Mike Rowehl: This is Mobility

I’m getting a bunch of pings from all sorts of advertising folks (probably because I’ve worked in the past at places like Overture and AdMob, worked in mobile for a while, and my current company is in the process of getting sold off). First of all let me say there’s nothing out there that I’m interested in on that end. Not least of all because I’m still involved with AdMob and there would be some nasty conflict of interest stuff going on if I were even talking to competing ad networks. But primarily because there seems to be almost as much cluelessness in this round of media expansion as there was in the last. I was working for a company doing primarily banner advertising around 1998, so that’s my basis for comparison - and we all know how that turned out. Allow me to elaborate on “cluelessness” a bit if I may.

I think the current round of confusion is being driven primarily by the amazingly high value that Facebook is commanding and the mistaken belief that that value is associated with the information that Facebook has about their users and the corresponding margin that supposedly translates to with advertisers. Just untrue top to bottom. Take this advertisement which has been darkening the valuable rightmost column of my Facebook pages for nigh on 6 months now:

If they were really paying that much attention to media rotation and customer profiling and advertiser fill they would have way more interesting stuff to put in there for me. None of this is to say that Facebook isn’t worth their enormous valuation, just that I don’t think that user data and advertising is the reason for it.

However the outcome of stuff like this, and other commentary floating around about mobile devices being able to provide the ultimate in personalized media (because they’re always on, always connected, one-person devices loaded with a wealth of information about their user) is that folks link that “the more I know about my users the more valuable I am as a media property.” That’s an awfully dangerous generalization. To see why allow me enumerate the basic models of advertising and how they relate advertisers (just the basics, there are enough shades of gray in there already):

• CPM - which stands for cost per mille, or cost per thousand impressions. Not cost per million, like everyone would assume. Which is the first bit of evidence that lots of advertising is based on some rather twisted and introverted opaque set of dogma if even the base terminology starts off confusing. If you’re running CPM stuff someone is paying you some set amount of money for every thousand times you show their ad. When advertisers are paying you for CPM advertising they love to know information about the people they’re advertising to, so knowing a lot about your users will help here - IF (and this is a big big IF) you have enough users displaying any given trait to make it worthwhile for the advertiser to buy against. The first part is easy: “I know I have 15 people in Kentucky who like to wear red shoes when they go line dancing.” The second part is hard, who gives a crap about that trait? Do they care enough to try to reach 15 people? Are they willing to pay you enough to reach them that it’s worth your while? See all the questions that come out of this? You’ve got to think these things through before you say “the more we know about our customers the more we can make in advertising,” cause for the most part you’re wrong.
• CPC - which stands for cost per click, just like you would expect it to. Whew. Advertisers running CPM ads eventually said “this is stupid, why do I pay you to show stuff as if I were advertising on a billboard, when this whole Internet thing is supposed to be about interactivity and measurement?” Thus was born CPC, where your advertisers pay you every time someone clicks on one of their ads. Hmmm. So how do you use all that information you have about customers here? Well you can ask your advertisers to fill out a bunch of info about who they think would respond best to their ads and to describe their services and offerings. But ultimately most advertisers don’t care, if they’re only paying when someone clicks they really don’t need to do your job for you. So when you’re selling CPC your customer information is useful TO YOU the property owner, but only when combined with the knowledge and algorithms to optimize the inventory you have (and a deep network of advertisers).
• CPA - and we’re back to slippery terminology, CPA stands for something like cost per action, or cost per acquisition. Advertisers again said “hey, why am I paying someone to drive clicks to some random webpage when what I really want is to sell widgets?” and thus was born the concept of CPA, where you as a property owner get paid once one of your users goes somewhere else and performs some desired action. Once again, like CPC the advertiser no longer really cares about demographics and user info. Hell, they care even less now. If the user is going to buy something on the advertiser site, great. The advertiser normally doesn’t care if the person is in Palo Alto or Pakistan (barring weapons classification of strong crypto of course, but that’s a different story).

See, as you climb up the scale of advertiser value the burden of using all that customer information shifts from the advertiser to the property owner. The game starts to come down to automated systems of optimization and segmentation and having enough inventory and fill to be able to shift values around in search of optimal methods for running the property. The inventory number requirements and complexity of response optimization really make this stuff best suited to a network of advertisers and properties working in unison - which is why Google and AdMob are doing so well. It’s a hard nut to crack for all but the largest media companies, and even they tend to bank as much on the demographic skew of their audience as deep information about their users (and still only directly rep part of their inventory frequently and backfill with general networks or piggyback on external sales). So being part of a network is the best way to make sure you have access to advertisers and the kinds of algorithms that drive pageview value, but it also restricts how you can use all that information you have about your users. Does not compute.

Right now I’m just not buying the overall pitch for much of the expansion of mobile advertising. Much of it seems like technologists building systems for their own sake without paying much attention to how the whole thing fits into the overall market - or even what the real desires of advertisers are. All these little niche networks keep popping up all over the place, trying to substitute depth of information about users for reach. That doesn’t work. You’ve got to have reach first, no matter how much information you have about your users, doesn’t matter. And you want user info, sure. That helps you more effectively manage the inventory you have and maximize yield, but it doesn’t drive the cost to the advertisers.

If you’re building a company planning to make money through advertising you can not count on depth of user information to allow you to command a premium unless you have a corresponding plan to reach large groups of motivated advertisers with a pressing need to reach exactly your users. When you talk to folks and they ask “so who are your users?” that’s the reason they’re asking. As technologists we tend to think the answer “Everyone! Anyone can sign up and use it” would be the natural answer all should be happy with. Definitely not.

Now, all of that being said, does that mean that all this jazz about always-on connected mobile devices being the way to the future of media is just a load of crap? No. Just that it’s going to take a model shift for that kind of customization and leverage of information to make it out to mass use. Someone or something is going to have to change advertisers minds about the fundamental way to interact with their potential customers and engage with an audience. We’ve been talking about that for a decade with the web already, and every few months we discover the next great new game changer. Over and over. And for reasons like that if you step back and look at the total number of dollars spent on advertising, the amount of adverting spend going into online vs offline is still pretty small.

So that’s it, you can stop calling me about mobile advertising networks now.

I used LinkedIn to try to help organize all the folks I run into at events and conferences, but recently Plaxo has really won me over. I love the ability to sync to local applications and manage my calendar, which I can sync with my phone as well. However, like everyone else, I have to struggle with the relative isolated nature of these online services. People still send me stuff through LinkedIn, and I don’t want to tell them to move over to Plaxo, cause why do they give a crap what I say about their use of online services?

Over the last week I’ve gotten a ton of pings over LinkedIn, which is great. Thanks to everyone who’s pinged me with interest or offers. However, the last few days when I’ve hit LinkedIn it’s been unusably slow. I thought maybe it was me, I’ve been mostly on a cellular connection for the last few days. But today I’m sitting attached to my high speed connection at home and LinkedIn pages are still failing to load. So sorry, but if you’ve pinged me through LinkedIn please do so directly instead. I might have to just ditch my LinkedIn account all together. You can get me at mike at rowehl dot com.

The reaction to Russ’s post about shutting down Mowser and his thoughts on the mobile web have me in a bit of a tizzy. Russ is a good friend, and I understand where his thoughts are coming from, so I’m not going to attack him. I would however like to bitch at the rest of the Internet for being unthinking sheep on the whole.

The reaction to Russ’s post seems to be “Oh god no! The mobile internet is dead!” For one of the folks who was a vocal advocate of the web on small devices to say that developing for low end devices is a waste of time, well then the whole game must be over and everyone should pack up their tents and go home. Come on! Here’s an example of something I like to call “independent thought”, all of you out there might like to try this exercise on your own.

First is to try to extract some core points from the stuff that’s laid down in the post. Try to draw some boundaries around what exactly it is that seems to be the issue with Mowser. First of all we were aimed at lower capability devices, the average handset you find in pockets from San Francisco to Mumbai. That meant we weren’t doing anything special for higher capability devices like the iPhone or recent N and E series Nokia devices. While that increases the audience you can address, it also takes a lot of the whiz-bang WOW factor out of the initial user experience for the power users. It also means that we were positioned against the current market movement, which sees more and more high capability devices out there. Does that mean that a content adaption service would never be able to survive in the world of high capability handsets? NO! It just means that two guys working out of their homes and trying to cover all aspects of the business weren’t able to come up with a method that covered high capability devices with a pleasant experience and served low capability devices adequately.

As far as the fact that many of the requests through Mowser were porn related, that’s not a unique issue, and certainly not “a problem”. Carrier networks try to filter out porn and restrict what people can do. Give them an open system and they’re going to go out looking for what they want. And if they want hot girl on girl action, hell, that’s what we serve up. Takeaway for that is that there’s huge potential in adult mobile content still, not all the business models have been explored there apparently. Does that means that “everyone browsing the mobile web” is a pervert? No, it doesn’t mean that at all. Does it mean that the only stuff you can put on the mobile web and get an audience is adult content? No. But it means that given the methods we used to grow organic traffic we got a lot more adult related searches than other terms. You can draw no other conclusions from it, and to do otherwise would be stupid.

Third is the monetization. Making significant scratch from a mobile website run as a media company is still pretty hard. The availability of mobile advertising networks has lowered the bar quite a bit. But not enough so that an adaption engine with monetization on interstitial pages and throwoff traffic on a directory of links is a no-brainer. We had too few methods of monetization build into Mowser, too few “advertising positions per pageview”, and hadn’t made it over the traffic hump where trying to represent any subsection of our property as a higher value media buy could have increased our margins. Because we did have lots of traffic from places like India and South Africa, which are less hotly contested in the advertising networks, much of the advertising we had a relatively low payout per thousand impressions (eCPM in the biz). We didn’t have a good mechanism for profiting off that traffic more directly, though I’m sure the right person would be able to find a way to make that work.

Finally there’s the team. Two guys who have been around mobile for years, very familiar with the web and current trends in online services, and well connected to the existing ecosystem. But also not sales and business development geniuses. A lot of this stuff fell to me, and I sent out a bunch of emails and called up the folks I was able to call up to pitch the idea. Some of them signed on, which was great. Mostly they were existing mobile publishers interested in linking out to resources on the web. But I wasn’t able to land any big online content networks. Lighting up About.com for instance with a mobile version for all their pages, or Wordpress.com. Getting integrated to mobilize outbound links from Facebook mobile, or MySpace, or LinkedIn. Someone with a better background than me might have been able to make those kinds of deals. Would it have changed everything? I’m not sure, but it’s an area of execution we failed at which could have shifted the balance. Should we have dumped the consumer focused side of things and instead sold transcoding as a service charged per thousand requests through the adaption engine? Should we have been looking to license this thing right off the bat? There were other methods we didn’t explore at all.

So take that all together and try to draw some conclusions out of it. Here are a few that pop to mind:

• If low capability devices are shrinking in overall market share it’s going to take more than minimal content stripping to make a compelling mobile experience. You should be planning for iPhone style devices to become more the normal and also figure out what needs to be done to service them. That means you probably need more than one engineer working on the problem, and should be looking at folks with traditional web experience in addition to those who know the details of mobile specific markup or you’ll paint yourself into a corner.
• You can’t represent a general adaption network as a premium property unless you find a way to segment. You can’t segment until you have a lot of traffic. Either try to grow in niches, starting in areas where you know you can monetize well and have models other than general network fill available. Or plan to run at a loss for a while and have a strong model for growth you can measure, and make sure you have control over it. “Google sends us traffic” does not count, especially in mobile. They can be sending you 100 thousand searches one day, and 10k the next. That’s their right as a search provider, they’re tuning as they learn as well. You need to have direct control over a majority of your traffic if your traffic is the way you make money.
• Selling the idea of making a mobile version to someone making money off web advertising is a very hard sell. They see small incremental revenue and a lot of technical complexity. Selling the idea of a free service that allows an existing mobile site owner to link to additional content, especially if they can profit from those links - that’s much easier. But even given that I wasn’t able to walk into LinkedIn and say “Hey, I have a great service for your mobile site that would allow you to include the links from the web profile pages in the mobile profile pages as well” and make that sale. I’m sure there are thousands of people who could. If you’re going to do those kinds of things, find those people early on. That means you have to be able to pick a strategy and commit to it.

See, look at that. Concrete helpful points instead of sensationalist hand waving. This was fun, thank you for joining me. I recommend trying this method of “thinking” whenever you can. Right now I’m going back to growing the mobile ecosystem wherever I can.

Russ posted already that we’re shutting down Mowser. I was chatting with him earlier, and we’re looking to sell what we have already rather than just toss it all. I still think it’s a great service that could fulfill a pressing need for a bunch of mobile sites out there. I’m not nearly as down on the idea of Mowser and the future of the mobile web as Russ is. Given the licensing fees associated with a few of the other adaption engines out there focused on mobile we could probably sell it for enough to get Russ out of debt and at the same time save some service provider a ton of money. If you’re interested let me know. miker at mowser dot com.

I was mulling some of the geo hacks for the N810 that are out there now. Maemo Mapper is a great open source mapping application, there’s a little app that geocodes photos as well. Then there’s Maemo WordPy for posting to Wordpress, and I was wondering if that allowed for geocoding posts. And I was pondering the user of the N810 as a geocontent production device. As well as wondering if the geoaware primitives we could use in mobile browsers would at all be helped by the evolving state of mobile Firefox.

All the little hacks on the N810 could really be solved more easily on the web if there were a way to hook the stuff together. What I started thinking about was hacking around with the new firefox release and see if I could get it to shove geoinfo into the outgoing headers. But then I realized most of the stuff I wanted to fool around with wouldn’t take the headers in anyway. So for now instead I made just a simple little python launcher app that pulls your current location from the GPS and launches a browser with Google Maps pointed at your current location. Very simple, but I imagine with some basic URL crafting you can use it to create geocoded Wordpress posts or geotag images uploaded to flickr. Maybe I can make it a little homescreen applet to display your location and launch one of a number of sites with your location fed in.

Thinking about the way the web facing geographic services have worked out, passing a URL with the location filled in seems to make more sense right now. There was a geo-headers ietf draft floating around at one point.. but I can’t find that as an official version. Are there services out there that use it? Or something else that’s common across services?

A few things I want to get down as much so that I know where to find them next time I’m looking them as anything else.

Installing Ubuntu using the version of Parallels I have a license for is a pain. My installs kept hanging and erroring out. Then I found a post that pointed out a few things, I needed to use 7.04 instead of the latest, and I had to tell Parallels that I was installing Solaris. Also very useful: how to change the resolution once you have it installed. I have it set to native resolution of my MacBook so that when I run fullscreen it looks like I’m running Ubuntu all on its own. Not fantastic, but cheaper than getting a laptop that actually works with Linux. I’ll have to pick up a new Thinkpad some time soon.

There’s a funky startup problem with Tomcat running under IDEA on Linux. It’s easy enough to solve, just add an extra environment variable when running the catalina.sh. Bit of a pain if you don’t know about the problem and are trying to figure out why a newly configured project isn’t running.

Even if you have the right repositories installed under OS2008, the GNU native toolchain doesn’t appear in the application manager. It does show up if you apt-cache search gcc however, and you can install it from the command line no problem. Don’t forget the libc dev headers don’t come down by default either, you’ll have to install them as well for just about anything.

Inspired by some screenshots that came through my feed reader the other day, I updated the firmware on my N810 and installed a few new bits of hackery:

Those are transparent homescreen apps, and more important than looking slick (if that’s possible) they’re written in python! Homescreen hackery here I come!

Getting the packages installed is a bit of a pain. They don’t one click install unless you install two packages by hand. I had to download python-hildondesktop and hildon-desktop-python loader from this site in order to be able to install and use them. And yes, download and install manually, the app installer doesn’t like them. Download them to your tablet somewhere, become root (I do that with an ssh to localhost myself), and run dpkg -i for both of them. Then you’ll be able to install the actual applets and enable them. w00t!

There’s a talk from Alex Russell about mobile web development up on IT Conversations. The title of the talk is “AJAX and Mobility”, which is actually misleading (as is my own title, since Alex isn’t really a Javascript developer only, he’s got experience all up and down the stack). He’s actually cautioning against going toward AJAX on mobile devices, having solid base primitives that really support the unique capabilities of a mobile device are way more important than trying to replicate AJAX.

His commentary is right on, and refreshing to hear it coming from someone who does a lot of Javascript online right now. Sometimes I think I’m insane for saying the mobile web isn’t ready for AJAX if the base functions don’t work yet. Maybe I’m being too conservative and I’ve somehow missed the boat on a new iteration of engineering on the Wild Wild Web. Maybe I’ve gotten too old. But then you see a few great new efforts flame out under the weight of their own development process and user base, and reality start to creep back in. Seeing an experienced developer or two express the same sentiments helps ease my nerves as well.

If you think I’m full of hot air for saying some of the things I saw, try out Alex’s take instead. His points about not having a reusable endpoint for application service on a phone is interesting. You have a phone number for voice and SMS, and you have a URL for web services, but there’s nothing really the same for installable applications (not yet at least, this is one of the reasons GetJar is interesting). I also thought his list of features the browser needs to grow directly were very interesting:

• Access to location information
• Access to the voice channel
• Access to the camera
• Access to contacts/address book

I keep hearing that list over and over again. Frequently the reason stated for not including the stuff is “security issues”, which is hard to argue against because so many folks currently would gladly give up tons of freedom in exchange for a little safety. But there’s no real reason for that. If I as a user want to send my location information to a service the system should allow that, and claiming security as the reason not to provide it is just a cop-out.

The point I disagree with Alex on is open source in mobile making a difference. I do see potential for folks like OpenMoko and the Firefox Mobile effort to really shift the playing field. If for no other reason than they can provide platforms that fix those four deficiencies in current mobile browsers. And once the mobile web is really setup to be able to win, things will tip in that direction. Users will bother to download and app and keep using it if that app is a genuinely advanced browser that gives them access to a whole bunch of other applications out on the web. Users will get a handset so that they can share their data online the way they want if the applications are out there to do so. I see open source as the only way to bootstrap that process. The carriers aren’t going to do it, existing manufacturers aren’t going to do it, and web app owners can’t do it on their own.

Listen to the first question from the audience too, it’s a gem. Some guy asking vaguely “On Windows Mobile with a GPS you can already use .NET to expose an interface to Javascript you can use from IE, why does location need to be built into the browser?” Ah, the world of mobile. Some people are just so beat down they’re completely incapable of thinking through anything new.

There’s a great post at TabletBlog critiquing the user interface of OS2008, the most recent revision of the operating system that runs on the Nokia internet tablets. To anyone who’s used both devices the points really ring true. Great writeup!

I’m a diehard Linux fan, and a mobile Linux fan in particular. My N810 is with me practically all the time. And people frequently ask me “Hey, what is that thing?” when I’m wandering around with it. Pretty frequently followed by “You think I should get one?” when I start gushing about the web browser, Python and Ruby packages, SSH, and how useful of a device it is for me. But my response is still “Well, are you a developer?” when they ask if they should get one. It’s just not a consumer device, not yet at least.

In theory us developers should be able to mold this open source platform and help make it something capable of appealing to the mass market. It takes time mind you, desktop Linux systems are only starting to make some real headway (thanks mostly to Ubuntu) after years of us geeks hacking around on the stuff. But once you have enough people pointed at an open platform interesting things do start to happen.

Part of the problem in OS2008 however is that the base environment and the bundled firmware applications are still closed source. If I end up disappointed in something like the RSS reader applet functionality I have options. The system is fully programmable so I could make my own RSS reader and create an applet for that which does what I want. It’s possible for me to do that. However if I had the source code to the bundled app the barrier to me trying out the configuration I want for my homescreen would be a lot lower.

I love the platform, and right now the N810 is a killer geek device for me. A real handheld Linux device that I can hack on, fantastic! However in the face of the platforms coming out I think they’re just going to get bowled over unless they do something to leverage the community they have to help them address a wider audience.

While I was upgrading to Wordpress 2.5 this morning I happened across this post I noted but never finished. So it’s a bit less timely, but the overall issues are still interesting.

Looks like modular handset manufacturer Modu is raising a bunch of money. It’s an interesting concept, but certainly not new. Back at the start of 2004 I was taking a look at a company called Antelope who had a product called a Modular Computing Core that promised much the same thing.

The normal argument around getting your data and service where you want them when you want them goes something like this in mobile:

1. Person 1 says that the problem is best solved by a set of services that run on each device/computer/gadget and allow you to synchronize your data and preferences information so that you don’t have to keep entering stuff over and over again.
2. Person 2 says that increasing network coverage and speed coupled with decreasing access costs and competition between access mechanisms for share of user traffic will result in an “always on” infrastructure that makes syncing unnecessary. Just put the data and preferences up in the cloud and always access them realtime online. Surely the network will be acceptable for any use like this before a realistic set of services and clients to sync become available.
3. Person 1 says something like “Bah, SyncML clients come installed on all kinds of handsets already.” And then they start bashing the networks as still being unreliable and unable to move large amounts of data for an effective cost and at the necessary speed.
4. Person 2 says “Ha! SyncML has been around forever and has yet to yield a workable direct consumer application.” Then something about it doing even less than Bluetooth to foster cross-service information semantics, which make it unworkable as a base for large scale cross-application syncing let alone cross-domain syncing.

The discussion always goes on from there, is always circular, and never reaches any conclusion. Normally I wander away by the second round if I haven’t already slipped out by the first round. Which isn’t to say that interesting work isn’t being done on both sides of the equation. Funambol writes some kick ass open source software that makes it more and more realistic to use SyncML every day. However, I still don’t. The Nokia implementation is broken in some way or something of the sort. I would have to setup my own server to get it working, and many of the services I would be interested in using are marked as very beta. And on the other side efforts like OpenID and OAuth are making it more likely that I can just store stuff in the cloud and use it where I like. But even basic interoperability for things like preferences and personalization don’t seem to have made it into general use. And while some of the efforts used to have a focus on mobile, very little seems to have even a consideration for mobile any more.

So in that context a modular computing core is interesting. It’s kinda like flipping both camps the bird and saying “I don’t think either of you are going to get your shit together in any reasonable timeframe, so I’m going to create a way to solve this problem completely in hardware.” Not sure I agree with it, software systems enjoy economies of scale and flexibility that hardware based solutions don’t. But in the meantime, it’s a novel way to endrun both issues and get your stuff where you want it when you want it.