Most freight forwarders I talk to have the same relationship with load planning: a warehouse supervisor with fifteen years of instinct, a tape measure, and a spreadsheet that someone built in 2019 and nobody fully understands. It works. Until the supervisor takes two weeks off, or the customer sends a pallet list with 340 SKUs and asks for a container count by Thursday.
3D load planning software sits exactly in that gap. It is not magic, it is not AI in any meaningful sense, and it will not fix a broken sales process. But it does a small number of things very well, and those things map directly to money. Here is what I have seen actually change, and what to ignore.
What the software actually solves
Four problems. That is the honest list.
Cube utilisation. This is the headline and it is real. When a human plans a 40ft HC by eye, the failure mode is almost always conservative: leave a gap, round up, book the extra unit. Software does not get nervous. It packs to the geometry. On mixed-SKU consignments with irregular carton sizes, a conservative improvement range is a few percentage points of volume per load, occasionally more when the cargo mix is ugly. On a lane running 20 containers a month, even a modest gain removes whole units from the schedule over a quarter. Do that arithmetic on your own lane before you believe anyone else’s case study.
Stacking rules. Max stack weight per carton, fragile-on-top, no-stack flags, orientation locks (this side up), temperature separation, dangerous goods segregation. A human holds three or four of these rules in their head reliably. Software holds forty. The value is not speed, it is that rule number 37 does not get quietly dropped at 17:40 on a Friday.
Weight distribution. Axle compliance on road freight, centre of gravity on containers, front-to-back balance. This is the one that stops being a cost topic and becomes a liability topic. A load that cubes out beautifully and puts 62% of the mass over the rear axle is not a good plan, it is a fine waiting to happen. Decent tools flag it before the truck moves.
Load sequence. Multi-drop is where the geometry gets genuinely hard. Drop 3 cargo cannot be buried behind drop 1 cargo. Humans solve this by loading in reverse drop order and accepting whatever cube loss that causes. Software solves it as a constraint inside the optimisation, which means you keep more of the cube while still unloading in order. On distribution work with four or more stops, this is often the single biggest win, bigger than raw cube.
| The problem | What the software does | What stays human |
| Cube utilisation | Packs to geometry, no conservative padding | Deciding whether the extra squeeze is worth the handling time |
| Stacking rules | Enforces every flag, every time | Knowing which supplier lies about carton strength |
| Weight distribution | Axle and CoG compliance checks pre-load | Local enforcement reality on the lane |
| Load sequence | Multi-drop order as a hard constraint | Customer relationships when drop 2 calls to reschedule |
| Quoting speed | Container count in minutes, with a visual to attach | The price itself, and whether you want the freight |
What it does not solve
This is the part vendors skip, so let me be blunt about it.
- Bad data in. If your customer’s carton dimensions are wrong by 3cm, the plan is wrong. Garbage dimensions produce a confident, beautifully rendered, useless 3D picture. Most pilot failures I have seen were data failures, not software failures.
- Warehouse behaviour. The plan says carton 214 goes in position 7. The loader puts it wherever is convenient. Software does not supervise. If you cannot get the floor to follow the plan, you have bought a quoting tool, not a loading tool, and you should price the decision accordingly.
- Pricing. Better cube tells you your cost went down. It does not tell you what to charge. That is a separate discipline, and if your margin logic is fuzzy, start with the 3PL pricing calculator before you spend anything on planning software.
- Winning the freight. A load plan does not generate a customer. It helps you convert one who is already talking to you. The sales problem sits upstream and stays upstream.
- Weird cargo. Rolls, drums, non-cuboid machinery, anything that flexes. Most engines model boxes and pallets well and everything else approximately. Test yours specifically if this is your bread and butter.
How to pilot on one lane in two weeks
One lane. Not the business. One lane, one customer, one loading bay. Anything wider and you will be arguing about edge cases for three months.
Days 1 to 3: pick the lane and pull the baseline. Choose the lane that annoys you most, ideally mixed-SKU, ideally multi-drop. Pull the last 20 loads: units used, declared volume, declared weight, any damage claims, any re-loads. Write the baseline down before you touch the software. Without a written baseline you will have no argument at the end, only vibes.
Days 4 to 6: clean the master data. Dimensions and weights for every SKU on that lane, plus stacking flags. This is the boring part and it is where the pilot is won. Budget more time than you think. If the customer cannot supply clean dimensions, that finding alone is worth the pilot.
Days 7 to 9: replan history. Run the last 20 loads through the tool as if they were new. Compare plan against what actually shipped. You now have a delta that cost you nothing in operational risk. If the delta is close to zero, stop here and keep your money.
Days 10 to 12: plan live, load live. Three to five real loads. Print the plan, hand it to the floor, watch. Record where reality diverged and why. Count the divergences that were the software’s fault versus the loader’s fault versus the data’s fault.
Days 13 to 14: decide. Write one page: baseline, delta, failure modes, cost per month, break-even in loads. If break-even is more than about a quarter of your monthly volume on that lane, it is not a yes yet.
The 7 questions to ask a vendor
- Which constraints are hard and which are soft? Ask them to name the ones the engine will silently relax to find a solution. Every engine relaxes something.
- How does data get in and out? CSV import is table stakes. Ask specifically about an API, and about whether your TMS integration is a product or a project with an invoice attached.
- What happens to the plan when one pallet is added at 16:00? Full replan, or incremental adjustment? The answer tells you whether the tool survives real operations.
- Can the loading crew read the output on the floor? Phone, tablet, printed sheet. A plan that only exists on a desktop in the office is a quoting tool.
- Who owns the data and where does it live? If you plan customer cargo, this matters contractually. Get it in writing, not in a sales call.
- What does the exit look like? Can you export your SKU master and your plan history in an open format on the day you leave? If not, you are renting your own data back.
- Show me a failed plan. Ask for cargo their engine handles badly. A vendor who claims none is either new or lying, and both are your problem.
Day-pass versus subscription, in general terms
Two pricing shapes dominate, and they suit different businesses.
Day pass or per-use. You pay for a window of access or a batch of plans. Good for project cargo, seasonal peaks, tender season, and for forwarders who plan loads occasionally but need to look sharp when they do. The cost tracks the work, which makes it easy to defend internally and easy to bill on to a customer as a disbursement. It gets expensive fast if usage becomes daily, and it tends to discourage the experimentation that makes the tool valuable, because every click feels like a meter running.
Subscription per seat or per site. Fixed monthly, unlimited or high-cap planning. Good once planning is routine. The real benefit is behavioural: when usage is free at the margin, your team replans, tries variants, and uses the tool for quoting as well as loading. That is where the compounding value sits. The risk is obvious, you pay in quiet months, and seat-based pricing quietly punishes you for training a second planner.
My rule of thumb: start on per-use through the pilot, switch to subscription only when you have three consecutive months of real usage data. Vendors will push the annual deal during the pilot because that is when your enthusiasm peaks. Enthusiasm is not a usage forecast.
What I would actually do
If you are a forwarder or a 3PL with steady mixed-cargo lanes, run the two-week pilot. The downside is a few days of admin, the upside is a permanent cost line reduced and a visual you can attach to every quote. That second point is underrated: a customer comparing three quotes remembers the one with the 3D plan in it.
If your bottleneck is not load efficiency but the fact that you do not have enough e-commerce shippers to quote in the first place, this software is the wrong purchase this quarter. Fix the top of the funnel first. Verified contact data for the lanes you actually serve, like the UK e-commerce database or the DACH database, will move your P&L faster than three percentage points of cube on a lane you only run twice a month.
Better loading makes you more profitable on the freight you have. Better outreach gets you more freight. Know which problem you are solving before you sign anything.
How this article was made — openly. This post was written by artificial intelligence and published automatically, without prior human editing. We disclose this under Article 50 of the EU AI Act. It reflects our practical experience in B2B logistics, but it is general information — not professional, legal or financial advice. Verify anything you plan to act on. Full terms.
Leave a Reply