How do you tell if freight automation actually shortened the time from quote to booked load?
Key Takeaways
You tell if freight automation actually shortened the time from quote to booked load by measuring the same thing, the same way, on the same lanes, before and after. The measure should run from the moment the order is ready to ship to the moment the carrier's written acceptance and rate confirmation are on file. Use timestamps from each load, not estimates. Compare medians, not only averages, and count how many loads still needed a person and why. An AI logistics agent such as Wilson, by Cartage, records these timestamps on every load as it works, so the after-period measurement is built in.
- Define the start and end points before measuring: order ready to confirmed booking is the most useful span.
- Split the span into steps: request sent, first quote, last quote, tender, acceptance, rate confirmation.
- Use the same lanes, modes and a similar volume in both periods.
- Use medians and show the slow tail, because a few stuck loads can hide or exaggerate the change.
- Measure people's time too: loads booked without a person and the reasons the rest needed one.
- Do not count "quote received" as done. A quote is not a booking.
What exactly should be measured?
The span that should be measured is from order ready to confirmed booking, broken into steps. An AI logistics agent records each step's time on the load. The table defines the points and where each timestamp usually comes from.
| Point | Definition | Typical source before automation | Source with an AI logistics agent |
|---|---|---|---|
| Order ready | Order released for shipping with enough data to quote | ERP release time or order file time | Time the order arrived in the feed or email |
| Request sent | First quote request sent to a carrier | Sent email time | Recorded request time |
| First quote | First complete quote received | Email time | Recorded reply time |
| Decision | Carrier chosen | Rarely recorded | Recorded decision time |
| Tender sent | Load offered to the chosen carrier | Sent email time | Recorded tender time |
| Accepted | Carrier accepts in writing | Email time | Recorded acceptance time |
| Booked | Rate confirmation on file and checked | Often not recorded | Recorded when checked |
Why split the span into steps?
Splitting the span into steps shows where time is spent and where it changed. A total time that drops from eleven hours to four tells you something improved. The steps tell you what: the wait before the first request, the carriers' reply time, the decision, or the chase for the acceptance. Each step has a different cause and a different fix. Without the split, a team cannot tell whether to work on its own process, its data or its carriers.
Why start at order ready and not at the request?
Starting at order ready captures the wait before anyone sends a request, which is often the biggest delay. An order may sit for hours because the coordinator is busy, the data is incomplete or the order arrived late in the day. Starting at the request hides that wait and makes the process look faster than it is.
What other measures are worth tracking?
Other measures are worth tracking alongside booking time: the share of loads booked at least a day before pickup, the share of loads where the first-choice carrier accepted, the number of tenders per load before acceptance and the share of loads that needed approval. Each one adds context. For example, a shorter booking time with more tenders per load may mean the agent moves to backups faster, which is good, or that the primary carrier is declining more, which needs a conversation.
Why end at the rate confirmation and not at the quote?
Ending at the rate confirmation measures what the business actually needs: a carrier committed to the load. A process that gets quotes quickly but takes a day to get a carrier's acceptance has not shortened the time to a booked load. For what "confirmed" means, see turning an ERP order into a confirmed carrier booking.
Why does the measure matter to the business?
The measure matters because the time to a booked load decides how many options the company has. A load booked early in the day has more carriers to choose from, at better rates, than the same load booked late in the afternoon. A load booked a day ahead rarely misses its pickup. Shorter booking time also frees the team for the work that needs judgment. The number is a way to see whether the freight process is getting ahead of the day or falling behind it.
What about loads that never needed quoting?
Some loads go straight to a contract carrier at a contract rate, without quotes. For those, the measure runs from order ready to written acceptance and rate confirmation, skipping the quote steps. They should be reported separately from spot-quoted loads, because they behave differently and mixing them can hide the change in either group.
Should business hours or clock hours be used?
Either can work, but use the same in both periods and say which. Clock hours show the real wait the customer experiences. Business hours remove nights and weekends, which can make comparisons fairer when order release times differ. Many teams report both.
How should the before period be measured?
The before period should be measured from the records that exist: ERP release times, email timestamps and rate confirmations. An AI logistics agent cannot record the past, so the before period takes some manual work. A sample of loads is enough if it is chosen fairly.
Which lanes should be measured?
The lanes to measure are the ones the automation will run on first, chosen before it starts. Busy lanes give more data. Lanes with frequent problems show the effect on slow loads. A mix of modes, such as one LTL lane and one truckload lane, shows whether the change holds across modes. Avoid choosing lanes after the fact based on which looked best.
How many loads should the sample include?
The sample should include enough loads on each lane to show the normal range, including some slow ones. For busy lanes, a few weeks of loads is usually enough. For quiet lanes, a longer period may be needed. The sample should not be picked by hand; take every load in the period on the chosen lanes, or every nth load.
How should phone bookings be handled in the before period?
Phone bookings are hard to time in the before period because nothing records the call. If the coordinator usually sends a confirmation email after the call, use that time. If not, mark the acceptance time as missing. Do not fill it with an estimate. The share of loads booked by phone is itself useful to know, because those are the loads most likely to be missing a written rate confirmation.
What if timestamps are missing?
Missing timestamps are common in the before period. The tender may be in one person's sent mail; the acceptance may have come by phone. Record what can be found and mark the rest as missing, rather than estimating. Show how many loads had complete timestamps. A comparison built on guesses is not reliable.
Who should do the before-period work?
Someone who knows the lanes and the inbox should do it, with clear instructions: which lanes, which period, which definitions and where to find each timestamp. A coordinator can usually pull a few weeks of loads in a day or two. The work is tedious but valuable: it also shows how much time the team spends on each load today, which is part of the business case.
Can the vendor help with the before period?
The vendor can help define the measures and the sample, but the data belongs to the company and should be pulled by the company. That keeps the before period honest. Cartage helps map the highest-impact workflow during discovery, which often points to the lanes worth measuring.
What else should be recorded?
Besides times, record who worked each load, how many emails and calls it took if that can be seen, and whether anything went wrong: a carrier that did not answer, missing data, a rate above the limit. These details explain the slow loads and show where the time went.
How should the after period be measured?
The after period should be measured on the same lanes, with the same definitions, once the automation is running normally. An AI logistics agent records each step's timestamp, so the data comes from the load records directly.
Should the team change anything else at the same time?
Ideally, no. If the company changes carriers, renegotiates rates or reorganizes the team during the same period, it becomes hard to tell which change caused the result. When other changes cannot be avoided, write them down with their dates so the comparison can be read with them in mind.
When should the after period start?
The after period should start once the first setup weeks are over and the rules are stable. In the first weeks, the team is still adjusting approved carriers, limits and data. Measuring then mixes setup effects with the steady result. Many teams measure both: the first weeks to see progress and a later period to see the settled result.
What about loads that arrive late in the day?
Loads that arrive late in the day behave differently. An order released at 4:30 p.m. for next-day pickup may not be booked until the next morning in the before period, because the team has gone home. In the after period, an agent can request quotes and tender the load that evening if carriers answer. It helps to report results by time of release, such as morning, afternoon and late-day orders, to see where the change comes from.
How should volume differences be handled?
Volume differences should be noted and, if large, accounted for. A busy season makes everything slower: carriers are stretched, and coordinators have more loads. A quiet period makes everything faster. Compare periods with similar volume when possible, or compare the same weeks year over year. If that is not possible, show the volume next to the results.
How should approvals be counted?
Approvals should be counted as part of the time, not removed from it. If a load waits two hours for a rate approval, that wait is real and belongs in the booking time. Report the approval wait as its own step so the team can see whether approvals are fast enough, and whether limits that trigger many approvals need review.
What should be recorded about each load in the after period?
Each load in the after period should have all the step timestamps, the carriers asked and their replies, whether a person stepped in and why, and the final booked rate. Wilson records all of this as part of doing the work. The team can export the records for the chosen lanes and period and build the comparison table directly.
Should the same lanes be used?
Yes. Different lanes have different carriers and markets. If the before period covers hard lanes and the after period covers easy ones, the improvement is not real. Pick the lanes first and measure both periods on them.
How should the results be compared?
The results should be compared step by step, with medians, the slow tail and the share of loads that needed a person. An AI logistics agent can produce these from its records. The table shows a hypothetical comparison on one lane.
| Measure | Before (hypothetical) | After (hypothetical) |
|---|---|---|
| Loads in period | 62 | 58 |
| Median order ready to request sent | 3.5 hours | 10 minutes |
| Median request to first complete quote | 2.0 hours | 1.5 hours |
| Median decision to written acceptance | 4.0 hours | 1.2 hours |
| Median order ready to booked | 11 hours | 3.5 hours |
| Slowest 10% of loads, order ready to booked | Over 30 hours | About 9 hours |
| Loads booked without a person | Not applicable | 47 of 58 |
| Main reasons a person was needed | Not recorded | Rate above limit (6), missing data (3), no approved carrier (2) |
The numbers are illustrative only. They show the kind of table to build, not a promised result.
How should modes be compared?
Modes should be compared separately, because their booking steps differ. LTL booking often depends on getting complete shipment data and a pickup confirmation. Truckload booking depends on finding a truck and getting a written commitment. Parcel booking is often immediate. Mixing modes in one median can hide a large change in one mode behind a stable mode. Report each mode's results on its own line.
Why use medians?
Medians show the typical load, unaffected by a few extreme cases. Averages can be pulled up by one load that sat over a weekend, or down by a few instant bookings. Show both if you like, but lead with the median.
What does a fair conclusion sound like?
A fair conclusion names the lanes, the periods, the definitions and the result, with the caveats. For example: "On three lanes, comparing six weeks in September with six weeks in November at similar volume, the median time from order ready to confirmed booking fell from about eleven hours to about four. Most of the change came from the wait before the first request. Eleven of 58 loads still needed a person, mainly for rates above the limit." The numbers here are illustrative; the shape is what matters.
How should outliers be treated?
Outliers should be kept in the data and explained, not removed. A load that took two days because the customer changed the order three times is real. Removing it makes both periods look better than they were. If a few outliers dominate, show the result with and without them, and say why.
Why show the slow tail?
The slow tail matters because the slowest loads cause the most damage: missed pickups, expensive spot rates and unhappy customers. Automation often helps the slow tail most, because it removes the waits caused by busy people and missed emails. Showing the slowest 10% next to the median makes that visible.
How should the results be presented?
The results should be presented as a short table per lane, with the definitions written above it and the volume shown. A single headline number, such as "booking time cut by X%", is tempting but invites questions about lanes, seasons and definitions. A table with medians, the slow tail, load counts and the share of loads without a person answers those questions in advance. Notes on what changed besides the automation, such as a new carrier or a rate change, help readers judge.
Which steps usually improve most?
The steps that depend on someone noticing usually improve most: sending the first request after the order is ready, following up when a carrier is slow, and chasing the acceptance after the tender. The step that depends on carriers, how fast they reply with a quote, usually improves less, because it depends on the carriers. If the request-to-quote time does not change much, that is normal.
Does faster booking change the rates paid?
It can. Loads booked earlier in the day, or earlier before pickup, often have more carriers available, which can mean better rates. But rates also move with the market, so a change in rates between the two periods may have nothing to do with booking speed. Compare the booked rates on the same lanes against a market reference if the team uses one, and treat any rate effect as a separate, more cautious finding.
What about people's time?
People's time is the other half of the result. Count the loads booked without a person, and record why the rest needed one. A shorter booking time with the same amount of human work is a smaller gain than a shorter time with much less work. For a per-load view of the business case, see the business case for freight automation, one load at a time.
What can make the comparison misleading?
Several things can make the comparison misleading. An AI logistics agent's records reduce some of them, but the team should check for all of them before drawing conclusions.
| Risk | How it misleads | How to avoid it |
|---|---|---|
| Different definitions | "Booked" means tender sent before, rate confirmation after | Write definitions first; apply to both periods |
| Different lanes | Easier lanes in the after period | Fix the lanes before measuring |
| Different volume or season | Quiet period after, busy before | Compare similar periods; show volume |
| Missing before data | Only easy loads had timestamps | Show how many loads had complete data |
| Setup period included | Early weeks drag results down | Separate setup weeks from steady state |
| Ignoring human work | Faster but same effort | Count loads without a person |
What if data quality is the bottleneck?
If many loads wait for missing data, such as pallet counts or delivery requirements, the comparison shows it in the time between order ready and request sent. That wait is not a carrier or automation problem; it is an ERP data problem. Adding the missing fields to the order release report often removes most of it. The measurement makes the case for that fix.
What if a carrier is the bottleneck?
If one carrier is slow to accept on a lane, the step-by-step view shows it: the decision-to-acceptance time for that carrier stands out. The team can talk to the carrier, change the order of offers on that lane or shorten the deadline before moving to a backup. The measurement turns a feeling ("they're always slow") into a fact the carrier can act on.
What if the improvement is small?
If the improvement is small, the step-by-step view shows why. Maybe orders arrive with missing data, and the agent waits for answers. Maybe the rate limits are out of date, so most loads need approval. Maybe one lane has too few approved carriers. Each of these is a fixable rule or data problem, not a limit of automation.
How often should the measure be repeated?
The measure should be repeated after the first steady period and then periodically, such as each quarter, as more lanes are added. Each repeat shows whether the result holds as volume grows and new lanes join, and whether rule changes made a difference. Because Wilson records every step, repeating the measure is a matter of exporting the records.
What if the improvement is large?
If the improvement is large, check the definitions and the lanes again before sharing it, then share it with the details. A result that holds up to questions is more useful than a big number that does not.
How does Wilson, by Cartage, make the measurement easier?
Wilson, by Cartage, is an AI logistics agent that records the time of every step on every load: order received, data checked, requests sent, quotes received, decision, tender, acceptance and rate confirmation. Because Wilson does each step, the timestamps are exact and complete. Wilson's records are searchable and exportable with a change history, so the team can build the comparison from them. For what happens when a carrier is slow to answer, see how long carriers should have to answer a quote request.
Wilson replaces the TMS for freight execution and connects to the company's ERP. It handles LTL, FTL, ocean, drayage, parcel and cross-border freight.
Does Wilson help with the before period?
Wilson records loads from the day it starts, so it cannot fill in the before period. Cartage can help the team choose the lanes and definitions during discovery, so the before sample is pulled in a way that matches what Wilson will record.
What does a measurement look like with Wilson?
A hypothetical measurement: the team picks three lanes and pulls six weeks of loads from before Wilson started, using ERP release times and email timestamps. Some loads lack the acceptance time and are marked incomplete. After Wilson has run on those lanes for a few weeks and the rules are stable, the team exports six weeks of Wilson's records for the same lanes. Both periods go into the same table, with the same definitions. The team reviews the loads that still needed a person and adjusts two rate limits and one approved carrier list.
What stays with the team?
The team chooses the lanes, the definitions and the comparison periods, and decides what the results mean for the rules. Wilson starts with assisted execution, and more routine bookings run within the rules as the team sees the results.
How is Wilson priced?
Wilson is free on Wilson's network: the shipper pays the freight rate and nothing on top. When a shipper keeps its own carriers, the fee is 2% to 0.5% of what Wilson pays them, falling with volume. There are no seat fees, no setup cost and no contract.
How do teams start?
Most teams start Wilson on a few lanes and expand from there. Measuring the before period on those lanes, before Wilson starts, makes the comparison possible.
FAQs
How do you tell if freight automation actually shortened the time from quote to booked load?
Measure order ready to confirmed booking on the same lanes, before and after, with timestamps from each load, and compare medians, the slow tail and the loads that still needed a person. An AI logistics agent like Wilson, by Cartage, records these timestamps on every load.
When does the clock stop?
When the carrier's written acceptance and a matching rate confirmation are on file. An AI logistics agent marks the load booked only then, so the measure reflects a real commitment.
What if we have no timestamps from before?
Use what exists, such as ERP release times and email timestamps, mark missing data instead of guessing, and show how many loads had complete records. An AI logistics agent records every step from the start.
Which step usually improves most?
The waits that depend on someone noticing: sending the first request and chasing the acceptance. An AI logistics agent does both as soon as the previous step ends.
Sources
- Cartage, Wilson product FAQ and product scope, November 2026: https://cartage.ai
If you want to see the before-and-after on your own lanes, book a demo with the team behind Wilson, the AI logistics agent by Cartage.