Recursive PMs: Using the Crowd for Cover

In the middle of May I speculated--reasoning from first principles only and with no inside information--about how prediction markets might help the book industry pick winners:

"For some types of books (and authors) prediction markets might be appropriate."

That sparked a note from a reader, tipping me off that:

"Confidentially, something's about to launch very soon in this area."

Sure enough, MediaPredict launched a week later, eliciting this highly positive piece in the 5/21 NYTimes: "Publisher to Let the Public Have a Vote on Book Projects".

Media Predict is soliciting book proposals from agents and the public, and posting pages of them on the site. Traders, who are given $5,000 in fantasy cash, can buy shares based on their guess about whether a particular book proposal is likely to get a deal, or whether Touchstone Books, an imprint of Simon & Schuster, will select it as a finalist in a contest called Project Publish. If either happens within a four-month period, the value of the shares go to $100 apiece; if not, the share price falls to zero.

The devil as they say, is in the details. Traders are voting not on how popular a work could become with the book-buying public but with the binary, intermediated and likely recursive question of whether the editors at Simon and Schuster will select it.

In other words, S&S editors are going to be breathing their own exhaust--now with more butt-covering camouflage. (Which reminds me a little of this.) It's the kind of conversation I've seen some couples get into:

Man (off-handedly and with a smile): What would you like for dinner, honey?
Woman: Oh, whatever you'd prefer is just fine with me!
Man (thinking burgers and beers with the guys): I'd prefer what pleases you, dear.
Woman: Well I don't know. What were you thinking?
Man: I was thinking I'd like to know what you're thinking.

And 'round and 'round and 'round it goes. That's an imperfect and terribly stereotyped analogy, and no, it's not my happy marriage. Well, OK sometimes maybe it is. But let's move on, shall we?   :)

The more abstract version of that analysis is this: The system S&S has set up with MediaPredict has no outside, arbitrating reference point--such as how many copies a particular book would actually sell.

Hey, now there's an idea: ask the market directly about its spending enthusiasm for a given title. (Print-on-demand technology, not to mention digital rights management, render completely moot the objection that only a few select books can be produced. The hurdle revolves around marketing and a legacy business model of annointing a few winners.)

Unfortunately, S&S's Touchstone Books does not appear to take that important dis-intermediating step. I doubt it's MediaPredict's fault. I suspect (without any proof except my experience with how Fortune 500 executives think and in particular how publishing executives think) that the use of prediction markets may have been de-clawed in this case to avoid dis-enfranchising S&S editors. Think about it:

The PM predicts Bestseller status for a book the editors decide to pass up. It gets picked up by another publisher, fulfilling the PM's expectations and making bundles of money for an S&S competitor. At that point, S&S editors not only have egg on their face, but the S&S board (and shareholders) are emboldened to ask: Tell us why exactly we need you if all you're going to do is to second-guess and stand in the way of this smart little prediction market thingy we've set up over here to perform the same function?

Justin Wolfers (respected PM guru at Wharton) notes the same basic flaw, though you'd have to read to the end of the article to find his critique.

[Wolfers] said that if Simon & Schuster relies on the traders' judgments to select a book, it could skew the bets themselves.

"If they say we find it really persuasive that everyone bet on book A, they're just looking at a book that everyone bet that everyone else bet that everyone else thinks is the best book," Mr. Wolfers said. "So you don't end up with the wisdom of crowds, but the infinite reflection of crowds looking at crowds."

This is at least the second time I've noticed a major media outlet burying what should have been the lead on a prediction markets story--and with it, Wolfers' critique. If I were him, I'd be a little peeved. Last fall, the Washington Post carried a piece ("The Top Pickers vs. The Pack") that completely mashed two irreconcilable ideas, as I noted in "WaPo Misguided on 'Experts' vs. The Swarm":

The article takes the academically and empirically well-proven notion of collective intelligence (aka the wisdom of crowds), and presumptively twists it into the more comforting and familiar (if statistically invalid) notion that "father-knows-best"--only we haven't done a good enough job of finding the right "father" up until now.

I don't know why the media's treatment of prediction markets is missing the mark, but they need to get smart faster. PMs are not going away. Not taking seriously the critiques of Wolfers and others isn't helping. The NYTimes digs itself further into a hole by taking a flawed comparison at face value:

Media Predict is modeled after other so-called prediction markets like the Hollywood Stock Exchange, which allows traders to bet on the four-week North American box office receipts of movies, or the Iowa Electronic Markets, which allow people to bet on election results...

The glaring difference: the HSX and the IEM ask traders to vote on highly distributed end-result questions. As I've observed many times (e.g., here, here, here and here) prediction markets--while very cool indeed, and applicable to many more things than those to which they are being applied today--are just plain lousy at calling things like Papal elections or the nomination of judges where information is closely held by a select group. I suspect it's even worse when that group has sponsored the market and each is trying to predict the actions and preferences of the other.

If the same principle were to be applied to American Idol, voters would weigh in with their opinions not on which singer(s) they liked best, but on which one(s) they thought Simon Cowell was going to like. That's not entirely without its entertainment value, but it adds little to what Mr. Cowell was going to say anyway. Simon (& Schuster) could take a lesson from what the other Simon (Cowell) lacks: humility. The crowd is often right. If they're paying, they're even more right.

As for MediaPredict itself, I'm reserving judgment. At this stage of market development, more new players trying more approaches, covering more questions and attracting more corporate attention is a good thing--and entirely expected. As the MP folks write on their blog: "Love us or hate us, we’re here…" That may be a little extreme. How about: I love that you're here.

Comments

[Media Predict]

We do encourage debate about Media Predict. In response to some remarks above, we would direct you to comments already made on Midas Oracle.

There, we said that the “infinite reflection” criticism is not a slam-dunk. The important point is that selection for publication is in many ways a distributed end-result question, albeit not on the scale of a national election: “The key thing for traders to bear in mind is that thousands of books are published every year. We’re just trying to predict whether this will happen to the proposals on our site. All books on Media Predict can secure publication via getting a book deal on their own, or via winning the contest.” Thus the projection at stake is whether these book proposals are of a quality to please the publishing community in general, Touchstone being one potential publisher.

In the context of your discussion above, choosing books according to sales projections is entirely possible, and indeed will be done at a later phase in Project Publish. However, lead-times are long in book publishing. A decision market based on sales projections would not run its course until two or three years from now. It would seemingly be difficult to recruit participation for such a long-term market at this stage.

In general Media Predict believes that markets can be used – in a wide diversity of ways – to improve the status quo in media content selection. In media, 10 percent of products generate 90 percent of revenue. The more we introduce markets into the process of content selection, the more this situation will improve for everyone.

[Erik Sherman]

I find myself of another mind here. I think there are limitations of what prediction markets can do. Just think about this: what’s the most unreliable type of polling/market research you can do? Asking people what they would do under a hypothetical circumstance. They pick the answer that makes them appear more the way they’d like to appear, and it’s all an unconscious psychological play. In this case, they're trying to be smart and pick the winner, because then they win as well. So it becomes a contest of outguessing everyone else, but, as you indicate, has nothing to do with what readers would actually buy, and that's the real question. Publishing houses have shown themselves inordinately capable of making bad decisions. That effectively reduces the prediction to how they'll screw up this time.

Readers Digest used to be bottom line smart about this whole issue. When trying to decide what types of articles they’d write, they would create a bunch of booklets on many topics. Consumers could get them for a nominal price, but they actually had to put money down. The results told the magazine where to aim its editorial focus. That way they spoke to their readers and got solid reactions. You could do the same with book readers. Offer articles or short stories on similar themes and see to which ones they gravitate.

Only, that still leaves the real problem - how can you ultimately tell what will burst out in success in ways that no one ever expected? Those are often the books that carry publishers into serious black ink.

This also makes me wonder whether publishers are any better at picking winning titles than they were when they were run by people who genuinely loved books.