Why are the rates never right? (part 3)

For all bits and bytes of information we can garner about the market, we’re always using past information. Yet, all our decisions are for the future. The real-time price processing desired today gets us closer insofar it’s nearer the spigot of expectation. Reality is the rubber meeting the road ever after that. No one knows the future. So we all guess.

Guessing blind is a strategy some dare try, the rest try to employ some mix of external guidance and/or internal history. As with other components in geography, density, and parameter there are considerations for how to use time in these calculations.

In part 3, I’ll close with how we treat the balancing rod of past and future balancing any model, whether driven by organic or artificial intelligence.

Past

The easiest way to think of these questions is what does this decision do to my pool of possible results? I don’t need much of an equation to tell me the pool of results will be bigger if I include anything from 100 days ago versus only yesterday. The decision between a couple days or weeks is another matter.

Everyone is making these decisions for sake of relevance. Three weeks history on the local Chicago run is useless for spot operation but may be all there is for anything worthy to meet criteria for the Massena, NY location.

The operator is sizing these considerations on autopilot. They are very good at reweighting parameters to zone in. Many systems the human then leverages are more fixed or systematic for figuring which series of information is most relevant. Both systems have to inevitably communicate and translate from the inputs. Answers are found in the process but there are never really THE answers for everything.

The long series trend below has monthly, 3, and 6 month rolling averages for an all-in dry van spot rate using DAT’s linehaul since 2010. They can be thought of as three different simplistic models of pricing trends.

Each of these lines contains the monthly number, but as results are added, things begin to behave differently as we saw in part one 2 with density. Add enough results and we’re back to a straight line like the dice throws.

All of these models are also correct representations of what has happened. Their use situation dependent.

AI is already incorporating some of these human-led decisions into models, co-packaging internal and external inputs. The result is a more sophisticated and catered experience for the industry. They too do their job very well, but cannot answer every user’s brand of risk tolerance or budgeting.

Which carries us into the future.

Future

“Past performance is not indicative of future result” is plastered across any financial product for good reason. There are easy bets you can make based on yesterday but no one makes them because there’s no incentive. Will the sun come up tomorrow? 100% Will there be any rain tomorrow? Ehhhh, 40% chance. Welcome, meteorologist.

For the sake of argument, we have perfect history and parsing of data to deliver a pristine market rate for everyone inside an organization to use up to the entire industry. I’m in NY and you in LA, but we both see Chicago to Atlanta at $1,100.

We have the same carrier pool or shipper base. The freight on hand is 10,000 lbs of easily loaded material in a good spot. If either submit this market rate no one wins out. Market forces would pull someone to offer $950 or $1,000 instead. One system suggests $975 based on some intelligent arbiter. It settles at $1025 for tomorrow. Pick at noon.

You walk in at 8am to a dreaded blown turbo in the wee hours of the morning. The market rate is not going to update. The system intelligences may. The receiver or end customer calls in between and it’s ok to kick it out a day or two. What then?

The ability to store and pass dynamic histories for reference are a whole other can of worms worthy of future coverage itself, but the lack of integration and visibility to update or expose information can be a culprit all it’s own. Absent of any care of what the final number is supposed to be.

At a 1:1 transactional level one may argue there is a route to a perfect match for a price on a shipment knowing all the details. They’re probably right. Any scalable system, however, has to address each “what if” thereafter, as it may apply to changes in the shipment in question to something fitting within the criteria of a search close to it. It has to be fluid.

The timing component has to run from very close to now to very far in the future. The search for today in four hours or four days or four weeks from now may or may not need to change. If a strike happens. Or a blizzard, that system will need to crawl around all of these events.

Once the pole lengthens for a termed rate, one that stays consistent over a period of 3, 6, or 12+ months the pooling and formulations can look quite different. $800 over 100 shipments for the next 12 months is an option both sides can’t guarantee delivery on. Both sides will come to a decision using some mix of inputs again, and will go the 180 or 365+ days backwards for that theoretical mean to guide.

Thereafter, no one is in discovery any longer. It’s all give and take. Until enough of the parameters change.

Convergence

Most tools and histories have to stop here because it moves away from discovery into discretion. It’s not their function. Knowing the market rate is laughable if needing to move 10 of anything expeditiously, it changes again whether they need to go today or through next week. It is nonetheless important to establish a gravity with a benchmark in mind or use.

Similar thought exercises break the golden rate theory figuring differences between committing to 10 or 100 of something, then again whether they go tomorrow or over the next year. Or if they reject for blades of grass in the trailer.

Which all gets us back to our own senses of capacity, operational acumen, and insights. Expecting complete alignment or an omniscience is a fools errand. It therefore comes down to the work of teams, vendors, and partners to have honest assessments of their needs and trade-offs in producing or leveraging scaled systems.

The secret sauce is found when everyone has a hand in the ingredients versus some expectation the next batch will just do the trick.