How to reach any conclusion you like

Some things are not politically correct to say, so lots of stats are presented in support of rash generalisations that deny those things. Often, they’re contrived to support a pre-conceived argument, rather than constructed to shed light on the matter.

That the environment is not the biggest crisis of our time is one case in point, which I’ve discussed before. You’ll see plenty charts that show, for example, what a 20-foot rise in sea levels would look like on a map. Or how hurricanes have increased in frequency or intensity. A close look at the data, however, and you’ll see signs of contrivances, such as selecting end-points for data series to support the pre-conceived argument, or presenting speculative conclusions as factual data. And how the environment manages to recover from catastrophes — such as volcanoes, meteor strikes, earthquakes, nuclear testing — when it really is a fragile system, perched precariously atop an unstable equilibrium, and sensitive to every little belch and burp of human life, is left as an exercise for the reader. How humanity came to be so prosperous and well-fed if our production methods are so stupidly unsustainable, ditto.

But let’s look at a very different example. Let’s consider Microsoft’s success. It isn’t politically correct to suggest that the monster from Redmond occasionally makes smart business decisions. How it came to be a large, successful company that made more millionaires than most is left as an exercise for the reader. How a a billion PCs came to be installed around the world in 20 years, a majority of them running Microsoft’s terrible products, ditto.

Now granted, I’m not convinced that its bid for Yahoo! was a smart business decision. It smacks of desperation. It has spent ten years trying to find a revenue stream to which the Windows-and-Office cash baton can be passed, but no convincing candidate has yet appeared. In its search, it has a habit of buying up second-best players in a market segment, and then swamping them with Microsoft branding until they’re never heard from again. My own prediction on Microsoft is that ten years hence it will be known as a very good gaming company.

But take a look at this analysis, by Max Freiert, highlighted recently by the folks over at Junk Charts, who spend their lives debunking — often in the most entertaining fashion — statistics abuse by companies, governments and the media.

It shows this chart:

Then Freiert assigns value only to new customers, which makes the $45 billion deal appear like a valuation per customer of $1 200. This is patently ridiculous. But I suspect it is designed to look ridiculous from the outset. That’s why it does not take into account that Microsoft might derive additional value from customers in the overlap area, or might offer additional value to its own customers, and that these might all lead to higher revenue for Microsoft.

If we redo the calculation, but with the (equally arbitrary, but more realistic) assumption that Yahoo! customers who also use Microsoft properties are worth half as much as new customers who didn’t use Microsoft properties before, and customers who don’t use Yahoo! at all have no additional value at all, we get a per-customer valuation of less than $200. And if you postulate that the deal might result in the ability to offer new revenue-generating value to customers that didn’t use Yahoo! properties before, at the arbitrarily-selected rate of one-third of the value of new customers from Yahoo!, you get $150 per customer, overall. Does that sound more reasonable?

But what about the new customer numbers? Does the chart above really reflect visually how many new customers Microsoft gets? How small is that top slice really? The guys at Junk Charts, instead of mentally calculating the area as a reader is meant to do, simply recast the data as a bar chart, as a statistician might do. It’s not pretty, but it does make the point: actually the growth in customer numbers is pretty darn decent.

Here’s their reworked chart:

Check out the last line of the table in Freiert’s analysis, and you’ll see that his result is the same: new customer growth will be a substantial 31%. And if you look in the next column, you’ll see a 89% projected rise in page impressions. In fact, Freiert makes a big deal of this growth in Microsoft-owned traffic later in the analysis. But if only the new customers had any value, why the massive discrepancy? This merely confirms that assigning value only to those new customers is a ludicrous assumption, apparently designed to support a pre-conceived conclusion.

The growth in customer numbers is larger than Freiert’s pie chart suggests, and the actual price paid for potential new revenue is much, much lower than the $1 200 he puts in the headline. None of this shows that the Yahoo! acquisition really was a smart acquisition, but it does show that at least part of Freiert’s neatly contrived argument for why it may have been a daft acquisition holds no water.

It was a daft acquisition for other reasons. I could create a chart to prove it, but I fear you wouldn’t believe me.