Social Media Root Cause Correlation

I think that it’s pretty much common knowledge now that we are working on permission based correlation system for social media, although many people won’t understand what that is, developers and engineers will be very familiar with correlation. Permission based correlation systems for social media platforms potentially have the option to solve the number one issue faced by publishers, e-publishers and everyone else embarking on the social media journey and that’s:

How to make money with Social Media

This is the one problem we are out to fix with our products and services, we are of course building out our advertising network of long tail social media sites but we have now deployed a number of products that allow operators of social communities to actively monetize their social community assets using our promotions, location based services, paid for classifieds and competition marketing. Ok so how does that relate to correlation -

What is correlation?

Data correlation is the process of understanding how data attributes are related, its normally used by banks and telecommunications companies to wade through vast amounts of data to find patterns, trends etc. When applied to social media you can define a rule that drives your advertising serving. For example let’s say you are John and you just placed a BMW 3 series for sale on a social media classifieds portal – at this point in time the correlation system says > JOHN = Selling Car = BMW = Location Dubai = Current location Abu Dhabi. Now if you are a Mercedes Garage in Abu Dhabi wouldn’t be an ideal time to show John your advertisement or send him a message or even better offer John a Free Test Drive? Of course it would – that’s the power of correlation. But >

Why permission is based correlation important?

John of course may not want your message, why would John want to be bombarded with advertising messages for something that he just doesn’t want – that’s where profile management has to come to play – its Johns personal preferences that should detail exactly what type of offers, ads and promotions he gets either 1) based on his profile or 2) if he wants based on his past or current activity, if using location based services Johns preferences also need to take into account when he wants to receive messages, for example between the hours of 9am and 9pm and when he is within 1km of an advertiser– Hey Presto thanks to correlation John now has time, activity and location based opt in advertising

Understanding Root Causes and Symptoms

Any engineering buff reading this will be asking “Isn’t this a simple SQL query?” In effect yes, running an SQL query automatically would give you the results, but of course you can’t run SQL queries across your database all of the time, it simply wouldn’t scale there is TOO MUCH data, so to apply correlation accurately one should bring an element of Root Cause to the table. Root Cause Correlation (RCA) is the method of understanding what the cause of the effects is, this can be a little hard to understand. Let’s go back to John, we know that he recently put his BMW up for sale, and quite rightly we made an association that John could be in the market for a new car. However what if John was selling his car because he was moving country from Dubai back to the UK? The ROOT CAUSE of John selling his car, placing his furniture for sale, booking a flight ticket etc is that John is moving to the UK. The fact that he is selling his car, furniture and booking a flight etc are symptoms of the root cause; so conceivably it is possible to scan a database for “Pattern’s of symptoms” that infer a root cause. By defining a root cause you can now provide John with all of the relevant services he may be interested in – in this case a flight ticket. Scanning for these symptoms requires an application that can scan multiple database instances concurrently, simply because John’s information is likely to be held in a number of social media applications. Associating John as an individual across multiple social community platforms is pretty tricky - it requires multiple applications either to have a single sign on identity for multiple platforms or for you to resolve the information on one common attribute. This attribute is normally Johns Email address.

What does this mean for people’s privacy?

What this means for your privacy is that using pattern matching correlation algorithms owners of social community applications understand exactly what you are up to. How they use the data is dependent on their privacy policy, this is why its critical for any user to read the community privacy policy prior to joining a social community. Within the H2O social communities we are planning to let users set their own privacy policy detailing how their data will be used.

Where is this going?

Where this is going is that publishers will require root cause analysis tools to deliver super targeted advertising to their customers. A publishers advertising products will need to be much more targeted to drive advertiser value in the future, its these extremely targeted high value applications for advertisers that will make up the shortfall in reduced print advertising revenue in the future. These are the technology applications that we are currently building, check our CRM & Social Media pages on our corporate site for more information.

For the nerds :-) Here is some relevant Google Tech information on the subject - I love this crap !