Thursday, March 7, 2013

New Ways Milliseconds Can Matter

Milliseconds long have mattered for high frequency traders, since a time lead of that magnitude can translate into hundreds of thousands of dollars of extra profit on a trade.

Such algorithmic trading is handled by computers, not live humans, so trading speeds  are limited by processors, software and sometimes even the distance between two computers that are parties to executing a trade.

But Adobe now believes such differences of milliseconds might someday be important for digital marketing platforms as well. It is a bit of hyperbole at the moment.

But Adobe argues that such speed advantages may separate winners and losers, according to Brad Rencher, Adobe SVP.

Adobe is aiming to build a millisecond of a lead on the competition.  It turns out that the millisecond is the one that occurs between the last piece of data a consumer “gives“ a system and the content with which the system responds.   

What happens in that millisecond?  The system needs to correlate, manipulate, measure and analyze all the various pockets of data is has on the consumer and then choose, assemble and display the relevant content to her.  

The content that these algorithms choose and that the infrastructure renders, must encourage the consumer to take a preferred action, especially buying something.

The analogy is not quite perfect, since milliseconds are not relevant for human cognition or perception. But milliseconds might be a reasonable of describing latency as computers experience it when crunching enormous quantities of data before presenting some sort of solution in a marketing context.

In other words, in the big data environment, when evaluating data and then assembling offers, for example, milliseconds might matter when trying to find meaningful signals about what people are doing right now, where they are, and what device they are using, and what they might want.

So milliseconds might matter if that data has be correlated with what is known about the customer from all available sources of data (CRM, social) to create a more granular view of a particular person’s values, past behavior and buying preferences. Then the objective is to use that understanding to predict what content must be delivered, immediately and in context to achieve a commercial objective.

In that sense, milliseconds might matter, even if humans cannot apprehend delays that small.

That might also imply that milliseconds could matter, in new ways, for communication networks other than those supporting high frequency trading systems.

No comments:

Many Winners and Losers from Generative AI

Perhaps there is no contradiction between low historical total factor annual productivity gains and high expected generative artificial inte...