Stop building strategy from the inside

11 firms out of 8,430, why a third of feedback makes performance worse, and the pilot I ran in the wrong branch.

In this week’s edition

1. Thematic Strategy: Building a strategy that gets stronger as the world changes

2. The Art of Leadership: How to give feedback that actually helps people

3. Operating at Scale: How to ensure a rollout lands everywhere, not just the pilot

Read time: 7 minutes

---

_{THEMATIC STRATEGY}

Building a strategy that gets stronger as the world changes

There are dozens of strategy frameworks that promise to help you outperform your competitors. For my PhD at ESCP Business School I went looking for the companies that had actually managed it, screening 8,430 firms listed on the US stock market across 81 industries for a dominant share of their industry’s profits over five consecutive years.

Just 11 firms cleared the threshold, 0.13% of the sample.

What those 11 have in common is the framework I now call Thematic Strategy. They were not better informed than the companies they beat, and their competitors paid attention to the same external changes. What separated them is that they committed the business to three or four of those changes and stayed with them for a decade.

Which reduces to a single question you can ask about your own strategy:

Have we aligned our positioning and capabilities to leverage the most important drivers of transformation?

You answer it with three questions, and the order is the framework:

1. Foresight. What long-term changes outside this business could open an opportunity or become a threat?

2. Positioning. Given those changes, how do you leverage them to differentiate yourself from your competitors?

3. Capabilities. What do you have to build to deliver on that position?

Most firms have a strategy, and most of them start by improving the firm from the inside. The best firms do the opposite.

Remember:

1. Start outside the firm

2. Align with massive, fast-growing transformations

3. Focus relentlessly on riding the wave

Then put this to your team:

Name the three changes outside this business that we are organised around, and the last thing we funded because of one of them.

The changes will come easily. If the funding does not, you have a foresight process that produces reports.

---

_{THE ART OF LEADERSHIP}

How to give feedback that actually helps people

In 1996, Avraham Kluger and Angelo DeNisi published a meta-analysis in Psychological Bulletin covering 23,663 observations. They wanted to know whether telling people how they were doing improved how they did.

On average, it did. But that average hides the finding that made the paper famous. In over a third of cases, the feedback made performance worse.

Feedback works by redirecting someone’s attention. Where attention is redirected to the task, performance improves because the person now has something specific to work on. Where it directs attention to the self, performance falls, because they start managing how they are seen instead of how they are doing.

Which gives you a check that takes about four seconds. Write down the main thing you intend to say, then find its grammatical subject. Is it the work, or is it them?

Here’s two examples:

1. You need to be more commercial at this level. The subject is the person. It names no piece of work, so they spend the following weeks trying to look commercial.

2. The paper didn’t say what we would stop doing to fund this, and that is the first thing the committee will ask. The subject is the work, and the work can be fixed.

The theory does not spare praise, either. You’re a star points at the person as squarely as you’re not detail-oriented does, and it directs attention to the same place. Encouragement that never names anything specific changes almost nothing.

Being specific requires you to have read the paper or watched the meeting. General feedback about a person’s character is what you give when you have not done the work of looking, and it feels kinder while leaving them with nothing to work on.

Sometimes the thing you have to raise really is about the person. Then anchor it to one occasion. Not you’re too defensive in exec meetings, but:

When Sarah pushed on the timeline on Tuesday, you answered the challenge rather than the question, and she stopped asking.

Name the occasion, describe what happened, then stop talking and let them answer.

And when you find that you cannot be specific, that is itself the finding. It means you are not ready to give feedback, and the conversation is premature.

This is a meta-analysis from 1996, drawn heavily from controlled settings rather than executive committees. It will not tell you which of your own conversations sit in the damaging third, and checking the subject of a sentence will not make you right about the person. It will stop you saying the worst version of what you were going to say.

---

_{OPERATING AT SCALE}

How to ensure a rollout lands everywhere, not just the pilot

In 2007 I was given 71 Halifax branches to run, leading around 1,300 people. I was 31, and it was the biggest thing anyone had handed me.

What I saw quickly was that the branch teams had almost nothing by way of a performance process. A job description, sometimes a target, barely any feedback, and no development discussion at all. That struck me as an obvious gap, and one I could close.

So I downloaded our standard forms and asked the manager of my largest branch to try them with her team of 55. It worked. She liked it, her people engaged with it, and I had my proof. I planned a proper launch and delivered it across all 71.

Then I did my usual round of branch visits, and I could not find it running in a single branch.

The feedback I got was harsh, and it was right. What works for a branch with a team of 55 does not work for a branch with six part-time colleagues and a manager who also serves customers. And the harder point: not everyone wants a development plan. Some of those people had worked in that branch longer than I had been alive. They wanted to come to work, do the job well, and go home.

I never did the arithmetic. Across 71 branches and 1,300 people the average team is about eighteen, and I piloted in a branch three times that size. I had not chosen a representative site, I had chosen the strongest one, because I wanted the test to work.

What I have used since is two sites and one question, and both are settled before the pilot starts.

Two sites means running the pilot in your strongest unit and your hardest one simultaneously. The strong one tells you whether the thing can work at all, which is worth knowing and is all I learned in 2007. The hard one tells you whether it can work everywhere.

One question is what you ask the manager of the hard site before either pilot starts, in these words:

If this were the only branch, what would you have to change to make this work?

Then listen. If it is something you need them to keep doing, change the process and leave the branch alone. If they cannot name anything at all, they have not understood what you are asking of them, and one conversation has made that clear.

I missed one thing in 2007 that this does not catch. The method tells you whether a process survives. Whether anyone wants it is a separate question, and a process that only runs while people are enthusiastic will stop when the enthusiasm does.

The other thing I got wrong. I had planned a launch, which is an event you run once. A rollout is months of going back to the sites least likely to have bothered.