Twitter Loves You 70% More Than Oprah

One of the first things you notice about twitter is what a positive, happy place it is. In fact, Tweeters are 51% more likely to show love than the world at large. Facebookers show 10% more love than average. Oprah.com, on the other hand, shows 11% less.

How can I possibly know this?

By using Google as a semantic mining engine to assess a web community's culture. While there's a lot noise and inference in this data, it's still interesting and revealing.

For the Google index at large, love dominates hate by an impressive 9 to 1 margin (that stat makes me happy). On Twitter, love wins by a wider margin: 14 to 1. On MySpace, love eeks out a more modest 6 to 1 spread.


I calcuated these ratios by using the results from these Google searches:
site:{sitename.com} love
site:{sitename.com} hate
To decide if the ratios were higher or lower than the world at large, I compared them to the simple ratio of "love" and "hate" search results on google overall. There are large sample sizes-- Google has 1.7 billion results for "love".

While Twitter gives a lot of love, some sites give even more, like ancestry.com (248% more) and modelmayhem.com (389% more). I guess we do love our families a lot, and fashion models even more.

There are some surprises among sites that give less love. Oprah.com shows 11% less love and BarackObama.com shows 55% less.

Moving beyond the headline-grabbing fun, this analysis is a small example of ways to assess the culture of an online community. By using variations of the same technique, you can gain some inference into issues such as:
-- how happy or sad does the community report being? (symmetrical antinonyms make for easy analysis)
-- how analytical is the community (for example, BarackObama.com has 3X more relative usage of the word "analysis" than the internet overall)
-- how active or passive are the participants in a community? (Tweeters use active voice verbs far more often than Facebook or MySpace.)

While the possibilities never end with this analysis, we are only at the beginning of a trend to study virtual community cultures. I think we will discover that these tools help us to at least partially explain and predict their relative success and failure.

For example, if Twitter is a more "loving" community, does this speak to its relative success, or not? And what about Twitter would make it a more "loving" community?

I suspect that the answers will prove to be similar to what we generally know:
- Community culture is very important in driving behaviors-- and is critical to community survival, growth, and success
- Community culture is determined by the participants and strongly influenced by the governance methods
- Critical issues become: Who governs? How is governance shared? To what ends? With what methods? How is governance allowed to evolve?

I give the Twitter founders a lot of credit for establishing a great culture, including its degree of love and fun. Now it's up to them and the Twitter community to maintain and adapt this culture through their hyper-growth cycle-- no easy challenge. How they choose to govern (including the degree of technological and management openness) will be important.

Search Revenues Will Bounce Back

Despite the worst advertising recession since at least WWII, both Google and Yahoo's search revenues increased 3-4% in the U.S in Q1 compared to a year ago. Once the economy truly bottoms, I expect search revenues to rebound nicely.

Here's why:

Search query volumes in the U.S. are still growing 30%+ per year.
Yields per search have declined by roughly the same amount, due to three factors:

1. CPCs down about 10% year over year
   


2. Click through rates down about 10%


3. Ad coverage down about 10% per year (see graph above)
   

While all three yield variables have been negatively impacted by the economy, that won't be true for long. Let's consider each variable separately since they will behave differently when the economy improves.

First, CPC's. This is the easiest call. When the economy turns, CPCs will first stabilize and eventually rise again. Competition for attractive keywords will rise; volume on the highest value keywords (like mortgages) will increase; and the long-term migration of advertising form off-line to on-line continues. Why am I so confident about this? Because the supply of high quality keyword searches, meaning the supply of highly focused consumer intent, is far more scarce than advertiser demand for that intent.

Second, click-through rates. This variable is trickier-- I would expect a lower rate of decline, though a reversal of trend is less clear. On one hand, a stronger economy will increase consumer buying interest and the likelihood of clicks; so will ongoing improvements in ad matching algorithms. On the other hand, more searches per user decreases the likelihood of click-throughs on a given search. And then there's the elusive issue of "ad blindness". On the margin, are people starting to "tune out" ads? While this might happen to generic display, search is driven by clear intent, so this effect should be less. (There's a lot more to explore in subsequent posts about intention-based ad models.)

Third, ad coverage. I would expect the rate of decline to moderate, but not necessarily reverse. On one hand, there are many small/local businesses (millions of them!) that are not yet using paid search-- but they will. On the other hand, I expect much of search query growth to be less commercial and more social in nature ("social search").

Putting these pieces together: I'd expect CPCs to rise again, click-through rates to moderate their declines, and ad coverage to continue to drift a bit lower. In sum, the negative 30% yield headwinds will moderate substantially (though not entirely reversing), allowing the search engines to convert some of their rapid search volume growth into revenue growth in the U.S.

International search revenues have an even brighter future. I'll tackle that later.

A couple of long term trends that will ultimately challenge search revenue growth are the rapid growth of mobile and social media-- and the intersection of both. (I'll tackle these topics later as well.) While these are very powerful long term trends that are already significantly at work, the "traditional" search market is far from fully penetrated itself.

Why Search Volumes Will Keep Growing, Big Time

Search volumes in the U.S. are still growing at roughly 30% per year. I expect chunky double-digit growth to continue, since in terms of widespread adoption, "search addiction" has only begun.

Here's what I mean: the average number of search queries per person in the U.S. is now about 3 per day; this is up from about 2 per day a few years ago. Heavier search users, representing the most active 20%, average about 12 searches per day. The other 80% average less than 1 per day.

I think that last statistic is remarkable: 80% of search users average less than one search per day. This suggests that the adoption curve of search usage is still very young. Better search tools and more relevant search results will drive incremental usage-- but a lot growth will come simply as users become more sophisticated.

How high could search intensity ultimately climb? I'm not sure. As a search-aholic myself (see my Google search history above), I don't find 12 searches a day to be "extreme" at all. As you can see, that's an "off day" for me, with many days above 20 or 40 searches; but, of course, I work on the internet.

In 5 years, if 40% of people (instead of 20%) are averaging 12 searches a day, search usage will have doubled. Since the whole distribution continues to shift up, I expect growth will be faster than that.

The search metrics I refer to above come via Larry Chase's interview with comScore's Search Evangelist, Eli Goodman.

Long Tail Getting Longer & Fatter

As users search more, the length of their queries grow. This is a good trend for long tail publishers.

comScore data shows search engine queries continue to lengthen. Query length looks to be increasing about 3-4% per year. This may seem relatively slight, but the impact on the nature of search traffic is enormous, since keyword combinations are exponential.

While the math requires several debatable assumptions, I'd estimate that the 7% increase in query length during the last two years would increase the average number of possible keyword combinations by 25-30% for a website (assuming the site did not increase its amount of content). This trend may accelerate, since the longer the average query gets, the more exponential the effect.

Sites with extraordinarily deep content (ie: long tail friendly) should benefit most from this ongoing trend.

I'm composing a 78 word query now to celebrate!

The comScore post has other interesting data as well.

Free Throw Accuracy - NBA Moneyball with Dr. Tom

Dr. Tom Amberry, at the age of 71, made 2,750 free throws in a row. He didn't miss, he was asked to leave, as they were closing the gym.

NBA players make 77% of their foul shots, a relatively disgraceful figure. (BTW, the NBA free throw percentage is basically identical to the WNBA.)

Dr. Tom's training regime seems to get results for others. He claims to have improved a junior high team from 40% to 80%+ in a week. If he could take an NBA team from 77% to 85%, he would be worth an astonishing 10 wins a season.

Here's the math:
- The average NBA team shoots 25 FTs per game
- Increasing FT % by 8 percentage points results in 2 extra points per game
- 12.5% of games are won by 2 points or less (includes OTs)
- 12.5% of 82 games is 10.3 extra wins

An extra 10 wins would have put any of these seven teams into the playoffs this year: Phoenix, Indiana, Charlotte, New Jersey, Milwaukee, Toronto, and New York. But no, they missed free throws and the playoffs.

This year, if Orlando had made one extra free throw (here's looking at you Dwight and Hedo) in regulation in Games 2 and 4, the Magic would have led the series by 3-2. Orlando did overcome the worst FT% in the regular season this year (71.5%), but it caught up with them in the end. Without any doubt, poor FT shooting prematurely ended the Magic's championship bid.

There are probably less than a dozen players in the NBA who are worth an extra 10 wins per season. All of them make more than $10 million per year than the average player. I don't know how much Dr. Tom charges, but I assume it is less.

This NYT article says that NBA free throw shooting hasn't improved in 50 years. This freakonomics post says that while the average player hasn't improved, the top players have improved a lot (which means the worst have gotten worse, yikes!).

Maybe some variable pay in the NBA would cause improvement. Or hiring Dr. Tom. Or both.


* Thanks to Tim Katlic for researcing this post.