Why are the rates never right? (part 2)

Now that a net is made, it’s time to collect some results. We spent part 1 considering the why and how of geographical coverage. There’s no one flavor to rule them all, but once the jurisdictions are made, analysis can ensue.

Within part 2 is the meat and potatoes of reasoning behind price variations at any point in time. Which come down to how much data is there, and what else is influencing it? Their categories are density and parameter. We’ll finish the series with how this all moves over time.

Density

I have a die, dice?, whatever, and I toss it onto the table. It lands on 6. I ask the room, what number will it land on the next toss? The nerd in the back yells 3.5!

Shut up nerd! what’s he talking about though?

Any next toss will be any possibility of 1, 2, 3, 4, 5, or 6. The average, when you add them up, dividing by the count of numbers, you do get 3.5. No toss will ever get us a 3.5 landing, however. This idea should be taken seriously when considering freight rate data too. That golden rate is just as theoretical a place to anchor.

The dice never lands on 3.5. Each new toss has the same 1/6 chance but if you toss it enough times the calculated average of results will settle here. A good visualization lies below.

Which brings us to density. If I roll a 6 and a 3 my average is 4.5. If the next is a 1 I’ve plummeted to 3.3 average, yet my spread has run the whole range of results. If the results were 1, 1, and 1 the reality may seem different between the two. The reality is they just don’t have enough observations to help create an anchor that doesn’t depend on sight.

To tie with freight, we’ll attach a lane rate per dice throw. Notice how results change amongst the averages as you try to compare them in the right column.

Each company has different results per throw, giving their own average. Then, if you take the average of each company/dataset over all throws, they’re all damn close. MATHGIC.

The lesson here is a smaller pool of results can offer variability, and when you compare two small pools together, it can muddy the waters further. Clarity, then, as with all things lie with experience and time.

Similarly, there’s a marginal return after enough time or result. Another key point in debating who has the most density to how one should use internal versus external data thanks to the Law of Large Numbers exampled earlier.

With more information comes greater confidence until meeting a point of diminishing return. Knowing the long term average of 3.5 doesn’t tell me to try 3 or 4 every time, rather, it keeps me from throwing down on 6’s in perpetuity. Each new throw is the same 1/6 chance.

The example above takes results on some fake localish runs over weeks. This is half a year’s worth. Series 1 moves it every week. Series 2 once every other month. Their results are near identical. How says the nerd from series 1, “I HAVE MORE DATA”. Chill out, my guy, it was a Carson to Rancho, FCFS 10K each time.

Conversely, if the competition, provider, or carrier only has sufficient coverage within the new series 2 above, but you’re running with series 1, who has better sense of that market? Some of these answers are more rhetorical or hard to chase down, but positioning and real estate chant the same mantra - location, location, location.

Trusting enough information and being honest about where there are strengths and misses creates more sound operation than tweaking anything to death or insisting if one drinks enough salt water they’ll inevitably quench their thirst.

Parameter

Let’s play a different game called guess the rate. I give the lane and you blurt out the rate.

Pull: Los Angeles, CA to Dallas, TX.

Shoot: $3,200!

Once the TI-83 is holstered, we go to the second round. This round is called how committed are you? Something fun is sequentially added to the shipment and you decide if you’re still committed.

It’s 45,100 of beverages!

Yea.. sure, scale on site right?

It picks up Thursday!

S-s-omeone will want to get home and hold ove-.

at 6am.

I said $3,200?

These things don’t shout at you on the dots and plots, but for anyone who has operated around flurries of daily volumes, they can recognize why variances exist in any pool or history of data. Below are real results for Chicago to Atlanta via DAT’s Ken Adamo by contributor and their density.

My lines and scribblings are added to provide operator vision to these variances. If it were merely the case the more volume you added, the better or worse the rate, these bubbles would have a much more orderly look to them.

Instead, realities of each line of business muddle against others going in the same direction and of similar grouping to offer our average. When it comes to a local California run, any average rate cuts through too much cake to be exacting, but when taken together against the unique variables of a shipment, the average and range can help with better positioning.

This applies for the sparse and little historied results too, they just rely a lot more on risk tolerance. Unfortunately, there’s little ability for nuance as what you see is what you get. It’s much harder to sense if witnessing a trend or tail event showing up in the data.

Three loads moved in the last 6 months from Perry, FL to Herkimer, NY may seem like a great fit when my search is Tallahassee, FL to Utica, NY, except they were all class 9, needing placards but technically non-hazmat. Are they a better guide than the Lakeland, FL to Syracuse runs of groceries going every other week? You’ll never have the insights of referential data to help explain these differences, but it doesn’t stop any organization from looking inward to view how well the apples are being sorted with or next to the oranges.