Why carrier quote emails arrive in 12 different formats.
Post a dry-van load to DAT and Truckstop at 8:45am. By 10:30am, 40 carriers have responded. You open your Gmail. The first reply is a clean DAT-format block with lane, equipment, and rate in separate rows. The second is a three-sentence inline pitch with the rate buried in the middle. The third is a PDF rate confirmation as an attachment. The fourth is a spreadsheet — a truck list — with 12 rows, one of which covers your lane. The fifth is a forwarded thread where a dispatcher is quoting on behalf of a carrier, and the actual number is three emails deep. The next 35 look like some remix of those five.
This is not a solvable problem at the format level. There is no standard carrier-quote email schema and no entity in the industry has the authority to enforce one. What is solvable is the extraction layer — getting the rate, lane, and equipment out of whatever the carrier sent, reliably enough that a broker can trust the ranked list without reading every thread.
This post maps the twelve most common carrier-quote email formats we see in live broker inboxes, explains why each one exists, and walks through what it takes to extract structured data out of all of them without a human doing the reading.
The twelve formats (in rough order of volume)
1. Inline plain text, rate embedded in a paragraph
Most common single format. A dispatcher writes three to five sentences — "Hi, we can run this for $2,950 all in, pickup tomorrow 8am, our MC is 123456, let me know" — with the rate fused into the prose. Easy to read manually, a pain to parse at scale because the rate position, format ($2,950 vs 2950 vs $2,950.00), and even currency indicator drift reply-to-reply.
2. DAT-format reply block
Structured block with lane, equipment, rate, weight, and carrier info in separate labeled fields. Easy to parse when the carrier is replying through DAT's native tools; much less easy when they've pasted DAT-style into another email client and the field alignment has collapsed.
3. Truckstop-format reply
Similar structure to DAT but different field labels and ordering. Also frequently re-pasted into generic email, losing the exact format. Parsers that handle DAT cleanly often mis-classify Truckstop replies because the heuristic keywords differ.
4. PDF rate confirmation attachment
Carrier attaches a PDF — often their own rate confirmation template or a signed version of the broker's load tender. The email body says "See attached" or is empty. All the useful data is in the PDF, which needs to be text-extracted (easy for typed PDFs, harder for scanned / flattened ones).
5. Spreadsheet attachment (trucklist)
Carrier sends a spreadsheet — usually Excel — with one row per truck and one of the rows covers your lane. Matching the row is a lane + equipment join. Not hard to do well, but requires actually opening the spreadsheet programmatically rather than just reading the email body.
6. Forwarded dispatcher quote
A dispatcher at a 3PL is quoting on behalf of a carrier. The actual pricing thread is a forward of an earlier exchange, and the relevant rate appears in the third or fourth quoted block of the email. Parsers that only look at the latest reply miss the number entirely.
7. TMS-generated quote email
Carrier is running McLeod, Tai, or a similar TMS that has a "reply with quote" feature. The email is machine-generated with a reasonably consistent structure, but the template varies per TMS vendor and per carrier brokerage deployment.
8. Short one-liner with the rate only
"$2,850." That's the whole email. No greeting, no MC, no pickup time. Extractable, but the downstream ranking needs to handle the absence of every field other than the number.
9. "Call me" reply with no rate
The carrier has not committed to a number in writing. Has to be classified as a no-rate response and ranked below any carrier who did quote, unless the broker has a historical relationship with the carrier that justifies the outbound call.
10. Multi-load batch reply
Carrier is responding to multiple load postings in one email, listing each with a separate rate. Parsers have to disambiguate which line item corresponds to the load the carrier is replying to.
11. Non-English reply
Spanish is the most common second language we see in the US broker dataset, followed by Punjabi. The content structure is the same — rate, lane, equipment — but the surrounding prose needs multilingual handling.
12. OOO auto-reply / delivery confirmation / unrelated noise
Not a quote at all, but it lands in the same inbox thread. Needs to be classified and dismissed, not ranked.
What it costs brokers
On our data, a broker triaging 30-50 carrier replies per posted load spends 20-30 minutes per load on the reading step alone — not the negotiation, not the booking, just the reading. A brokerage posting 40 loads a week loses 13-20 hours of broker time to this task. Almost all of that time is a function of format variability. If every reply came in the same structure, the same broker would triage the same volume in 3-5 minutes per load.
This is not theoretical. We've instrumented broker workflows and the format-variability tax is real, measurable, and recoverable.
Why you can't just regex your way out of this
Brokerages that try to build their own rate-extraction layer always start with regex. It works fine on format #2 (DAT) and format #7 (TMS-generated), which together cover maybe 25-35% of volume. Everything else — especially formats #1, #4, #5, #6, #10 — needs a model that can read prose and make inferences.
The modern answer is LLM-based extraction, scoped carefully so it does not hallucinate. Keelway's extraction runs on a constrained LLM pipeline: the model can only output numbers that appear in the source, tags uncertainty explicitly, and flags any reply where the extracted rate confidence is below threshold for human review. Accuracy on our evaluation set is above 95% for numeric quotes that exist in the email.
How Keelway handles all twelve
Every inbound carrier reply goes through the same pipeline: classify format, extract rate and lane and equipment, check FMCSA trust, compare against market rate, rank against the other 39 replies on the same load. The broker sees one ranked list and makes the decision. Whether the input was a plain-text one-liner or a 12-row Excel attachment is invisible at the decision layer.
For the full capability spec, see Carrier Email Automation. If your brokerage runs on Tai, McLeod, Aljex, Revenova, Turvo, or Rose Rocket, we write the accepted carrier + rate back into your TMS once the decision is made.