How can predictive analytics leverage social customer behavior?
Predictive analytics, by definition, deals with analysis of vast amount of historical data to capture relationships between variables and use it to predict future trends and behavior patterns. These predictions can be used to recommend actions that will drive outcomes to meet business goals.
While predictive analytics has a wide range of applicability, from churn management to budgeting and forecasting, one of its most popular application is in the marketing process to improve efficiency of campaigns and identify cross/Up-sell opportunities. So how is social customer behavior impacting these typical applications?
Most predictive models are based on the vast amount of historical data already residing with the organization to predict recurring customer behavior. Social media interactions and transactions are disrupting these complex models by adding new/unknown dimensions of social customer behavior. Social web is highly dynamic, with rapid viral spread of news, trends, fads. These can rise, evolve and vanish rapidly without giving the traditional predictive model a chance to factor them into its analysis.
It has become necessary for the organizations to recognize and capitalize on the changing customer behavior. Else they run the risk of their predictions and forecasts becoming less and less accurate.
One way is to tune its conventional predictive models to factor Social media metrics into the modeling. Eg: A customer who is actively engaged on the product's community, follows the brand on twitter and is a fan on Facebook, may have a less churn probability than a customer who is not. This higher level of customer engagement with the brand may be a representation of improved loyalty. Therefore, the churn management model needs to be tweaked to include these (and possibly other) social media parameters.
There are a few challenges with incorporating social media parameters into predictive models.
First challenge is the lack of data availability. Being a relatively new phenomenon, the amount of social behavior data being tracked by an organization is likely to be low. Also, not all customers are 'socially enabled' leading to data availability only for a subset of customers.
Second challenge is the changing nature of social media itself. Today, customer activity on communities may be one way of measuring engagement. Tomorrow, there will be other social media channels that may become more popular in terms of customer usage.
Fine tuning the predictive models to incorporate social customer behavior is certainly not an easy task. It would be an iterative and incremental approach. However, given the power of social media, organizations can hardly choose to ignore these customer behavior influencers.
Comments
Whilst there is some logic in this, it is also fair to say that the only time many customers will interact via Social CRM type tools (if that is even the best way of describing it) is if they are either asking a question - where they might indeed use a community belonging to a vendor or when they have something critical to say. Either way, the vendor community may be used - such as the above mentioned Facebook page etc BUT people who have a grievance are much more likely to just shout it. Via Twitter, their own Facebook stream or wherever. We vendors have to be very aware that we can't control where people talk about us. As such, the predictive usefulness of social data may grow of course. It may and probably will have increased usability. What it won't do however is cater for the issues such as Toyota had last year - no way of predicting that through historical data and it won't tell you what people are going to buy - only what they previously bought. This needs to be thought through very carefully before investment.
Posted by: Chris Butler | January 3, 2011 6:07 PM
The article gives a good insight at high level. Though, it is still time before organizations will actually start building models using social data.
As of now, the market is still gearing up for social CRM, but yes thought process of this level surely will help organizations to look ahead and see the sustainability model of social CRM.
Posted by: Nishith | January 4, 2011 10:26 AM
Thank you for your comments. While customers are extensively using Social networking sites to get information and raise issues, they are also expressing their preferences, likes, dislikes, product ratings etc. This information is being tracked by enterprises and possibly integrated into their customer master data. It will certainly take some more time for social behavior data to be of significant use in predictive analytics. Also, it may be more useful for certain predictions such as loyalty/churn vs others such as inventory management as Chris has rightly pointed out. I will post some thoughts regarding the same in the next few weeks.
Posted by: Shyamalee Samvatsar
| January 6, 2011 12:14 PM
I think the idea of using predicitive analytics tools for customer engagements in the social media is interesting but a little pre mature at this time. Traditional predicitive tools are largely based on analysis of large amounts historical data which is present in plenty in more traditional channels like voice, email and where it's easier to store such interactions. Customer social interactions and behaviours have new dimensions altogether (sentiments etc), and also unique identificaion of each customer in various social media sites itself remains a challenge for many enterprise customers. it will be interesting to see developments in this area. Thanks for posting this interesting topic
Posted by: ruchin | January 10, 2011 6:26 AM
I believe predictive in social will come from integration of various social media / mobility applications. For example, today most of the socio- mobile applications are point solutions and capture " in the moment" customer behavior/ interest, integration of such varied applications will help form a pattern to the consumer behvior and hence "predict" consumer actions.
Posted by: Khushbu | January 11, 2011 10:11 AM