It’s Day 90. It’s gone badly. What happened?

"The line between disorder and order lies in coordination". - Sun Tzu

Hi from Ireland. Derek here.

Welcome to the very first Day 90. Thanks for reading. I’ll try not to waste your time.

Here’s the idea. I want to help you de-risk your digitisation and automation decisions in multimodal transport and logistics. Integration risk is huge. Dashboard fatigue is everywhere. Solutions that don’t really solve anything are everywhere. And I’d like to work through it with you, as a community. On that note, please reply to the polls in each edition - it helps us build the picture.

A quick word on who’s writing this. I’m Derek O’Sullivan. I live in Fenit, a small fishing village with a deep-water port on the west coast of Ireland, and I started out designing container cranes at Liebherr. These days I watch those same cranes leave my local harbour for every corner of the world. Since then I’ve built a company, bought my share of software, helped another logistics supplier test for product/market fit, and now I work with CaDi Systems on that same problem. I’ve sat on both sides of the table, and I’ll always tell you when that matters.

I’ve spent most of my career designing hardware and software I thought would solve a real problem. Sometimes I got it right. Sometimes I didn’t. The wins and the losses taught me to ask what breaks before I ask what’s possible. That’s what Day 90 is for.

In today's Day 90:

Every failed rollout gets a post-mortem. Almost none gets a pre-mortem.

After the fact, everyone can explain what went wrong. At Felixstowe in 2018, forwarders were doing it within weeks. One said the new system was set up for an export-dominated operation when Felixstowe is import-dominated. Another said four areas passed testing on their own, then failed when run together. In hindsight, both look like things someone could have raised.. (More on that case below.)

So who could have flagged them before go-live, and did anyone ask? Here's a clue. The Loadstar and Raft asked supply chain people about AI this year. Among VP-level and above, 77.5% were upbeat about what it means for their careers. Among individual contributors, 37.5%. Same technology. Very different mood. The people who choose a system and the people who live with it don't see it the same way, and they're not always in the same meeting.

So here's something to try before you sign. It takes 1 hour and costs nothing.

Get frontline staff in the room. I mean a checker, a dispatcher, a yard planner. Not the project sponsor. Then say this:

Give everyone three minutes to write down their answers, quietly, on paper. What they think could happen. The methodology matters. Otherwise the most senior person sets the answer in the first thirty seconds and everyone nods along. Think of an athlete kind of ‘prehabing’.

Then read them out. Then read them out. Patterns show up fast. Those are the questions to take to the vendor.

It's actually called a pre-mortem. A psychologist called Gary Klein wrote it up in Harvard Business Review back in 2007, for a simple reason: people keep their doubts to themselves while a plan is being made. Once the plan has already failed, it's easy to be honest. A pre-mortem gives people permission to be honest while there's still time to do something about it.

Then, with the vendor in front of you, swap the polite questions for one proper one:

A slide can't answer that. A real customer can.

Is it worth the hour?? Gartner surveyed 306 logistics professionals and found that responding to team resistance and building feedback in raised the odds of a transformation succeeding by 62%.

Pushing ahead with a top-down, urgency-led approach cut the odds by 47%. Listening isn't soft. It's the cheap option.

I have the scars to show for not getting the ultimate users involved in the decision. They have to feel part of it, and they have to buy in. It’s the cornerstone of successful change management.

Try it this week, even on a small tool, and hit reply to tell me what came up. I read every reply. Be aware of the dreaded confirmation bias.

Let's look at one that went wrong in public, so you don't have to go through it in private.

In 2018, the Port of Felixstowe in the UK swapped its terminal operating system. That's the software that tells the yard where every box is and what to move next. Hutchison replaced the existing Navis system with its own, called nGen. By the time Seatrade Maritime reported on it on 26 July, the disruption had been running for six or seven weeks.

So what went wrong? According to that report:

And the cost? Vessels "cut and run", leaving containers on board, and cargo was diverted to London Gateway, Liverpool, Southampton and Tilbury. Hauliers made long empty trips. Forwarders lost hundreds of thousands of pounds, and the report says jobs were lost too.

One forwarder put it neatly: "nGen has been successful elsewhere but it was not comparing apples with apples." The harbour master, Ashley Parker, said: "We're getting more used to the system and tweaking it as we develop it."

What I take from it. The report doesn't say what testing or training Felixstowe did, or how the changeover was run, so I won't guess. But every one of those problems turns into a question you can ask before you sign. Ask the person on the yard, not the sponsor:

And if you only ask one thing, make it the Day 90 Question: show me a customer like us at Day 90. What broke, and what did it take to fix?

Durban is about two months into its own Day 90

Durban Gateway Terminal moved to Navis N4 in mid-August. What was planned as a 12-hour changeover turned into more than a month of delays, according to Daily Maverick. Cargo volumes from 14 August to 10 September ran about 30% below the previous four-month average. By 8 September, ships were waiting 5.8 days on average at anchor off Pier 2, and some cargo was rerouted to Cape Town and Port Elizabeth. The terminal said the system had stabilised by 26 August.

But it isn't only an IT story. The same report says only 37 of 110 gantry cranes were reportedly operational, and wind put the waterside on standby in mid-September. That's the thing about new systems: they never land on a clean slate. They land on top of everything else that's already going wrong.

Having begun my career in the crane business, and having spoken to multiple terminal operators both then and more recently, those percentages really really suck.

Day 90 for Durban falls around mid-November. I'll check back in then.

"It is ultimately the forwarder or the broker"

One for your next AI demo. In The Loadstar's interview with WiseTech CEO Zubin Appoo (4 September), the question was who is liable when an AI decision turns out to be wrong. His answer: "It is ultimately the forwarder or the broker."

It's a month old, but it's worth taking into the room with you. If the vendor's agent makes a bad call on your shipment, whose problem is it, and what does the contract say?

Disclosure: I work with CaDi Systems, which integrates with CargoWise.

One question, one click, about ten seconds. And I'm genuinely curious, because the answers will shape what I write next.

Thanks for reading the first one. It means more than you'd think. If you take one thing away, make it this: before you sign, get a frontline person in the room and ask them how it will fail.

Two small asks:

Next Tuesday: the reference call. How to ask a vendor’s customer the questions that get you the truth.

Until then - fair winds and following seas,

Derek