About 10 days ago I had an interesting twitter conversation with @wimrampen and @grahamhill about the added value of collecting loads of data about your brand, your target demographic and your customers on social networks, such as twitter, facebook, friendfeed etc. That conversation, and @martijnlinssen’s excellent blogpost on Social CRM got me thinking, and I’ve decided to write down some of my thoughts in this blogpost.
What is social CRM again?
I asked myself, what added value does collecting data about your brand or customers provide and what user groups of that data could exist within a company? Collecting data just to build a giant datastore seems rather senseless. If we look at the body of work that is currently called social CRM we can clearly identify two valuable applications of Social CRM:
- Social Service: using social networks / social media to reach out to customers that have an issue with your product or service. Using social networks provides ease of use to your customers and provides them with yet another way to get their problems resolved quickly when they need support. Provided you monitor social networks and your customers have know how to get in touch with your of course.
- Social Sales: Promoting your products online, starting competitions on social networks (facebook / twitter). We have all seen clear examples of succesful social sales campaigns using twitter (like dell, starbucks etc).
I know my explanation of what these two succesful applications are, does not do justice to the value it can deliver to companies and customers alike, but this post is not about explaining social crm and how it can benefit your company. Most social crm initiatives can be divided into the above mentioned categories however.
Social Data Mining / Gathering Social Network Data
Social Data Mining is something that has popped up a bit more recently however. Social Data Mining could be defined as scouring all kinds of social networks and raking in all kinds of data about users that tweet, post or blog about your brand/products or services. The aim is to use all that data to analyse and report on trends, customer segments, and discover possible new markets for your product.Social Data Mining could also be used to gather social network id’s for your customers, much like you collect phone numbers or addresses. Sounds terrific doesn’t it? But to me, this is all a load of bull. The data simply isn’t reliable enough to use for data mining / marketing initiatives.
Getting back to one of my initial questions, who would want to use all this information within your company? Your marketing department? Your product development department? Social Data Mining is usually heralded as a new data stream for your marketing department. Your sales and service departments are probably already happy with Social Sales, Social Service and Reputation Management tools such as Radian6 or Nimble. Let’s look at the challenges associated with using social network data for marketing and segmentation.
Besides the obvious privacy challenges, I feel that the two main challenges with social data mining are Identifiability and reliability.
Let’s look at identifiability first. So, you’ve gone ahead and bought some tools to tap into Twitter’s Firehose of tweets. You are filtering everything out based on a couple of intricate algorythms and storing them into a large database for analysis. What kind things do you need to use this data for segmentation for instance? Well, demographic information would be nice: what’s the persons age, is he/she male or female, where is she from etc. How do you figure all that out from ElRioGrande’s tweet that he simply loves to drink coca cola? The answer: you don’t. Most people have an unidentifiable social network profile because the love the anonimity of tweeting about things, without someone finding out who they are. They purposely choose a vague account name and withhold as much personal information as possible. The only provide the information they want to share with the outside world. They are in control. Pretty hard to use that firehose of tweets for segmentation if identifiability is an issue.
One other thing we like about data that we analyse is that it is reliable, we know where it comes from and we’ve identified (there’s that pesky identifiability issue again) a number of characteristics of the person talking about our product or service. I myself would love to work with data that I know is from a customer (past or current). How reliable is LittleStevie13’s tweet about the fact that he hates eating at taco bell? Is stevie actually 13? Might he be a competitor? We simply don’t know. We could use his tweet to contact him to see if we can resolve his issue with taco bell (maybe he would like his taco a bit more spicy) and offer him a free taco, but we can’t rely on it to form a basis for customer segmentation or marketing initiatives.
So, in my humble opinion, gathering all kinds of social networking data for data mining and segmentation is a load of crap. Sure, we can use it for reputation management, reaching out for service issues or selling our product, but we simply can’t use it for marketing. Perhaps in the future we will develop tools that allow us to cross reference social networking data with actual customer data to close the identifiability and reliability gap (and start up a whole new discussion about privacy implications of that cross referenced data) but right now I don’t see an added value in social data mining and if there’s a consultant or vendor knocking on your door trying to sell you social data mining tools, tell him your busy figuring out social service and social sales and to come back in about 5 years or so.
Data mining image courtesy of http://www.mbaknol.com