Twitter is doing two things with this:
1) Writing their own applications such as the clients on the web, mobile clients on iOS and Android, and I believe a host of other clients for other devices that use the Twitter protocol in the way the company currently envisions their network being used
2) Providing an API for other application developers to create their own applications, that use Twitter data as well as data from other sources (such as Facebook, RSS, photos, videos) to create rich applications.
There's a limited number of #1s and lot of #2s. Though Twitter's own applications will always have a lot of traffic, I think we will see the number of applications in category #2 continue to increase as the ecosystem figures itself out.
There's a lot of speculation that Twitter will start to put ads directly into the stream based on what the user is interested in, placement of which will probably be based largely on which Twitter accounts the user follows. As twitter wraps more data to go through their own redirectors at t.co, I imagine it will also be based on what the user clicks on.
If you go to t.co it says:
Twitter uses the t.co domain as part of a service to protect users from harmful activity, to provide value for the developer ecosystem, and as a quality signal for surfacing relevant, interesting Tweets.
It's the last part that is most relevant. Remember, whenever anyone lists three reasons, the first two are just filler.
So, while I'm sure that Twitter will experiment with putting interest based ads into the stream, I'm somewhat less confident that direct response based advertisers will find the click through rates and conversion rates they would hope for there.
Brand and awareness advertisers may find more success here, but their usual medium is through display advertising such including rich media units with audio and video. Those type of units in-stream where the payload is 140 characters just don't make a lot of sense to the user experience.
Integrated advertising is certainly all the rage, but I think it's difficult to put integrated advertising into a protocol, especially when the applications that use the protocol are myriad and vary greatly in presentation.
I actually think the right approach in the end will be not to monetize the protocol, but the applications that use the protocol.
My hunch comes from looking at a lot of "in-stream" data over the years in RSS feeds that ran through the FeedBurner Ad Network and Google's AdSense for feeds products, which share some of the same characteristics from a user's point of view.
Performance of ads varies greatly from application to application and it's naive to think that a one size fits all approach in the stream is going to give user or advertisers exactly what they want.
A lot of this has to do with user intent at the time they are interacting with the content, as well as the mode and mobility of the user. So an ad you put in the stream better be different on the desktop than it is in the mobile client which must also be different in a third party application such as Flipboard if you want advertisers to keep coming back for more.
I think ads in the stream will just be part of the story.
So back to the title of this post which is "Why Twitter should create their own ad network" and what I mean by that.
I think for Twitter the right approach to monetization should be to look at all the different use cases of client applications that use the Twitter protocol, and build appropriate ad units for ALL those specific types of applications. They should of course use these units in their own applications but then also allow developers to embed the best ad unit for the developer's application with a revenue share.
So for instance, I would slice up the ecosystem into Web Stream Readers, Tablet Magazine Applications, Small Screen Phone apps, Television Applications, Vertical News Aggregators, Curation Applications, Photo Sharing Applications and everything else that the Twitter Business Development team has seen come through the office - and then build appropriate ad units for those applications with the right offerings for advertisers on the back end.
There should be something for mobile local transactional advertisers, as well brand display advertisers. but it shouldn't be one size fits all.
Specialized advertising networks always outperform generalized advertising networks. This is why at Google, each particular advertising product that exists in it's own medium and mode has it's own full time product manager (sometimes many!) whose job it is to analyze the specific data for their product and then tune and optimize the heck out of it with their engineering team(s) to optimize revenue and performance.
Twitter should organize the same way and on the back end of this all should be their own relevance engine, which uses all the available signals, such as the aforementioned list of subscription topics and click data, to create the best advertising product for each specific application use case.
There's nothing stopping a third party from creating this network, but I believe Twitter's strategy should be to simply create the best ad network that outperform's any potential competitor.
They are in the best position to be able to create this network as they have the most complete view of all the data and applications that are in the pipeline. Their own applications will serve as a proxy to performance for each type of unit, and then they will also have data signals from all their developer's applications that use the network.
If they can start to pump advertiser money back into the ecosystem, it will fuel development and innovation even further in that same ecosystem. Not to mention silence the pundits.