Logistics Optimization

What data do you need before you optimize a transport network?

November 19, 2026

Access to data is the most common first hurdle for optimization projects. For last-mile operations, this looks like access to the systems where you data is held, or the data isn't being collected. Some companies have collected the data from a "we can figure this out later" perspective, and now that it's "later", the data isn't in the right format, or doesn't work together.

Optimization software can only show you savings that exist in the data you give it. If you're planning a carrier tender, a DC consolidation or a fleet redesign in the next year, getting the inputs right is the first piece of work to complete.

This guide covers what to collect, how much history you need, and how to scope the first analysis so it produces answers in weeks rather than quarters.

The core data checklist

1. A year of daily order data

Order history is the foundation of any network model. You want at least 12 months at a daily level, not weekly or monthly totals. Daily data shows the peaks, the quiet days and the day-of-week patterns that drive how many vehicles you actually need. Monthly averages hide all of that.

For each order, aim to capture:

  • Order date and requested delivery date or window
  • Origin (DC, plant or depot) and destination (customer site, with a full address or coordinates)
  • Volume in the units your vehicles fill up on: weight, cubic volume, pallets, cartons or a mix
  • Product or temperature class, if it affects which vehicles can carry it
  • Customer or channel type (for example, B2B retail versus direct-to-consumer)

Why a full year? Because seasonality changes the answer. A network that looks over-resourced in February can be badly short in December. If you only have six months, a model will still run, but you'll be making long-term decisions on half the picture.

2. Carrier rate cards by equipment type

If you use third-party carriers, the model needs to know what each move actually costs. That means current rate cards, broken down by:

  • Equipment type (rigid, semi, refrigerated, van, B-double and so on)
  • Lane, zone or distance band
  • Rate basis: per trip, per pallet, per kilometer, per hour or a combination
  • Fixed charges and accessorials such as fuel levies, waiting time, tolls, tail-lifts and multi-drop fees

Equipment type matters more than most teams expect. The cheapest network on paper often assumes larger vehicles that some customer sites can't take. Linking rates to equipment lets the model respect site access while still finding consolidation opportunities.

3. Price ladders

A price ladder shows how a carrier's rate changes as volume, distance or vehicle size steps up. It's what lets you test questions like "what happens if we move two lanes from a part-load to a full-load rate?" or "at what point does a dedicated vehicle beat paying per pallet?"

__wf_reserved_inherit

4. Historical performance data

The final piece is what actually happened. This lets you benchmark the current network before you start changing it, and check that the model reflects reality. Useful sources include:

  • Actual routes driven, with stop sequences and timestamps
  • Carrier invoices or freight spend by lane or region
  • On-time performance and failed or redelivered orders
  • Dock times, loading times and time spent waiting at customer sites
  • Vehicle counts and utilization by day

Don't skip the timings. In network analysis, time at the dock and at the customer often explains cost differences better than distance does. A consolidation that adds a few kilometers but removes hours of loading delay can still be the cheaper option.

Nice-to-have data

These inputs aren't essential for a first pass, but they make the results more precise:

  • Site constraints including receiving hours, vehicle size limits, dock counts, and unloading equipment
  • Driver rules including shift lengths, break requirements and award or contract conditions
  • Planned changes such as new sites, lost or won customers, or expected volume growth
  • Emissions factors including vehicle fuel types and efficiency if you also want to report the carbon impact of each scenario

Start with one DC, not the whole country

A national model is slow to assemble, slow to validate and hard to trust in its first version. One missing data feed from one region holds up the whole project.

A better approach is to start with a single DC you know well, then add one or two contrasting sites. For example, a dense metro DC plus one in the north and one in the south of your network. Choose sites that differ in density, distance to customers or carrier mix.

This gives you three things:

  1. Speed, three sites' data can be gathered and cleaned in a fraction of the time
  2. A trustworthy baseline from which your team can sense-check the results against sites they know intimately
  3. A pattern you can extend if the savings drivers are similar across three very different sites, you can estimate the national opportunity with more confidence, then roll the model out site by site.

Common data gaps, and what to do about them

Addresses that don't geocode cleanly

Flag low-confidence locations rather than guessing. A handful of misplaced customers can distort a whole region.

Volumes in the wrong unit

If orders are recorded in cases but trucks fill up on pallets, agree a conversion before modeling.

Rate cards that don't match invoices

The gap between contracted and invoiced cost is often a finding in itself.

Missing timestamps

If you don't have actual arrival and departure times, use planned service times as a starting point and refine later.

__wf_reserved_inherit

Getting started

Adiona's Fleet Simulator models depots, fleets, carriers and service rules side by side, so you can test network changes before you commit to them. Talk to our team about scoping a first analysis, or try Adiona free for 14 days.