Wherever it may eventually go — and my bet is it will be a mixture of pro accounts and advertising — the advertising model in Twitter might be something very innovative, both in targeting and delivery.

Targeting – Easy Semantics and The Realtime Factor

Targeting a Twitter user is very convenient. They don’t have to assume or speculate anything about a user, they don’t have to track cookies and collect Behavioral Targeting data and try to determine if a user is action prone or not, or try all the other good old targeting tricks, such as geo-targeting and figuring household income. The simple nature of Twitter messages, which is very short and informative, makes them very easy to process semantically — aside from lolspeak and l33t, sentences are simple, usually noun/pronoun-verb-adverb-adjective. And the best part is, the user gives information about himself voluntarily — location, likes and dislikes, actions performed, etc.

Take me for example, in this imaginary (but could-be-true) scenario:

• In SFO, waiting to board flight for JFK. Coffee anybody?
• Just saw the new Pearl Jam album, love them!
• @johndoe Let’s meet later this evening for dinner

Do you realize how much Twitter knows about me in the 5 minutes I tweeted these three? They can advertise to me flight tickets, coffee in SFO, Pearl Jam and similar music, and places to eat dinner in New York.

Now, you might say that Facebook or other social networks can provide similar targeting, but the realtime nature of Twitter makes it even more powerful. I might ditch Pearl Jam in a month in favor of Soundgarden, and Twitter will know that immediately. I might be in a conference and suddenly crave a steak. The relevance of the advertising is much better when my realtime wish or craving is in the equation.

Delivery – Unobtrusiveness and Flow

The way Twitter will deliver the ads will have a very high impact on the user responsiveness to the advertising. And the fact that Twitter has an open API and a lot of users using 3rd party clients to access it, will force them to embed the ads in the Twitter stream. They could be “full tweet ads” or they could be “tweet-appended ads”.

For example:

• In SFO, waiting to board flight for JFK. Coffee anybody?
  • From twitter: Drink Coffee at Starbucks@SFO!
  • From twitter: Next time, save on airfare with MyImaginaryTravelAgent!
    
• Just saw the new Pearl Jam album, love them!
• @johndoe Let’s meet later this evening for dinner

Or:

• In SFO, waiting to board flight for JFK. Coffee anybody?
• Just saw the new Pearl Jam album, love them!
• @johndoe Let’s meet later this evening for dinner
  • From johndoe: sure let’s eat a steak! (From twitter: check out the SteakHouse on 5th and 34)

I’m For It, The World Is Ready

I would not care receiving both these forms of advertising, if they are well integrated in the flow of my Twitter stream, and especially if they are so well targeted. I believe the semantics tools today are good enough to process meaning from a user’s short and simple under-140-characters sentences, since there is no hard contextual analysis to perform. And the realtime factor makes the targeting temporal-aware — they know what I need when I need it. Regardless of what payment model they choose (CPC, CPA, CPM), the targeting and delivery methods are the winners here.

There, I’ve contributed my part to the Twitter business model obsession.