July 10, 2007
Follow me on Twitter

Nielsen is wrong, but not for the reasons you think

Jakob, being Jakob, stirred up a hornet's nest with an essay recommended that experts avoid blogging. Jill looks askance, and Scoble is fighting mad.

The problem here isn't the recommendation, or Nielsen's trademark of over-generalizing in order to Make Good Enemies. The problem is that Nielsen is talking about a formal model of expert blogging, and that model is almost certainly wrong.

Nielsen posits a population of 1000 topical bloggers whose ability and expertise are normally distributed.

Let's assume that a given writer's posting quality is normally distributed, with a mean representing that person's level of expertise and a standard deviation 3 times as large as the SD for expertise among people. I don't know what the actual number is, so this is just a rough estimate. But it's reasonable to assume that posting quality is more variable than expertise for several reasons:

  • Sometimes people toss off a posting in a minute. Other times they spend hours.
  • Sometimes a writer happens to know a lot about the topic at hand, possibly because they've just spent several months working on that exact problem. Other times people know nothing--which doesn't keep them from voicing their opinions :-)
  • Sometimes people are lucky and get a blinding insight. Other times they post more out of duty than anything else.

For some reason or other, Nielsen runs a simulation of this model to discover that the very best blogger sometimes posts things that are below average. There's no need to do a simulation; the statistical problem can surely be solved in closed form. But we can do well enough on the back of an envelope.

  1. Of expert's ten posts, we know that (about) 4 are likely to be a standard deviation wide of expert's expected mean.
  2. That means two are likely to be a standard deviation worse.
  3. Since the standard deviation of post quality is three times the standard deviation of expertise translates to three standard deviations worse in terms of base expertise.
  4. So, if the world's top expert is three standard deviations better than the average pundit, one of those ten posts is likely to fall somewhere around average.

So, sure, if these are the rules, your expert (who is maybe three standard deviations better than average) is going to look very, very average in one post out of ten.

The factor of three is arbitrary. It's plucked out of thin air. And I don't believe it: writers are not this inconsistent in other fields. Newspaper reporters working on deadline face similar pressures: did that mean Red Smith or Peter Gammons went to print about three times a month with a story that was below average -- not their average, but the average for all sportswriters? Not all Hemingway is equally good, but the worst of Hemingway's seven novels (To Have and Have Not) is surely better than the average novel of 1937.

Expertise doesn't work this way. You don't hire an expert for a single reason, any more than the Red Sox hire "a ballplayer". Sometimes you need one judgment, but it's got to be the best possible judgment — just as sometimes you need to have a slugger on the bench who can pinch-hit for your wiry shortstop in the bottom of the ninth with a man on third and one out. Sometimes you need reliable, consistent good sense, just as sometimes you need a utility infielder. Sometimes, you need a chance at an astonishing, industry-changing insight. Sometimes, you need the contacts. Sometimes, you want the expert to tell you what everyone else knows. And sometimes you want a fourth for golf.

For many definitions of expertise, you want to demonstrate consistency, breadth, and awareness. Blogs are good for that.

Nielsen's definition of expertise is pernicious. He writes:

We can measure expertise as some combination of intelligence, education, experience, correct methodology, professionalism (say, avoiding profanities and politics), and willingness to be frank.

Nearly every clause in this catalog is mistaken. I don’t care if my expert is well educated or intelligent, any more than I care that they are good looking: experts are supposed to provide the best possible answer to questions within the ambit of their expertise. If a reluctance to be frank impedes giving me the best possible answer, the expert is stealing from his employer.

And while interjecting politics where it doesn't belong is unprofessional, we all live in the world. Sometimes, an expert needs to answer the question: which side are you on? Sometimes, not to answer is a very bad answer.

Nielsen’s model entirely overlooks consistent excellence. We don’t subscribe to an expert's weblog because of occasional flashes of brilliance or hilarity. We subscribe because, day in and day out, the expert tells us what we need to know. Digg and Reddit will find the spectacular post-of-the-moment whenever we like; expertise is for keeping us informed all the time.

Nielsen's argument rests on a rhetorical convention: he redefines weblog to mean "a bad weblog" and contrasts his own practice with a straw man. Nielsen's alert box is a weblog; it merely happens to have posts that run a little long, and he writes a little less frequently, than is typical of some other bloggers. The crux of Nielsen's essay is the contention that

Blog postings will always be commodity content: there's a limit to the value you can provide with a short comment.

Instead, Nielsen urges people to write well and at suitable length. But that’s obviously what any writer — including weblog writers — should do. And it’s what they do today. Sometimes, demonstrating your expertise requires an extended argument. Sometimes, a sentence or two will do the job nicely. Sometimes, you just need an equation.