Halfway through the engagement, the partner reviews the team's analysis. The first hypothesis doesn't survive the data: the centre that looks most expensive per call turns out to be the best per customer. The team tests a second idea, looks for the one sentence, and decides how to tell the client that the work is two weeks behind.
Right. Before we look at any charts, tell me where we are. In two sentences, Owen.
The analysis is mostly done, and it doesn't say what we expected. And we're about two weeks behind, because the Glasgow data arrived late.
That's two very different pieces of news. Which one should worry me more?
Honestly, neither. The first is good news, even if it's uncomfortable. The second is in our assumptions. But I'd like to start with the first.
Judith asks for the status before any detail, and Owen gives two facts: what the analysis says, and where the plan is. Leading with the bad news, instead of hiding it after the charts, is what makes a manager trust the rest of the update.
Fine. Then why did the Glasgow data arrive late? I want to understand it before I hear it from Philip.
Glasgow's phone system doesn't record repeat calls. Joanna's team had to check five hundred cases by hand. They did a careful job, but it took an extra week.
And the visits? What did you see that isn't in the data?
In Glasgow, I sat with an agent who took a call about a damaged parcel. She kept the case, and the next morning she called the customer back herself. The customer thanked her by name.
And in Leeds, I listened to a customer explain the same problem to three different people in two days. Nobody did anything wrong. The process just passed him on.
Good. Then nobody did anything wrong, and we say exactly that. Meera, show me what it says.
In the war room, our first hypothesis was that most of the saving comes from closing Bristol and Glasgow and moving the work to Leeds. On cost per call, that still looks right. Glasgow's calls are about twenty-five per cent longer than Leeds, and it's the most expensive centre per call.
So what?
So on cost per call, the answer is Leeds. But then we looked at repeat calls. In Leeds, twenty-two per cent of customers call again about the same problem. In Bristol, nineteen. In Glasgow, eight.
Judith's two words ask Meera to move from a fact to its consequence. A number on its own is not an insight. In consulting reviews, "so what?" is asked again and again until the analysis says something a client can act on.
Eight, against twenty-two. How sure are you of the eight?
It's a sample, not the full system, so I'd say it's between six and ten. But even at ten, it's less than half of Leeds.
Then let's test the hypothesis properly. For everything to move to Leeds, what would have to be true?
Leeds would have to take Glasgow's customers without more repeat calls. And today it can't. If Glasgow's customers get the Leeds repeat rate, a lot of the saving goes on answering the same customers twice.
Instead of arguing whether a hypothesis is right, Judith asks what conditions it needs. If those conditions are false, the hypothesis falls by itself, without anyone having to defend or attack it. It is one of the most useful questions in problem solving.
So the first hypothesis is gone.
Before we bury it, what does it do to the money? Philip started with four million.
On cost per call, everything in Leeds saved about three million. When I add the extra repeat calls, it falls to about one point eight.
So the client's number was four, our first estimate was three, and the real one may be under two.
And we haven't told anyone a number yet. That was the reason for page five.
Exactly. Meera, do you remember what you said in the war room, when I asked what would change your mind?
I said, if one of the smaller centres does something better than Leeds, something Leeds can't easily copy. And it does. In Glasgow, the first agent keeps the case until it's closed.
Then your hypothesis was wrong, and your test was exactly right. That's the job.
Then let me try a second idea, just to test it. What if we did the opposite? Keep Glasgow, and move everything there.
I looked at that. Glasgow has sixty agents and the oldest phone system. It would have to grow to almost three hundred. Nobody thinks that's realistic, including the people in Glasgow.
Fair. So not one centre. Two?
Two looks more interesting. Leeds is the cheapest per call and it's where the head office is. The real question is the second centre: Bristol or Glasgow.
And what does the data say?
On cost, Bristol and Glasgow are close. On customers, Glasgow is clearly better. But there's one more number. HR sent the length of service yesterday. In Bristol, almost half the agents have been there more than five years.
That's exactly what Javier told us at the workshop. If the work leaves Bristol, those people won't move with it.
Javier's warning at the workshop now appears in the data. Owen says so openly. Recognising that a sceptical client was right about something is not weakness; it makes the final recommendation stronger and easier for that person to accept.
So closing Bristol would cost us exactly the people he warned us about.
Unless we plan for it. A transition offer for the best Bristol agents, and enough time to move the knowledge before the work moves. I don't know yet if that's realistic.
Then that's the next thing to find out. We're not recommending anything today. We're deciding what to test next.
Then I'll ask HR for the Bristol numbers by role this week. And Joanna should know her old centre is the reason the first idea fell.
Make sure she gets the credit for pointing us there, and in front of Philip, not in a corridor.
I will.
Now give me the one sentence. Meera, you first.
Our analysis shows that although Glasgow has the highest cost per call, its lower rate of repeat calls means that the cost per customer problem is...
Stop there. Can Philip repeat that to his board?
No. I couldn't repeat it to myself.
That's an honest answer. If you can't say it simply, nobody else will be able to repeat it.
Let me try. The cheapest centre per call isn't the cheapest centre per customer.
Better. Much better. But be careful. That's a finding, not a recommendation. We don't know the answer yet. We know the question was wrong.
Owen's sentence is clear and memorable, and Judith accepts it, but labels it correctly: a finding describes what is true; a recommendation says what to do. Presenting a finding as if it were a decision is a common mistake in the middle of a project.
Then where does it go in the deck?
On the first page of the steering committee pack, as a finding. And the cost of the first hypothesis on the second page, so they see why we changed our minds. Never hide the idea you dropped.
Then I'll redo the first two pages tonight.
Now, the two weeks. How do you want to tell Philip?
As a change request, at the steering committee. Two extra weeks, because the Glasgow data needed a manual sample. It's what our assumption says: if the data arrives late, the plan moves with it.
Good. And write it so it's about the data, not about Joanna's team. They did more work than anyone asked them to.
The delay was caused by the client's data, and the proposal said that delays would move the plan. Judith still asks Owen to describe the cause, not the people. A change request that blames the client's team wins the argument and loses the relationship.
And the telephony option? Philip decides that at the same meeting.
Then we need to be clear about the total. If he buys telephony, that's two more weeks on top. I don't want him to hear "two weeks" twice and feel tricked.
Agreed. One page, both decisions, and the total at the bottom. And price the extra weeks exactly. No round numbers.
One more thing. I promised Javier he'd see the numbers before anyone else. I'm meeting him on Thursday.
Good. He'll like the repeat calls. He won't like what they might mean for Bristol.
I know. That's why I want him to hear it from me, and not on a slide in front of Philip.
What will you say to him first?
That his instinct about customers was right. Before I show him anything about Bristol.
Good. Start with where he was right, and he'll listen to the rest. Start with Bristol, and he won't hear anything after it.
Can I come to the meeting with Javier? I worked on the sample with Joanna's team, so I can answer the detailed questions about how we counted.
Yes, please. He'll trust the numbers more if the person who counted them is in the room.
Then we're done. Meera, good work. Owen, send me the steering committee pack on Monday, and the change request with it. I want to read both before Javier sees anything.