Byline: Rand Schulman. In the online marketing world, everyone is always asking me about real-time data, insisting they need real time. I say it’s a myth.
“Real time” is one of those things vendors and users have always been confused about. Users of web analytics tools have asked for “real-time performance” while, at the same time, not really understanding why. Vendors, ever eager to one up their competition, are willing accomplices in this little charade.
This has manifested itself in a range of real-time functionality, like “real-time alerts,” which offer a minimum level of value. Emergency, red lights flashing, conversion rate dipping below threshold: beeper goes off! Yeah, and now what do you do?
True, there are some metrics which you could collect and display that will indicate general site conditions, but beyond that, real time alerts are of minimal value. Yet we hear the request all the time from users for real-time features in our products.
So, what IS real time and HOW MUCH of it do you need to get your job done?
Let’s define our terms. Real time is defined over two dimensions: collection and display. Vendors have applied this definition to mean anything, believe it our not, from four hours, to session-based results of milliseconds to 30 minutes. Early on there was a notion that all the behavior we collected about a person, and their purchase data, could be utilized to create offers that would yield higher conversion rates (as measured by sales or leads). That promise didn’t materialize for many reasons, including inadequate databases, slow bandwidth and lack of raw processing power. Firms that held great promise like NetPerceptions, Personify and Engage largely went by the wayside due to unfulfilled user expectations.
What are some of the big issues facing us with regards to real time… and, are they moving us forward? First of all, there’s the issue of data collection versus data reporting: how much you can collect and report on in real time? Many, if not most, systems today collect page view data in real time, but now we are heading toward event-based collection. But capturing every event is just as irrelevant as capturing every cursor move. It becomes more problematic if one does not fully appreciate the orthogonality of these RIAs, Ajax, flash and streaming media events as compared to relative ease in understanding and optimizing page view-based events. One simply can’t collect and display the data in real time with any level of relevance.
Then there is the issue of endless querying and cubing of the data. How much data needs to be sliced and how quickly does one need to take action on it? Many vendors charge exorbitantly for this capability, when the requirement does not, in fact, exist. You simply can’t make this truly actionable in real time. Emergency! Change the content!
Here’s a little vendor secret. Vendors play this game because users have not been explicit about their needs. We say to ourselves, “Users are asking for real time, but what they REALLY mean is real “enough” time.”
So what have we learned? I hope we have learned to ask up front, “What do you want to do with this data?” and “HOW will you take action with it? Is it manual or automatic? What other systems of record are involved? How will I measure the uplift and results?”
The collection and display of real time data is very important in some call center applications where the offer can be targeted to the caller during the call. There are other uses of real time data, based on aggregated information and displayed as “the wisdom of the crowds.” Product search results are a good example. Beyond that, there are few examples of real-time personalized data that are practical for increasing and optimizing conversion events -- unless you’re okay with four hours as “real time” or confine results to the changing of only two or three objects, frames, or blocks of content. I’d like to hear your real time feedback.