Cogblog

Outside.In is launching a new tool to make it easy to become a local newspaper.  Is that smart?  They convince AllThingsD to make it sound reasonable by sharing a spreadsheet. Great marketing!  I thought the spreadsheet was super wrong, so I revised it:  Google Doc Spreadsheet

Their original spreadsheet seemed wrong to me in a few respects:

• 40m pageviews is the average number of pageviews?  That means they get as much traffic as the LA Times web site.  Is that realistic?  Here is a spreadsheet with pageviews by newspaper.  This data is two years old, but very few get that many pageviews.  Compete.com indicates that last month the Baltimore Sun got 3.5m unique visits.  That means it is quite unlikely that they are anywhere near 40m page views.  If an institution with a huge marketing vehicle, tons of marketing support, and city institutional status can’t get anywhere close, it is really, really hard.
• The network model is preposterous.  First, they imagine that their sales force can build an ad network of 80m additional monthly impressions.  Then, they can do a 50/50 rev share with these publishers and get all their inventory, even though they only have a 20% fill rate.  Then, their sales force can sell it for $10 cpms.  Is there really that much local inventory available?  Seems like if it is, it is across a long tail, requiring substantial sales and publisher service to support and acquire.  Next, is getting it at 50/50 realistic?  I would expect a publisher would probably ask for more.  Furthermore, if you only had a 20% fill rate, then the publisher would probably put you on a daisy chain and use networks themselves.  Why share 50% of your revenue with this local entity if they are just collecting pass-through from Google or Advertising.com.  Finally, can the sales force really expect to sell weakly branded network inventory for $10 cpms?  I have found that sales forces accustomed to selling a premium brand struggle to sell network inventory due to the changes in the business model.
• Is the remnant monetization realistic?  They claim to get $5.00.  Certainly there are parts that are contextually effective and will see an aggregate page CPM of $5.00.  Other inventory will go for $0.50.  In fact, I would say that most inventory will go for $0.50.

Let’s just say that their model is not conservative.  It seems pretty aggressive, actually.  If you decrease all of these numbers just a little bit, the business is suddenly bankrupt.

Fred Wilson, an investor in Outside.in, says:

As you might imagine, it’s a “honey we shrunk the kids” story. The topline goes down by an order of magintude and so do the costs. The profits are still there (at least in theory). In Mark and Peter’s strawman model, a local media business with 40mm monthly page views does about $7mm in annual revenues and almost $3mm of pre-tax income. You can go click on that link in the above paragraph if you want to see the model.

…

And if that is true, and I think it is or will be, then the local media companies that leverage their audiences for their content, create communities and conversations, will win. And they’ll be profitable businesses worth owning and investing in.

Fred is an awesome VC – one of the best.  There is not a chance he makes a single investment in a business like this.  Even if things went their way, is investing in a business that, at scale, generates $3m profit really a business?  Fred might invest in the roll-up of 300 of those companies.

I walked into my local Starbucks the other day and realized that they had made some serious changes to their menu: They had devoted 1/3 of the menu to Frappuccinnos (iced slushy beverages), 1/3 to teas, and 1/3 to coffee.  The result of this menu distribution was they no longer offered all of the “flavored” coffee drinks in a non-Frappuccinno form.  When I asked them about it, because I am somewhat partial to their “Espresso Truffle”, the barista’s indicated that they still, in fact, had all of these drinks, they are simply no longer on the menu.

My reaction to that was, “how are they going to sell any of them if they don’t tell people they are for sale?”

This is an interesting decision by Starbucks that gets at a larger sales force decision.  To wit: Starbucks could have simply added more menu space, covering their walls with the hundreds of drinks they serve, but would exposing every drink they offer actually cause more or less confusion?  Would it increase sales?

Compare and contrast this with In-N-Out Burger.  They have super simplified their menu.  It only has 4 products that are not drinks on it: Double meat/double cheese burgers, cheeseburger, hamburger, fries.  Doesn’t get much simpler than that.  They also have a “secret menu” that you can order from if you are “in the know”.  Amusingly, even the secret menu on their web site does not have all of the secret things you can order.  There are literally hundreds of discussions about the secret menu all waiting for your google query.

Let’s bring this back around to the enterprise.  At AOL Advertising, we literally have hundreds of products for sale.  Almost any combination of behavioral data (and we can generate basically any behavior) and web sites (and we work with almost every web site), geographic targeting, custom dynamic banners, whatever.  Netblocks, clutterbusters, morphing logos, sponsorships, takeovers.  People can buy this on a CPA, CPM, or CPC basis.  Besides display, we have a search product, an affiliate product, a mobile product, and a video product.

How can our sales force, much less our customer base, hope to understand the myriad products available for purchase.  Yet, if you split the sales force, you have multiple sales reps calling on the same customer competing for a single budget.  Not having one person sell everything de-optimizes the sales force in the eyes of the customer.  Certainly, AOLs attempts to address this have been well-documented.  And every customer wants something a little different.  I am partial to the Espresso Truffle.  If AOL Advertising took some products off the menu, some customers would be angry, some would think we lose business because people want those products, and some would not buy from us because that is what they want and they just don’t know it is available.  Yet our customers only have a limited amount of time to spend learning about our products.  What products should we talk about if we only have an hour to educate the customer?  30 minutes?  15?

If there is one thing I have seen, it is that customers love the new stuff.  Our sales force would love it if there was a new, hot thing every two weeks that they could call their customers about.  Yet every new thing makes something fall off the back that a salesperson forgets he could sell.  Keeping the right things top of mind in such a world is a challenge for every sales organization.  And typically the new stuff is not the best stuff, it is just the new stuff.

I know, for me, I try very hard to avoid introducing new products.  And I run New Product Development!  How does your company manage the challenge of keeping the menu manageable and optimized?

Great post on the O’Reilly site on how “Twitter is Not a Conversational Platform“.  It made a number of points that resonated with me:

• I can count on one hand the number of people that I have “met” via Twitter.  In fact, it might be zero.  I haven’t thought of one yet, but I am not saying unequivocably.
• I suspect that this has increased since Twitter made the decision to cut down on feed spam and stop showing @replies.  Now, your ability to serendipitously discover new Twitter users has decreased dramatically.  At least, with the new Facebook, I am reading comments by people I don’t know and get to know them a bit vicariously.
• The nature of how information is shared, but relationships are not really built is certainly true, and it made me want to go read Clay’s book.  Twitter is rarely building shared experiences.  In the same way, while Cogmap allows people to “collaborate” on charts, there tends to be little actual collaboration, much like Wikipedia.  The result is a site with lots of fans, but not a lot of community.

Interesting to think about how to drive truly shared experiences in a way that constructively builds community.

A wave of articles are coming out marketing the value of more intrusive web advertising.  To wit:

Here Come TV Commercials Between Web Pages (TRI)

“…ShortTail Media is testing a new ad unit that allows publishers to insert 15- and 30-second video ads between pages on their sites. Reuters has already signed up for a beta test this summer. MSNBC.com and The Weather Channel are close to joining in, reports AdWeek.…”

Furthermore:

“If all this sounds like it will hamper the user exeperience, then too bad, says ShortTail Media CEO David Payne.

In a speech at the Advertising Bureu’s(sic – Cogmap editor) annual meeting, writes AdWeek, Payne told publishers they needed ‘ to adopt bigger, bolder creative and to be less sensitive to user experience.’”

Pretty enlightened, huh?

MediaPost releases an editorial from the creative director of PointRoll the next day and his message is:

“We’ve got this great new medium. Let’s start using it to its full advantage. Little rectangles and squares, many of which appear on the same page at the same time all vying for audience attention, are never going to compete on value with television commercials and large, splashy print and outdoor advertisements.”

Martin is a little better balanced.

There was a recent article in the New York Times about how Wired, the print magazine, is crashing and burning.  One of the points they make is that Wired, the print magazine, has less than a million readers, while the on-line version has more than 11 million readers and makes less money.

Is the answer a worse user experience, as implied by David Payne?  Or is it something else, as implied by Martin Betoni of Pointroll?  Looking at this tiny set of data, the answer could be both:

• One solution could be more intrusive ads.  If ads that truly wrecked the wired.com user experience were introduced, maybe they only have 1 million web site users a month, but the web site is now a going concern.
• Another solution could be smaller changes that avoid significantly impacting the user experience.  Looking at these numbers, you only have to deliver 1/10th the value per user to generate the same revenue.  (Of course, Wired is losing gobs of money, so maybe that doesn’t get you there.

I will say this:  The “remnant business” fits in here somewhere.  Not sure where, but somewhere.  Here are a few facts:

• There are more web pages being viewed every day (supply is increasing)
• Of those web pages, 20-40% are sold for premium prices (~$20 cpms), and 60-80% are sold as “remnant” (~$1.00 cpms)

More intrusive ads may have the twin benefit of decreasing supply (making people hate the web) and increasing premium sales, resulting in far higher sell through rates, but that doesn’t seem like the optimal use of the internet as a resource.

An analysis more in-line with the current evolution of Internet demand would imply that pricing needs to be more elastic, implying the use of exchanges.  The challenge with exchanges is that they work best when inventory is commoditized.  Custom advertising experiences are difficult to sell in an on-demand exchange context.

Where am I going with this?  Not sure.  But I am pretty sure that no one wins when people start saying “we should make the user experience worse.”  I am excited to watch the Internet evolve, but if I have to watch a 30-second interstitial to see it, I am going to unplug.

Today’s MediaPost covers anannouncement by Lotame with an article titled: “Lotame Mines Social Data From Facebook To Target Ads“.  In the article, they lead with:

Imagine the insight that ad agencies and publishers could gain to target consumers if they knew the exact type of person who signed into Facebook at 8 a.m. each day to update his profile and then went to The New York Times to read the news.

Lotame has released a feature in Crowd Control, dubbed Stadium, that it believes will give advertisers and publishers insight into human behavior through social media data from sites like Facebook and others to help target ads. The application has been under development for two years.

The article even ends with an analysis of how fast Facebook is growing.

But the actual Lotame sound bites don’t mention Facebook.  This sounds very different and I suspect that the reporter somehow ran with an example and turned it into an angle that doesn’t even exist.

If Lotame was licensing Facebook behaviors, that is HUGE.  Enormous pool of granular, valuable, US interest data.

Who cares about their new product, they just announced a relationship with Facebook!

I took a cursory look at Facebook and unlike Bebo, a known Lotame partner, there appears to be no Lotame tag.  As a guy who has licensed a fair amount of behavioral data in my time, I would be shocked if Lotame were able to license Facebook data.

Two lessons here:

• Lotame PR management must have blown it.  Even to mention Facebook as an example here implies something unfairly to reporters.  Someone has to catch that.
• The reporter was in a hurry and not too experienced in covering this particular nuance of the online advertising market.  Lotame was trying to push a product line extension and it turned into a Facebook data licensing deal headline.  If that were really the case, trust me, Lotame would have been pushing that.

I could be wrong here, but it would be a huge story.  Frankly, I would be pleased as punch for Andy, who is a great guy and deserves to exit for $1 billion.  Go Andy!

That is a nice, provacative title, isn’t it?

Anyway, I had an idea yesterday that made me say, “Is it really as simple as that?”

And ever since then, I have thought, “Yeah, it is.”

In managing a marketplace, your average advertising network, in many different ways, has to “choose sides”.  Are they aligned with publishers or are they aligned with advertisers when it comes to how they deal with certain issues.  One issue seems like it might be more important than any other in answering this question: Publisher payment models.

Many (most?) advertising networks pay a revenue share to publishers.  To that end, they are incented to extract the maximum revenue they can from advertisers and efficiency of the buy and optimization of the buy for advertisers is only important to the extent that they can use it to justify extracting more revenue from advertisers.  Some networks are poster boys for this model, like a Federated Media.  Some are less transparent about it, but I think it works out just the same.  Right Media is probably an example.

A few networks pay flat CPMs.  When you pay a flat CPM to a publisher, that aligns you with squeezing value out of the inventory.  That is advertiser alignment.  Figuring out how to help advertisers get more bang for their buck on that inventory suddenly is really, really important.  If you are paying $0.10 CPMs and generating $0.12 RPMs, the ability to optimize and drive that to $0.20 is much more valuable to you (creating a much bigger incentive to you) and when you do that, some advertiser is reaping the benefits of that improvement.

What networks pay flat CPMs?

I know that usually I focus this whiney blog on maligning things other people say, but I want to actually be constructive for a change.  Let’s start with maligning things: Facebook and their newsfeed interface for application developers.

They offer two ways to create newsfeeds: A template wizard or an API.  Both are only somewhat useful because templates cannot be edited.  They can only be deleted.

Facebook strongly pushes developers to use the wizard because the wizard allows them to validate templates more effectively.  Unfortunately, using the wizard is fine for template creation, but if you want to recreate a template with a small change, it requires navigating the entire wizard process again and is cumbersome – using the wizard requires creating dummy data for every field and JSON formatting it.

The API works better, but if you need to rewrite a newsfeed template, you must delete that template, create a new template using API calls, and then put that API template reference in all of the points in your code that invoke that template.

Interestingly, I would like to propose a third solution: A generalized solution that could easily be layered on top of the existing template architecture, leverage loop holes in the current architecture model, and allow easy edits to template models.

The specific loop-hole that this would exploit is that Facebook allows applications currently to invoke other applications templates.  This type-checking is probably a good idea because it allows test versions of applications and production versions of applications to share templates without requiring recreation of templates, managing template IDs between applications, and associated testing risks.

My proposal is to create a set of generalized feed templates.  For example, one template for short stories might be:

{oneline1}{actor}{oneline2}{target}{oneline3}

Obviously you would need the reverse as well:

{oneline1}{target}{oneline2}{actor}{oneline3}

You would need versions of these templates without associated short stories and full stories, and all the various n-cases.  With and without action links.  Pretty straightforward.  Publish those template IDs and all of the sudden there is a flexible, editable architecture for crafting newsfeed stories on Facebook.

I am currently using a reference implementation of this architecture and it is working dandy.  Unfortunately, I hard-coded the action links and am far too lazy to redo it.  If I get around to it, I will do that and republish this.

Am I missing anything?

AdExchanger is the hot new advertising blog.  Have no doubt.

I was reading their interview about MediaMath, and I noticed the part where they said, “we don’t view the spot market as remnant!”  This got me thinking, “that is true, but true because they take an advertiser perspective.”

Remnant is a publisher concept.  Advertisers do not buy “remnant”.  They buy cheap, they buy high-frequency, they buy run-of-x.

Furthermore, publishers would imply that advertisers are choosing between their “premium” and “remnant” inventory, but this is simply another publisher conceit.  Truly, advertisers are choosing between better impressions and worse impressions and those impressions are scattered in between the premium and remnant inventory.  The first several impressions of a unique tend to perform better.  Does buying premium get you that?  Nope.  Many publishers don’t sell frequency caps as part of their standard IOs.  Of course, frequency caps are a tiny part of the optimization equation.  Things like DFP should, theoretically, be optimizing for publisher value – of course, I think they frequently don’t really optimize for much of anything at all.

Does that mean there is room for aggregators who can algorithmically add value for advertisers?  Absolutely!  Long live the revolution!

Jeremy Liew writes a post about techniques for targeting that Google does not offer and how effective they are for performance advertisers in other situations.

Like any pathetic blogger, I must comment.  My reaction to this was not that these targeting mechanisms have demonstrated effectiveness at all.  Rather, we have discovered a key modality that has unique value to direct response advertisers: People who aren’t doing anything at the moment.

IQ Quiz is a great example of a hard core DR advertiser.  Generally, they need conversions that fit one of two profiles: Expensive clicks that convert fantastically or really, really cheap clicks that convert okay.  Search clicks typically perform poorly for these direct response advertisers because the expensive clicks don’t convert well enough.  Clicks on social networks are very, very inexpensive (Google typically turns off my bids when they get this low) and the conversion rates are relatively high due to a unique user modality: People on social networks aren’t typically doing anything important.

People on Facebook are just hanging out, far more so than people Googling things.  These people are uniquely susceptible to a “have fun, play casual games message, meet people, look attractive, check your credit score” that is used by so many direct response advertisers.  Does behavioral targeting or more accurate demographic targeting deliver great value for a DR marketer?  Certainly, but typically any sort of premium involved crushes the CPA objective of the DR advertiser.

Effective reporting by social networks could improve the targeting purchases by direct response marketers.  For example, here at Platform-A, we offer end of campaign reports to certain advertisers showing the kinds of behaviors that worked most effectively during their campaign.  This evidence-based post campaign diagnosis can drive more effective purchasing in the future.

Absent this detail, typically these kinds of advertisers are simply on the hunt for cheap clicks.  It will take more than just targeting, but great reporting and analytics to push advertisers in this direction.

When you aren’t updating org charts on Cogmap, using spreadtweet will make it look like you are doing something else productive.