Tag Archives: data

The Financial Times shows how data-driven is done

The Financial Times launched its metered model years ago, which puts it way ahead of the current paywall curve.

After reading through this Mashable piece, it’s clear that all of the FT’s paywall experience — and, importantly, all of its related data — has made the organization quite savvy. For example:

Looking through some of the reader data — the FT’s data team now numbers more than 30 across three groups — the FT was able to recognize the kinds of patterns readers display before purchasing subscriptions. “We would see the sort of articles they were reading and the frequency they were reading those articles, for instance, and we began to map those,” [CEO John] Ridding explains. “People do behave in predictable ways.”

“… the FT was able to recognize the kinds of patterns readers display before purchasing subscriptions.”

That, right there, is how you put data to use.

The full article is worth a read.

EveryBlock is gone. Maybe it’s time to admit hyperlocal doesn’t scale

EveryBlock is no more.

The announcement was made via a friendly-but-succinct blog post:

Within the world of neighborhood news there’s an exciting pace of innovation yet increasing challenges to building a profitable business. Though EveryBlock has been able to build an engaged community over the years, we’re faced with the decision to wrap things up.

This was a surprise to EveryBlock founder Adrian Holovaty, who left the organization last year.

Hyperlocal is rough. It seems like there’s so much potential. So many local businesses and groups just waiting for the efficiencies of the Internet to revolutionize their efforts. And presumably, there are plenty of local advertisers who want to reach local people. That’s why we still have local radio and local TV and local newspapers.

But there’s something about efforts like EveryBlock and Patch and others that just doesn’t click. I realize I’m generalizing. And I know Patch is still around. And I know these sites all have different approaches and business models. Yet they all have that “hyperlocal” thing in common, and to date that’s been problematic.

The knee-jerk reaction to a hyperlocal failure is to blame the outsider approach. This is a case where I think that knee-jerk reaction is the right reaction. The more hyperlocal failures we see the more I’m convinced the trick to unlocking local is being local. The hub-and-spoke approach these organizations take is fundamentally flawed because they’re trying to create a model that can be plugged into any location. But is that what local audiences want? These outlets are franchising when they should be customizing.

Back to EveryBlock. Whatever the ultimate cause of the service’s demise — it’s weird how quickly it shut down — it’s still sad to see it go. EveryBlock was “data driven” long before everyone jumped on the big data bandwagon.

Pixar’s “Blue Umbrella” and sneaky YouTube views

The Wall Street Journal’s YouTube channel had a hit recently. But not really.

According to Ad Age, a teaser for Pixar’s animated short “The Blue Umbrella” was posted by the WSJ and that clip has generated more than 2 million views.

This is why you can’t take digital analytics at face value. The WSJ didn’t produce this video. Pixar did. The draw was Pixar. (And beyond that, the teaser itself is underwhelming.)

Notable things: Tumblr’s pride is justified, but misplaced; newspaper ad sales are on a very long slide; indoor navigation has an accuracy problem

Here’s the headline: “Tumblr boasts nearly 170 million monthly visitors

Only that’s not quite right. Those 170 million monthly visitors aren’t going to Tumblr.com for the sake of visiting Tumblr. They’re looking at this kind of thing (and rightfully so, because it’s awesome).

That’s an important difference. I have no issue if those numbers are meant to show the rise of Tumblr as a publishing platform. But if the stats are trying to place Tumblr in the same domain as other top sites, we need to take a step back and consider the context.

Here’s Quantcast’s* current list of the top 10 sites “based on the number of people in the United States who visit each site within a month”:

  1. Google; 194,407,568 [monthly people in the U.S.]
  2. YouTube; 174,158,768
  3. Facebook; 140,719,136
  4. MSN; 98,480,592
  5. Twitter; 91,263,448
  6. Yahoo; 79,030,880
  7. Amazon; 76,791,592
  8. Wikipedia; 68,114,712
  9. Microsoft; 63,044,600
  10. Huffingtonpost; 61,289,024

Tumblr is a publishing platform / discovery tool. The only other sites in the top 10 that compare — and this is a reach — are YouTube and Twitter. Both of those sites are also utilities — a significant portion of their engagement and distribution occurs off-site via embeds and external tools. Tumblr doesn’t really work that way.

Tumblr is closer to WordPress.com and Blogger, and that comparison is where things get interesting.

From the same Quantcast stats:

No. 15: Tumblr; 51,947,516 [monthly people in the U.S.]
No. 17: WordPress: 51,182,896
No. 19: Blogger: 48,293,848

Tumblr certainly has something to celebrate, but it isn’t the thing that’s being played up.

*I’m using Quantcast data because that’s the source of the “170 million” figure. The validity of Quantcast’s numbers is beyond the scope of this admittedly feeble examination.

Alan Mutter says newspaper ad sales have fallen 25 quarters in a row:

It is a testimony to the legendarily high operating margins of the [newspaper] industry and the considerable cost-slashing skills of contemporary publishers that nearly all the newspapers in business in mid-2006, when the trouble began, are still plugging along today.

The full piece is worth a read.

Last week I said I need an app for finding products in stores. Sadly, that’s an itch that will remain itchy for some time:

Analysts caution that the technology is still immature, with high costs and accuracy issues keeping more prospective customers on the sidelines. Adding more Wi-Fi access points and other hardware is expensive. Most indoor positioning systems, even using Wi-Fi, still miss the precise location by several feet. And there aren’t enough high-end smartphones in the market that can handle indoor positioning. [Emphasis added.]

“Several feet” isn’t good enough when you can’t find the damn Tobasco sauce.

Want to know what Google is up to? Here you go

GoogleI’ve seen lots of hand-wringing and sweaty prognosticating about Google. What will it do? What does it want? Is that don’t be evil mantra for real?

Funny thing is, Google’s strategy has always been in plain sight. There’s no obfuscation. There’s no misdirection. Heck, this New York Times piece spells it out:

Google has used a similar approach — immense computing power, heaps of data and statistics — to tackle other complex problems. In 2007, for example, it began offering 800-GOOG-411, a free directory assistance service that interprets spoken requests. It allowed Google to collect the voices of millions of people so it could get better at recognizing spoken English. A year later, Google released a search-by-voice system that was as good as those that took other companies years to build.

See what Google did there? It released a free service so it could gather huge amounts of data that could then be used in another product. That’s what Google does. Free leads to data, data leads to another product. Repeat over and over and over and over again.

The Long Tail and iPhone app usage: Nothing surprising here

From The New York Times:

The average iPhone or iPod Touch owner uses 5 to 10 apps regularly, according to Flurry, a research firm that studies mobile trends. This despite the surfeit of available apps: some 140,000 and counting.

I’ve seen the same stat mentioned before. Heck, I referenced that stat in a piece I wrote. But what I find surprising is that anyone is surprised by this. It’s the behavioral equivalent of the Long Tail: a few apps get frequent use — the blockbusters — while the others wane after post-installation popularity or, even worse, don’t get downloaded at all.

Instead of this broad-based stuff, what I’d really like to see is data that links up people’s interests/professions with their most-used apps.

Followers aren’t readers, so let’s stop fooling ourselves

Anil Dash follows up his great post on Twitter’s suggested user list with an equally great piece that politely challenges Twitter follower counts. As he notes, analytics and inflated self-importance are nothing new:

It’s a bit like when I worked at a newspaper: Every reporter thought “Well, our circulation is a million copies, that must mean a million people read my column.” Facing the reality that only 10,000 of those people read the column, or that perhaps only 1,000 of them were reading the advertisement on the opposite page, forced a useful and important reckoning into some false assumptions that were underpinning that industry’s workings.

The key here — and Dash mentions this in his post — is to dispel overblown notions so analytics become useful. Follower counts have value, just as page views, uniques, user-session times, circulation figures and subscription numbers do. But all those numbers have to be filtered through the realities of passivity and engagement.

Yes, But How Do You Feel? Sentiment Joins the Web Analytics Toolset

The New York Times examines sentiment analysis:

An emerging field known as sentiment analysis is taking shape around one of the computer world’s unexplored frontiers: translating the vagaries of human emotion into hard data.

This is more than just an interesting programming exercise. For many businesses, online opinion has turned into a kind of virtual currency that can make or break a product in the marketplace.

Amy Martin briefly mentioned sentiment during her presentation at Twitter Boot Camp in June (the sentiment stuff is in slide No. 9). The concept caught my attention because it strays from typical number-centric measurements like page views, user-session times or velocity. For someone like me, who believes numbers and non-numerical “soft” analysis must exist in harmony, it injects a much-needed psychological component into the audience dynamic. This commingling of data and feelings is why NBC Local’s mood tool is so interesting.

But let’s not get ahead of ourselves with the touchy feely business. Sentiment’s power as a data point is limited because it’s a loaded concept with infinite variations. If my “positive” could be your “neutral,” how can a measurement tool adequately capture sentiment on a broad, numerical level? It can’t. Not reliably, anyway. Wild swings and spikes will appear in graphs, but small percentage shifts between open-ended terms are too ambiguous to rely upon. That’s why sentiment needs to function as a general data point for online engagement. It’s a single tool on a big analytics workbench.