The Crucial Source of Information: People

A couple of days ago we received an enquiry from an executive: his impression was that a competitor was executing the strategy of systematically targeting countries, one at a time, directing a disproportionate level of marketing resource at that country until the firm achieved the No. 1 position in that country, then moving on to the next country.

The shift to the next country necessarily involved the reduction of resource aimed at the first country, so that the resource could be transferred to the next. Presumably the customers and channel partners in the country either wouldn't notice the reduction or, if they did, had become so loyal to the vendor that they wouldn't bother to switch to another vendor.

Was the executive's impression correct, he asked? And if so, how could we counter the strategy?

Led by our director, our initial position was the null hypothesis: the vendor wasn't targetting countries one at a time, but instead behaving consistently across all countries, and over time. Indeed it's very difficult to find evidence that suggests the contrary:

• We forwarded the question to our worldwide counterparts, who are normally assumed to be wise in all matters. They took some time and some pushing to respond. When they replied, they knew of no serial targeting strategy in Europe by the vendor.
• There was no secondary research, no single report, that focussed on the vendor's country-by-country strategy or gave any impression of the timing of such a strategy.
• There was nothing that we could look for in our own financial records that could possibly demonstrate a single vendor exercising a country-by-country strategy. If our results go up or down, it is usually due to a load of market effects and the actions of multiple competitors. To show the movements of a ghost vendor through our financial records seemed impossible.
• We found several news clippings from 2007 that indicated that resellers liked this vendor for its consistency.
• We didn't approach any employee who had previously worked for this vendor, because to do so would be unethical.

A first draft of a 'No Evidence' presentation was forwarded to the original requesting executive. A second version, polishing the English and improving the flow of the argument, but basically maintaining the 'no evidence' position, was produced.

Then suddenly, a contrarian view came through. Our European channels specialist, who had been copied on the request, not by us but by one of our respondents, said he had spoken to a respected external analyst, who said the vendor did, at times, employ a targeted, country-by-country approach.

So much for serendipity. My personal view is that the exernal analyst is so experienced and respected that we should take his word for it. Besides, he has direct contact with the vendor itself.

So the moral of this story is that much of the time in market intelligence, you'll be relying on the judgements of people. Numbers have their place, but too often they are historical—too historical to act upon. The more accurate they are, the older they are likely to be. If you try to work it out all on your own, you're almost bound to miss some important points. So ask for input, almost always.

Rewarding good market selection

I'll be brief tonight, as it's been a long day and I haven't had a chance to think deep thoughts.

I spent the day in the City of London, as guest of one of our investment advisor firms. They were educating us about interest rate swaps and inflation swaps—financial tools that UK pension funds are increasingly employing.

But perhaps the most intriguing nugget emerged from a discussion in the morning: our equity fund managers don't get directly rewarded for good performance—e.g. a financial incentive for exceeding the benchmark for that asset class. The reward they get is the continuation of their contract with our pension fund. But, unless the manager's performance is absolutely appalling, it tends to take at least three years to collect enough data to indicate that the manager should be sacked.

I guess this shows what a random process stock-picking is assumed to be. So, nothing like market selection in the world of business strategy, then ...

The Unmeasurables

Pick a typical action-oriented decision:

1. Whom will you choose as your next partner?
2. Whom will you vote for?
3. Which CD will you buy next?

How are you likely to make that decision? My guess is that no-one would make any of these decisions on entirely numerical criteria. You might have included some numerical factors—e.g. 'height greater than 150cm' or 'jazz band with at least two lead instruments'—but I would imagine that you would also include at least one personal judgement criterion, such as 'personality that I would rate at least 8 out of 10' or 'any new CD by an artists whose previous albums I have been at least 70% satisfied with'.

If you're committing your future or your money to a decision, you wouldn't leave it to entirely numerical criteria. You might let Amazon recommend a CD purchase, but you wouldn't let a computer make your purchase decisions without consulting you first.

This refusal to make personal decisions on entirely numerical criteria makes me highly suspicious of any business decision-making tool which uses only numerical criteria. Would you buy or divest a business on purely numerical criteria? I hope not—there would presumably be many qualitative factors, such as your judgement about the degree of fit with your existing business and the calibre of its management.

But in my job, I see an increasing number of tools being proposed for internal usage which aim to produce a composite number out of a series of purely numerical factors. Some of these tools generate a number, and others produce a 'traffic light' colouration of cells in which the decision to go for red, amber or green is based on a number. Other tools promise to create a 'heatmap', whatever that is. Many of them lhave a passing resemblance to a half-hearted GE multi-factor Matrix.

What they all share is the implicit message that if you can't measure it, it shouldn't be a consideration in the invest/divest decision. They don't trust the client management to supply qualitative judgements on unmeasurable factors (such as 'quality of indirect channels') because they fear that these managers will adjust their input to get the result they require.

And yet proper use of the GE multi-factor matrix should always involve some input criteria which require qualitative judgement. People seem to think they can include what they like in these tools, without realising that their proper construction has some intellectual basis. For instance, the 'Boston Box' arose out of the Experience Curve, though I can't recall the precise connection at the moment. It must be time to finish.

Tomorrow I'll discuss these matrices further, or something similar, or perhaps something quite different.

Make sure you ask the question very clearly.

Q1. What is the minimum number of people you need in a room before it is more likely than not that two of them share a birthday?

The answer is 23. Bypass the next paragraph if you either are not interested in how the answer is calculated or already know the solution.

Without substantially altering the answer, we'll assume the year is 365 days long. We'll focus on the chance of everyone in the room having a different birthday.

• The first person enters the room. The second person has 364 out of 365 chances of having a different birthday to the first person.
  
• The third person has 363 out of 365 chances of having a different birthday to the other two.
  
• And so it goes on. The 23rd person has 343 chances of out of 365 of having a different birthday.
  
• Multiply all these probabilities together to get the probability of all the events happening. You can do it on your calculator.
  
• After 22 people have entered the room, the probability of them all having different birthdays is still greater than 0.5. But multiply that number by 343/365 when the 23rd person enters the room, and now the probability is less than 0.5, so it more likely than not that two of the people in the room share a birthday.
  

Q2. What is the minimum number of people you need in a room before it is more likely than not that one of them shares a birthday with you?

The answer is about 254. Each person who enters the room has 364 chances out of 365 of not having the same birthday as yours. If x is the number of people in the room needed, then we have to solve the equation (364/365)^(x-1) = 0.5.

What intrigues me is that the questions are very similar, yet the answers are very different: 23 versus 254. Indeed the questions are so similar that many people might not detect the nuance. If we put the questions through an automated translator (e.g. Babelfish), and translate them into, let's say, Spanish and back again, we get:

Q1. Whoever is the most minimum number of persons that you need in a quarter before you are more likely that no those two from her part in the birthday?

Q2. Whoever is the most minimum number of persons that you need in a quarter before you are more likely that no that one of the parts in the birthday with you?

Both sentences no longer make sense, although they still differ. Will anyone understand the questions? It's highly unlikely that anyone will come up with 23 and 254.

The moral of the story

I'm not trying to make a point about not giving numerical tasks to foreigners. The key issue is about clarification of the question. English can be a very compact language—small differences in wording yield very different answers. Mechanical translators have a long way to go before they can understand these nuances. (And I have to admit that it is mathematically interesting that the two answers are so different.)