By Luis Lopez, AI transportation consultant, CEO of Go Hub.io Holdings Corp and subsidiaries, and host of the Freight Guru Podcast
Every freight brokerage sends out quotes that never turn into loads. Most teams shrug and move on to the next request. That is a waste, because the pattern of quotes you win and lose is one of the most useful data sets a brokerage owns. It tells you where your pricing is off, which customers are really shopping you, and which lanes you should stop chasing.
Win-loss analysis is the discipline of learning from that pattern. AI makes it practical at scale, but only if the basics are in place first.
Step one: log every quote and its outcome
You cannot analyze what you did not record. Many brokerages quote by email, phone and portal, and only the quotes that become loads ever reach the system of record. Lost quotes disappear into inboxes.
A usable win-loss data set needs, at minimum:
- Customer and contact
- Origin, destination and lane
- Equipment type and mode
- Pickup date and the date the request arrived
- Quoted rate and the estimated carrier cost behind it
- Time from request to response
- Outcome: won, lost, no response, or customer canceled
- Loss reason, when the customer gives one
The outcome field matters most. “No response” is not the same as “lost on price,” and lumping them together hides what is actually happening.
The features that drive wins and losses
Once quotes are logged, you can study the variables that move win rates. The usual suspects in freight:
- Lane. Some lanes you win easily because you have reliable carriers there. Others you rarely win.
- Lead time. Same-day requests behave very differently from requests a week out.
- Equipment. Specialized equipment may carry less competition, or more uncertainty in your cost.
- Price versus market. How your quote compared to a market benchmark for that lane and date.
- Response time. In many spot situations, the first credible quote gets the load.
- Customer. Some shippers award on price alone; others value reliability and history.
For context on how a quote gets built in the first place, see AI freight quoting and pricing.
Models that estimate win probability
With enough history, a predictive model can estimate the probability of winning a quote at a given price. Feed it the lane, lead time, equipment, customer, response speed and price relative to market, and it returns an estimate. Run it at several price points and you get a curve: how likely you are to win at each rate.
That curve is valuable because it connects price to margin. A rate that wins almost every time may leave money on the table. A rate that rarely wins wastes your team’s time. The useful zone sits between the two, and it is different for every lane and customer.
These models do not need to be exotic. Fairly standard classification methods work well on this kind of tabular data. What matters more is clean inputs and honest outcomes.
Finding lanes where you are always too high or too low
One of the quickest wins from win-loss analysis is spotting systematic pricing errors. Group quotes by lane and look at two signals:
- Lanes with very low win rates where your quotes sit consistently above market. Either your carrier costs are high on those lanes or your pricing logic is adding too much margin.
- Lanes with very high win rates where you consistently quote below market. You may be winning freight you could have priced higher, or your cost estimates may be too optimistic and you are losing money on execution.
The second group is easy to miss because winning feels good. Comparing quoted rates to what you actually paid the carrier closes that gap. For more on building benchmarks, see AI lane rate benchmarking with historical data.
Customer-level patterns
Win-loss data also tells you how each customer behaves. Some patterns to look for:
- Customers who request many quotes but award very few, which often means you are a price check, not a partner.
- Customers whose win rate drops sharply after a certain lead time.
- Customers who award consistently on certain lanes and never on others.
- Customers whose behavior changed recently, which can signal a new competitor or a new buyer.
Sales teams can use these patterns to decide where to invest time. A customer that awards on reliability deserves a different conversation than one that awards on price alone.
Building the feedback loop
Win-loss analysis only pays off when it changes behavior. A simple loop looks like this:
- Log quotes and outcomes automatically wherever possible.
- Review win rates by lane, customer and rep on a regular cadence.
- Adjust pricing rules or margin targets where the data shows consistent errors.
- Measure whether the change improved win rate, margin, or both.
- Repeat.
AI can help with each step, from extracting outcomes out of email replies to summarizing why a lane is underperforming. Pricing changes should still pass through guardrails. This piece on AI pricing guardrails for freight brokers covers how to keep automated pricing from drifting.
Pitfalls to avoid
Missing loss reasons
Most lost quotes come with no explanation. Customers rarely tell you why they went elsewhere. Do not fill the gap with guesses. Record “unknown” honestly, and ask for feedback when the relationship allows it. A short, polite question after a lost bid sometimes produces the most useful data you will get all week.
Selection bias
Your data only includes the quotes you chose to send. If your team ignores certain lanes or customers, the model never learns about them. It also does not see freight the customer never sent you. Treat win-probability estimates as accurate only for situations similar to your history.
Chasing volume at bad margin
A model that maximizes win rate will happily recommend lower prices. Volume is not the goal; profitable volume is. Always judge pricing changes on margin and execution cost, not on wins alone.
Ignoring execution
Winning a load you cannot cover reliably is worse than losing it. Win-loss analysis should connect to service outcomes, so you can see whether certain wins regularly turn into late pickups or costly recoveries.
Keep human judgment in the loop
As an AI transportation consultant, I see win-loss analysis as a decision aid, not an autopilot. A model cannot know that a customer is about to open a new distribution center, that a carrier partner just added trucks in a region, or that a relationship is worth protecting through a soft season. Experienced reps and pricing managers bring that context. The best results come when the model surfaces the patterns and people decide what to do with them.
For more conversations on freight and AI, subscribe to the Freight Guru Podcast.
About the author: Luis Lopez is a Miami-based AI transportation consultant and logistics entrepreneur, the CEO of Go Hub.io Holdings Corp and subsidiaries, and host of the Freight Guru Podcast.