(Over) Counting Widgets

I got an email from a friend the other day.

"Did you see that Clearspring has served 3bn widgets in the past five months?"

I hadn't actually seen that news, but here is the press release my friend was referring to. It's clearly an impressive number and Clearspring is doing some great things in the widget market.

But before we start putting Clearsping in McDonalds territory (billions served), let's get something straight. Serving widgets generates huge numbers quickly.

All you need to do is look at Photobucket, Slide, and RockYou's numbers for their photo/slideshow widgets to see how powerful the widget model is. I don't have access to the actual numbers for these three photo widget services and I'd prefer not to print the rumors I've heard, but I'd venture a guess that their numbers for widgets served each day will make Clearspring's 3bn number look tiny.

I do know the numbers for FeedBurner's widgets and they are well north of what Clearspring is serving. But this post is not about whose you know what is bigger than whose.

It's about the challenge of understanding what is what in the widget market.

Max Levchin, Slide's founder and CEO, is one of the most thoughtful people in the widget market and I've had some conversations with him about the challenges of measuring widgets. He's frustrated that there isn't good third party data on the widget market. I've also talked to the team at comScore about this issue because it's starting to become an issue everyone is paying attention to and there just isn't good data yet.

You can get unique visitor counts from comScore on widgets like Photobucket, Slide, and RockYou, but what does it mean? Here is the comScore data on the photo widget sector (the top three players).

Are those 18mm uniques that are attributed to Photobucket being seen on Photobucket.com? Or are those 18mm uniques the number of people that are being exposed to the Photobucket widget wherever it is being embedded (MySpace, Beebo, etc). I don't know the answer to that simple question, but it's an important one.

And what's the right number to look at? Should Photobucket get credit for having an audience that sees its widget on other services pages? When that page includes five to ten other widgets? Or should it just get credit for those who interact with the widget in some way?

The bottom line is we need better numbers on widgets, we need some standards, and we need them now.

Comments

[Chris]

There may be an alternative to applying "standards" of measurement to widgets, not that I discourage audience measurement firms from trying -- they're after it already.

The internet's most (financially) popular "ad unit" is the search result. I am certain there are gadzillions (tm) served every day, but we don't know, since there's no "standard" of measurement. Instead search results are valued by the "click marketplace," where eCPC's and yields reign supreme. How many "served?" Who cares?

No question that impression-based media buyers will need "standards" and data that substantiates their budget outlays, but the marketplace will likely grow to value these "impressions" prior to any measurement firm, and that will be interesting to see.

[Mike]

I'm a big believer in the value of usable metrics and appreciate all the challenges there are in getting accurate measures for web pages in a world with AJAX / DHTML / RSS, etc. The world of widgets adds further complexity.

However, I'm willing to offer up at least one solution. Since the term "widget" is used generically to describe a variety of things, can we start calling the metrics for widgets "Wetrics"?

[Michael Hoffman]

You raise important issues here and I am not sure they are solved by simple numbers. This can get complex -- if the widget is served on a page full of widgets, if it's below the fold, etc. then you take some discount? And I can imagine that on their home site you can have advertising and sponsorship that you won't be able to have everywhere, further complicating the issues.

All of this leads to the big question which is, what's the business model?

[Ro]

Thanks for the mention.

Actually, we (RockYou) are working with Clearspring on that exact issue, tracking widgets installed, impressions and placement. Another strong player in the widget measurement and management space is Widgetbox. Internally we track well over 100MM views a day across our network of 10M+ of widgets placed across social networks.

As for Michael's concern, our flash-based widgets don't launch until the viewer of the page actually scrolls to the widget itself. I.e. if we have 5 widgets on a page, but the viewer only looks above the fold where 1 widget is placed, we only count/track that as 1 view, since the other 4 widgets never launched.

[Grand Egress]

This post alludes to but doesn't actually drill down to an equally big problem: double-counting of visits and visitors.

If I view a web page, say like the home page of AVC, which contains multiple widgets, how many visits and visitors occur? Duh. One, and one.

But current metrics reporting and the understandable desire for self-promotion by new media companies would have us all believe that multiple visits and visitors are occuring (and the hype-loving media then seems to validate it all by repeating the questionable numbers ad infinitum.)

That is, in my example, AVC claims a visit and a visitor when I view the main page (as it should.) But every widget distributor also claims a visit and a visitor.

The truth of the situation comes out when the hype-ignoring financial analysts step in, as just happened with the MySpace deal to acquire Photobucket, which may be a great deal for everyone, but which the financial analysts noted will bring to MySpace nearly zero new unique visits and visitors -- because the essentially 100% of Photyobuckets visits and visitors were actually just loads of the widget on MySpace pages...

[Bogus Jones]

There may be some value in a better "wetrics"...but each type of widget has MUCH more going on than simple view/refresh/loads and eyeballs. It's not the web. It's widgets.

In my humble opinion, analysis of widgets has to come down to business model. Sure, a 'best new band' widget might have 27,000,000 views...but did they do any business?...did they generate any revenue?...did they deliver any advertising?

Anyone who has seen the new Google Analytics inter-face knows just how long it has taken to create meaningful metrics for web activity. "Wetrics" are going to demand a whole new den of analysts and anlysis. I just hope that it doesn't make the horse race more important than the horses...lest I dream.

-Bogus Jones

[Judah]

The big problem with getting metrics on widgets is that you are looking in the wrong place. comScore does not have reliable or accurate data. At all.

With a real web analytics system, such as Unica NetInsight or Visual Sciences, which employ valid data collection methods, like javascript page tagging, packet sniffing, or log file analysis, you can get these metrics by extending your data model to track "events" subordinate to the page view.

Using external audience panel measurement data, in general, is a faulty approach for measuring this level of on-page granularity. External metrics are like looking through a telescope at your front door, then asking the alien astronomer on another planet to identify what people did in your home. I prefer to use verifiable, auditable data recorded by the video cameras in my house to know who sat on my couch, such as what I get when I use a real web analytics tool.

[Brandon Watts]

I agree wholeheartedly with you. With widgets at the heart of some of the revolutions that are happening on the Internet right now, you'd think that we'd have full details about their results, but as of yet, those details are still rather amorphous. To be honest, we like, use, and promote these things that we essentially know nothing about.

However, when the full details are revealed, I'm willing to say that people are going to start taking widgets even more seriously.

Brandon Watts
Criteo Evangelist