The importance of Return on Attention

There has been a lot of discussion recently about the impact of Search Engines, and specifically Search Marketing, on the Web. Some people have asserted that Search Engines are “leeches”, feeding off of other people’s hard work. Others have defended Search Engines as wholly beneficial to the vast collection of web publishers, providing an income and distribution source that they otherwise would not have.

This discussion got me thinking about the real incentives provided by Pay-Per-Click Search Marketing. Nick Carr touches on this in a recent post:

“The economic incentive for the content producer therefore is not to produce content that simply engages a large or demographically attractive audience, but to produce content that (a) attracts an audience likely to click on a valuable advertising link and (b) increases the odds that such lucrative clicks will actually happen.”

This is an important point – these incentives are real, and can lead to two outcomes for content providers:

1 – Content providers change the type of content they provide to get a more valuable audience (for example, emphasizing product reviews instead of how-to articles)

2 – Content providers change the content of what they publish to more explicitly steer consumers to click on advertising. For example, writing good product reviews for marginal products, or writing articles with integrated “product placement”.

In either event, the safeguards that traditional publishers have had in place for years to minimize the impact of advertising on editorial content will not exist.
This is where Return on Attention comes in. If we can measure Return on Attention, we start to give consumers more control. How?

First, feedback. By letting other people know how valuable you found a site, you are strengthening the quality of the content (not in a News Hour with Jim Lehrer way, but perhaps in an audience-friendly Fox News way).

Second, quality. With click fraud and rising keyword prices, advertisers will start to demand and seek out high quality lead flow, not just a high quantity of leads. In other words, they will start to reward sites that send them eventual customers, not sites that send them leads.

In a world focused on clicks, all that matters is getting the click. Advertisers and search engines have become very skilled at increasing clickthrough rates on their ads. However, that results in low quality leads. The coming focus on high quality (but perhaps lower volume) leads is a disruptive change for the market. Return on Attention is the critical guide to success for publishers. Focusing on Return on Attention – and therefore focusing on quality content – will yield the greatest return on online publishing assets in this new environment.

Is PPA the answer?

It would be a stretch to claim that Google is in big trouble.

However, there has been considerable negative chatter recently about the implications and risks of the Pay Per Click (PPC) business model employed by Google and the majority of other online marketers (search engines, comparison shopping sites, etc). See, for example, Blodget, Nielsen, and Carr.

The objections include a) Click fraud is damaging advertiser returns while benefiting Google b) the PPC model yield a high volume of low quality leads c) search engines benefit from improvements in conversion rate on advertiser websites, even though they do not contribute to these improvements. Fundamentally, when you boil it down, the issue is mismatched incentives. Google does not have the same goal as its advertisers.

If (when?) this becomes a real, widespread issue for advertisers, Google will have a problem. It will have to start seriously considering employing other advertising models. They’ve started playing with CPM (cost per impression) based models. CPM is pretty old school, Web 1.0, relying on eyeballs. However, reverting to 1999 (even an auction-enabled version of 1999) is not a solution. CPM reduces the direct pay for performance aspects that people love about PPC, and it does little to eliminate gaming – it just changes the game.

PPA – pay per acquisition – models may be a much better alternative for the advertiser. In this model, the advertiser pays a commission when someone buys something. This has existed for a long time as affiliate programs. Amazon, for example, uses this approach when it allows people to put book ads on their sites. Amazon – the advertiser – only pays when someone buys the book.

PPA-based business models do not have the same problems as PPC. Sites are no longer compensated for low quality leads. Click fraud is a much smaller issue – few companies will be willing to pay people in India and China to actually buy stuff.

Indeed, PPA aligns the interests of the search engine and advertiser to an extent impossible via other models. And it is not that far out. Mainstream ad agencies have started talking about using this type of model for traditional advertising. It is, if anything, easier to do this online.

Anyone who figures out how to really apply PPA models on a broad scale will have an eager advertiser customer base. It’ll be interesting to see which of the search giants take the initiative on this.

“We were eager to help them shut us down”

“We were eager to help them shut us down,” jokes Sun-Times Publisher John Cruickshank. “They’re buying ads. We like that.”

Today’s Chicago Business reports on an expansion of the tests Google was running a few months back, this time focusing on newspapers instead of magazines.

“In a quiet and small-scale experiment, Google is running classified-like ads in the pages of the Sun-Times, which so far is the only newspaper participating in the Web-search behemoth’s test.

… Now, in the Sun-Times, Google is running ads in proximity to relevant content. On Dec. 12, for instance, Google ads touting ticket brokers, White Sox apparel and Chicago Bears memorabilia ran in the Sports section.”

A peek at VC returns

It is always tricky to really understand the world of VC and other Private Equity funds. Specifically, the funds typically go to extreme lengths to not disclose the returns on their portfolios. However, the Ohio Workers Comp fund has just published details of their returns. All in all, they made a total of 4.5% since 1998.

I think they need a new investment strategy.

The article also shows the vast difference between funds:

“Some funds, nonetheless, have already generated high returns. The bureau’s $8.2 million investment in a fund run by Quad-C Partners of Virginia in 2001 produced a $1.9 million profit check and a March 31 balance of $13.7 million – a return of 89 percent. Its 2000 investment of $14.3 million in the Carlyle Partners III fund of Washington, D.C., was up 49 percent as of March 31.

None of the five Cincinnati funds managing bureau money showed a positive return as of March 31.

Its $13.6 million stake in the Blue Chip Ventures IV fund in 2000 was down 16.8 percent. Its $11.1 million stake in the Fort Washington Private Investors Equity III fund was down 4.2 percent.

Meanwhile, the bureau’s $2.4 million investment in River Cities III declined 9.2 percent. Its $900,000 stake in Triathlon Medical was down 47.1 percent. Its $350,000 investment in Charter Life Sciences declined 42.9 percent.”

Declining Search Engine Marketing Effectiveness?

Jeff Matthews has a fascinating post deconstructing a recent FTD Press Release. Worth reading the whole post, but I will excerpt the key points here. The press release says, in part:

“During the Christmas season, certain online search engine costs increased significantly over the prior year, and as such we made the decision not to pursue the resulting high cost order volume.” stated Michael J. Soenen, President and CEO of FTD Group, Inc. “As a result, despite this slight decline in order volume for the Christmas season, we are reiterating our EBITDA and EPS targets for the year. Further, we have begun making additional investments in our marketing staff to help build a more diversified marketing portfolio.” (emphasis added)

As Jeff rightly points out, this likely means “the marginal cost of a new customer has reached parity with the marginal profit from that customer. Which is not something anybody expected happening any time soon.”

This is a potential huge problem for search engines – if there is no net value, growth will slow and possibly even go negative.

How widespread an issue is this? I don’t know. But there are some reasons to believe it could be an emerging widespread problem. After all, search queries are pretty concentrated in a small number of topics, including shopping. And, as Nick Carr pointed out a few days ago, the click economy incents publishers to figure out how to get consumers to click. Quality of lead flow has not traditionally been an issue of concern for search engines; their model is pay per click, not pay per acquisition (unlike, say, Amazon’s affiliate model). It will have to start being a concern soon, if this data point starts to morphs into a trend.

UPDATE: Battelle and Carr have both picked up the thread…

Return on Attention

Ed Batista at AttentionTrust wrote a great response to my post on the importance of Context (Beyond WHAT) in the analysis of attention data. In it, he says:

“I’m also not ready to concede that context can’t be wrung from the clickstreams and other attention data.”

I completely agree with Ed – I am not ready to concede that either. But the key, I think, lies in “other attention data”. I don’t think clickstreams are enough. Looking at clicks just tells you what someone has paid attention to in the past. In order to make sense of it, you need to know why they paid attention to it, and even more importantly, how much they valued the experience. In other words, to borrow John Hagel’s expression (not sure if John coined it or borrowed it from somewhere else), you need to know their Return on Attention.

In a stable environment, you can figure out the Why and the ROA through analysis of a large enough set of clickstream data. If someone returns to a source often, it is likely it provides them with a high Return on Attention. If someone buys coffee at Starbucks every day, it is safe to infer that they like Starbucks.

The vast majority of environments, however, are not stable. Ed alludes to this by noting that analysis of December clickstreams would likely lead to incorrect conclusions, because of holiday shopping. This is a great example of a very common phenomenon. People’s tastes change; Their contexts change, and not just at Christmas time. Take a vacation, work at home, buy a present for a friend’s birthday, go to a charitable event, think about joining the Army, have a kid, get married, change jobs – all of these are events that dramatically alter your context. Every one of these events diminishes the value of time-series clickstream data by some small amount, and makes it more likely you will draw the wrong conclusions. They also make driving actions much harder – you can see patterns, but with no understanding of the context changes, it is very difficult to provide recommendations that increase your Return on Attention.

(This is one of my big frustrations with the Amazon recommendation system, by the way – a little while ago, I was shopping for a present for my 3 yr old niece – no occasion, just a random present. I bought one from Amazon. Ever since then, I have been getting recommendations about other kids books, videos, toys, etc, even though I have relatively little ongoing interest. This only stopped when I went in to my ratings and history and provided Amazon with more context – specifically, that I wasn’t interested in those items).

This is not to say that clickstreams are not important or useful. They are. I have used clickstream data in the past to segment consumers, to target recommendations, to improve offers. It is certainly a part of any solution. But in dynamic environments, the value of clickstream data becomes hugely amplified when we have “other Attention data”. Specifically, data about Why and Return on Attention. They help us make sense of the clickstream data, to understand what still applies and what does not.

Where we at yapaZOO net out on this is that the issue is not about how much data we have, but the right types. Each set of data provides insight into consumers. Putting them together, however, yields far more than the sum of the parts.

BTW, also mentioned in Ed’s post is that he has no idea what we do, because we haven’t said anything about who we are or why we exist. I know our silence is annoying, and we’ll be coming out of stealth soon, probably in a few weeks. But we are working generally on this problem; we don’t claim to have a complete solution to all aspects of it, but we think we’ll provide a good step forward. We’re also huge supporters of what Ed and the others at AttentionTrust are trying to do, and will be joining as soon as we’re a little further along.

Beyond WHAT

From the AttentionTrust blog today:

” …Two of her [Elizabeth Lawley] comments really stood out for me. First:

Syndicated subscriptions are an attention filter… There are students of mine, colleagues of mine, who would love to be able to subscribe to my attention filter. ‘What is Liz reading?’

… Something implicit in Liz’s comment–and something I know Steve will discuss–is the importance of inattention. If I’m interested in knowing what Liz is reading (and I am), I’m just as interested in knowing what she’s NOT reading. “Attention filters” will help us find the good stuff and avoid the dreck, using attention data generated by us, our friends and colleagues, and although syndicated subscriptions are moving us in the right direction, we have a long way to go.

Liz was also careful to draw an important distinction between social networks and “attention networks,” and she posed a compelling question:

How do we meld attention networks with search?

This becomes a huge question when we start thinking about how “attention filters” might evolve. For example, my blogroll is a simple attention filter that gives you a very general sense of the issues that interest me, but it won’t tell you which authors I read the most, let alone which individual posts are most relevant to my interests.”

Hinted at in this entry is the concept of context. You want to know not only WHAT someone is reading, you want to know WHY someone is reading it and how much they LIKED it. In other words, you want to know a) was this worthy of their attention?; and b) did they get a good Return on Attention?

Simply capturing clickstreams cannot give you the context. But to make the concept of Attention Filters practical, we need to get to that level. Data without context doesn’t have meaning.

Search as Platform

Ok, just a quick post for now, until I more fully digest this. But this looks really cool.

Amazon, whose A9 search engine is cool but complex, has launched their Alexa Web Search Platform. Details via Battelle.

This is pretty cool stuff. If it works as it is advertising (a relatively big if, I suppose), it could dramatically lower the barrier to entry for Vertical Search Engines (like SimplyHired, for example). That would be pretty cool, because it would continue the shift away from technology and into community. It may also destroy some of these company/site’s business models, but it is not entirely clear they had one to start with.

This also solidifies the talk about Search as a new platform. Amazon has a lot of experience with this; by any measure their Web Services have been a huge success, and this merely extends their strength into new areas. The more people can build on this platform, the more people will naturally tend to bias their services towards Amazon.

Ok. I am starting to say things that everyone already knows. It’s late – sorry for the incoherence!

Let’s just agree that this is pretty cool, and we’ll think more clearly about why in the morning.

Cookies Seen as Privacy Threat

From AdWeek :

“Well over half of users believe Internet tracking cookies invade their privacy, according to a new poll.

In a survey of 150 Internet users, 64 percent said cookies represent an invasion of their privacy. The poll was conducted by online ad network Blue Lithium, which recruited participants through banner ads.”

I haven’t seen the details yet (and recruiting via banner ads does not seem terribly sound and that is a small sample size. But I digress…). But the Behavioral Advertising folks everywhere should continue to be worried. Cookies have been around a long time (“The Blue Lithium survey… found 83 percent of respondents claim they are familiar with cookies”), and rightly or wrongly, consumers still have concerns. Unless the industry figures out a way to make more people comfortable with this, they are at serious risk for a backlash and/or regulation.

This is also going to continue to be a large roadblock for brand advertisers. As a friend at P&G told me a little while ago, it is just too risky when consumer sentiment is so negative.

In exchange for attention data…

Couldn’t say it better myself, so I figured I would just alert everyone to jd’s post at his Grain of Salt blog:

“…ultimately, I want every bit of every (AttentionTrust member) page I view to be 100% relevant to me. In my perfect world, this relevancy would come in the form of ad-free content. But I’m not so foolish as to think that my inattention to (in the form of not clicking on) the ads you display will cause them to magically disappear. So if you insist on cluttering up the content that I came for with things that I did not, at least take the time to analyze my existing attention data and make use of it.“

(emphasis added)