Last Monday we saw an opportunity and took a pretty substantial risk by adjusting our pricing model to get more aggressive during a time when you’d seasonally expect rates to strengthen. This decision wasn’t based on changes in any of the rate analytics tools. In fact, had we relied only on benchmark data, we likely would have missed this opportunity. Here’s some of what we saw that made us confident in the decision: -Our DFM matching metrics showed a substantial increase in match efficacy. I posted about this on 10/12, and it’s held since. -Model bias had trended from a 2-3% under-prediction to a 2-3% over-prediction. -Gross margin on loads priced by the model expanded against relatively flat cost inputs. -Spot conversion on our pricing bots hit a wall and was regressing fast. The results over the last week+ have been remarkable. Conversion is up over 350bps, revenue is up 42%, and margin is holding within 25bps of the week prior. I’m not posting this to glorify a decision we made, but rather to show how the yeoman’s work of spending months building instrumentation and reporting can allow you to spot opportunity well before the competition and capitalize with confidence. You can’t manage what you don’t measure, and you can’t measure what you don’t monitor.

Was it pretty? No. Do I feel good about it? No. Was it a win? Yes. Go Bucks!!! 🌰 🐐

Following up on yesterday’s post with some early results from this week. This is strictly API/portal bidding off our pricing model, compared to last week. Conversion rate: +370bps Gross margin: +20bps Model error: -330bps Model bias: 2.8% over to 2.1% under Roughly 10-15% of the cost basis came out of the model WoW on a same-lane basis, and margin still held. That’s incredibly telling. No idea how long this lasts, but we’re going to stack like IHOP while we can.

Anyone else having trouble chasing the spot market down to preserve volume and conversion? Our spot conversion was down ~250bps last week versus the week prior and is continuing to trend down this week. Margin, meanwhile, was up ~110bps. We’ve been changing the cost blend and manually tweaking rates down all week, and while we’ve made some progress, I still feel like it hasn’t been enough. This seems seasonally abnormal, but maybe things have shifted around a bit and it’ll pick up leading into Halloween. Curious what everyone else is seeing. Here’s an example of a previously high-converting lane that’s fallen off a cliff over the last week or two.

DFM efficacy data is lowkey a great indicator of market tightness. Friday felt surprisingly loose and our spot models were chasing the market, so I was curious to see what the matching metrics looked like this morning. Sure enough, match rates were up substantially on Friday. What I found most interesting: matches as a share of total attempts barely moved, but matches as a share of loads covered more than doubled. I’m curious whether adding this as an input to the pricing model would actually lift model fit, or just add noise and circular/redundant signal.

It’s a bad day to be a turtle. Go Bucks 🌰

Find new data sources, build tools, and see real results in under a week. That's where we are now. I don't know how anyone sees this as anything less than amazing.

The AV/EV summit today with @einrideofficial and @EaseLogistics was really awesome. There are some real freight people at Einride, which makes me very optimistic about where this goes over the next several years. I can't imagine standing there watching this work in real life and not walking away with a sense of wonder and excitement. It felt like watching a robot vacuum clean the house but with an actual truck on the road. We have a long way to go before this scales and meets a standard that will quiet the critics, but it's moving at a breakneck pace right in front of our eyes, and I think that's pretty cool.

Imagine being bearish.

Bullish on goat data.

The team here at FreightNerd Media has obtained an exclusive photo of the moment the algorithmic pricing doomers heard about the CHR/RXO deal.

Champions.

OTR Solutions put together a really great event in Atlanta this week. I think they have one of the most promising product roadmaps in the space. The only thing really holding them back is their affiliation with the SEC. Back to brokering...

We've been developing a first pass at digital freight matching (DFM) over the last couple of weeks using AI-assisted development practices. Last Friday was our first chance to take it for a spin. The model is still very rudimentary, but I'm honestly shocked by the results. I'd estimate our total investment in this project so far at under $5000, which blows my mind. The results from Friday's test rollout: 1,313 loads offered 13,875 offers sent 523 loads covered by any method 81 covered by the DFM tool In summary, just over 6% of total loads offered were covered with an assist from the DFM tool. Looking at just covered loads, the share was 15.5% (and we'll have additional chances to cover the remaining loads). It's truly remarkable how quickly tools like this can be built, tested, and iterated with the help of AI coding platforms.

It’s a bad day to be an Illini. Go Bucks! 🐢 🌰

Sweet Fancy Moses...

Product development is a contact sport. I'm a firm believer that the best innovation happens as close to the user as possible. Practicing what I preach, I've been booking loads all week and iterating on new tooling to make the process more efficient. Needless to say, I haven't showered or eaten a hot meal this week. If this keeps up through Friday, please send help.

The more I use AI for product development, the more I'm starting to realize that most of the debate about this technology is happening amongst folks who aren't using it to build. Once you spend any amount of time with it, I don't understand how you can feasibly conclude that we aren't standing at the precipice of one of the biggest stepwise functions in the history of software products. I'm not saying that AI can turn some hobbyist into a world-class software engineer, but the productivity and speed gains are truly unbelievable. If you find yourself struggling to pick a side in this ongoing discussion, my suggestion would be to go chase an idea and build something. I'm guessing you'll be totally flabbergasted.

Go Bucks 🌰