Skip to the six decisions
Ghil · Labs Open a model

Freight decision models · 06

Freight cost is
either explained
or excused.

Six working models. Not slides. Each one takes a moment where a logistics team stalls on a decision and opens it up — every assumption on screen, and the ways the model can be wrong written down before the answer.

Start with the first decision How they're built

// no install · no login · runs in your browser

Case 04 · Budget variance Illustrative
Over plan+$6.4M
Explained$6.4M
Residual$0
Factors6
Owners assigned6 / 6

// sales blames volume, logistics blames fuel,
// finance says it's neither.
// force the residual to zero and the argument ends.

01  Starting point

Begin with the question, not the software.

02  Assumptions

Every input is visible on screen and editable.

03  Limits

Where the model breaks is stated before the result.

Index of decisions

Already sitting
on your desk

All six companies are invented. The situations are real ones, the figures are illustrative, and the model behind each is fully interactive. None of it is presented as a client result.

Northstar Appliance Parts · fictional parts distributor

“Four hundred shipments. Do we really need four hundred decisions?”

They were hitting their cut-offs. That wasn't the problem. The problem was tendering near-identical orders one at a time, and nobody could say which ones could have shared a truck without breaking a delivery window, an equipment limit, or a driver-hours rule.

The model searches stop count, fill rate, mileage and service constraints together. It does not promise the optimal answer — vehicle routing is NP-hard, and anyone who tells you otherwise is selling something. What it shows precisely is how far the current structure sits from a better one.

  • Consolidation
  • LTL → FTL
  • HOS
  • 400 orders

Open the modelPlain-language guide

HarborCare Medical Supply · fictional wholesaler

“Do we need an Atlanta warehouse to promise two-day delivery?”

Sales saw a Southeast gap. Operations had to separate geography from carrier performance before anyone signed a lease. The two problems look identical on a service report and have completely different fixes — one costs a building, the other costs a conversation.

  • Coverage
  • Service promise

Open the modelGuide

Arbor Home & Garden · fictional seasonal-goods company

“Freight says five DCs. Finance says three. Which answer counts the whole cost?”

A fifth site cuts outbound miles. It also duplicates inventory and adds fixed cost. Until both effects land on the same page, the meeting repeats itself every quarter with the same two numbers and no decision.

  • Network TCO
  • DC count
  • Inventory

Open the modelGuide

BluePeak Nutrition · fictional food manufacturer

“Freight is $6.4M over plan. Is growth really the whole story?”

Volume, mix, network, rate, fuel and accessorial were blended into one number. Which meant every team could point at another one, and the review never converged on anything a person could own.

  • PVM bridge
  • Accountability

Open the modelGuide

Meridian Electronics Distribution · fictional distributor

“Revenue is growing. Why is distribution profit still falling?”

Small rush orders, detention, reconsignment. None of it appears in a customer margin report, and allocating cost as a flat percentage of revenue guarantees it never will. The accounts that look best are often the ones paying least for what they use.

  • Cost to serve
  • Customer P&L

Open the modelGuide

Redwood Consumer Products · fictional shipper

“How much capacity should we commit before the next disruption?”

The cheapest calm-market mix is the first one to fail when tender rejection, spot inflation and a seasonal volume spike arrive in the same month. Cheap and durable are different answers to a question people usually ask only once.

  • Risk
  • Dedicated / Spot
  • Monte Carlo

Open the modelGuide

Method

Why you should
distrust this, carefully.

A model that hides its assumptions is a slide with animation. Every number here comes from inputs you can see and change, and each tool opens with what it cannot do before it shows you what it can.

Distances are great-circle estimates with a circuity factor, not routed road miles. Consolidation is a heuristic, because the underlying problem is NP-hard. Cost allocations depend on drivers you have to supply from your own data. None of that makes the output useless — it makes it checkable, which is a different and more useful property.

Use these to frame a pilot, structure a sourcing event, or take a better question into a finance review. Not to sign anything.

  1. 01

    Inputs are visible

    Rates, volumes, constraints and service targets sit on screen. Defaults exist so the page loads, not so you trust them.

  2. 02

    Limits come first

    Each tool opens with its accuracy band and its failure modes, above the results rather than in a footnote.

  3. 03

    Relative, not absolute

    These compare scenarios against each other. They do not forecast next year's freight spend, and treating them that way will burn you.

  4. 04

    A person still decides

    Contracts, capital, headcount and customer promises need an owner. The model narrows the question; it does not answer it.

Inside a model

Reading surfaces stay light.
Working surfaces go dark.

Pages you read and panels you operate are told apart by value, not decoration. Accent colour only ever carries meaning — green is improvement, red is loss, amber is an assumption.

Total linehaul
$2.41M▲ 8.2% vs plan
Cost / order
$41.07▼ 6.1% vs base
Load fill
78.4%▲ 12.9 pt
Out-of-route
11.2%± 0.0
Loads
184▼ 216 vs LTL
Base linehaul$1.98M
Detour / OOR$0.22M
Accessorial$0.13M
Spot premium$0.08M

Six ways in

Which one is
waiting on you?

Pick the story closest to yours. Two minutes to understand the situation, then the model is right there with nothing to set up.

Back to the six decisions

Elsewhere in the system

Back-office decision models · 04

Four working back-office decision models for contract review, invoice close, vendor risk, and expense audit.

Explore Back-office models →