This section lists the work, not prices, because the effort depends on your tools, team and carrier count.
| Connection | Setup work | Recurring work | Typical owner |
|---|---|---|---|
| EDI 214 | Partner setup on the VAN or AS2, carrier’s guide reviewed, mapping rows per carrier, test exchange | 997 monitoring, new codes from carriers, version changes, VAN mailbox and certificate upkeep | EDI coordinator |
| Carrier API (poll) | Credentials, client code per carrier, polling schedule by milestone, error handling | Credential rotation, API version retirements, rate-limit tuning, changes to response fields | Integration developer |
| Carrier API (push) | A public endpoint, signature checks, subscriptions per carrier, retry handling | Endpoint availability, subscription renewals, signing secret rotation, replay of missed deliveries | Integration developer and infrastructure |
| Telematics location | Carrier consent process, provider contract or carrier link, geofences per facility | Consent per carrier or load, geofence edits as facilities change, privacy and retention rules | Transportation team with IT |
| Mapping table | Internal event and reason lists, first rows for each carrier | Working the unknown-code queue, replays after fixes, review when a carrier changes guides | A named business owner |
A few observations follow from that table:
- The mapping table needs an owner outside IT. Someone who knows what “arrived at delivery location” means on the dock should approve each row. In our engagements we start with that ownership decision, because requirements come before the build.
- APIs break on the carrier’s schedule. USPS, for example, says its Web Tools API platform was retired on January 25, 2026; the dated list for each carrier is in carrier API migrations. Budget for at least one forced change per carrier API per year as a planning assumption, and adjust from experience.
- EDI is slow to change and slow to break. That is a strength for a status feed. Its weak point is code drift: a carrier adds a code, and your mapping does not know it.
- Telematics adds a people process. Consent, privacy and retention are decisions for your transportation and legal teams, not only an integration task.
Write the owner of each connection into an integration inventory, next to the method and the reason it was chosen. When a feed fails at 6 a.m., the inventory is what tells the on-call person whom to call.
What to do next
Start with the carriers you have, not the architecture you want. A short, ordered list:
- List carriers by mode and volume. Note which ones already send a 214, which offer an API, and which share location. The integration inventory worksheet has the columns.
- Collect every carrier’s guide. Get the 214 implementation guide and the API documentation for each, with version and date. Ask about DSDC standard support.
- Write your internal event and reason lists. One sentence per event. Mark each as actual, estimate or appointment.
- Build the mapping table. One row per carrier code, approved by its business owner. Unknown codes go to a queue.
- Fix time handling first. Decide how you infer time zones for 214 events without a time code, and test with shipments that cross zones.
- Set milestones and timers. Use the tracking milestone map to record each milestone, its source, owner, expected time and exception rule.
- Test with live shipments. Compare your event store with the carrier’s own tracking page for a sample of real shipments per carrier before anyone relies on it.
- Write the runbook. What to check when a feed goes silent, how to replay, and who owns each step.
If you are building the intake and model yourself, the patterns above are enough to start. If you want help writing the requirements and building the interfaces, see our ERP, WMS, TMS and EDI integration work and shipment tracking software, or contact us to scope it.
