Social Unstructured Data Is Not Unusable
Written by Evan SchumanJust as certain a fact as stating that many of today’s social network sites will be gone in two years is the fact that new social sites—invariably much more niche and focused—will replace them. Hidden in plain sight within the millions of posts in dozens of languages of these huge number of sites is every trend, every individual customer profile and every hint of what customers will buy—and perhaps even their desired price range—that your chain could ever wish for. There’s only one problem: There is no simple spreadsheet-friendly way to access that data.
You can read it without limits. But to automate that process and to process the data in a way to get anything meaningful out of it, that’s difficult. We are deluged with products and services that are trying to solve problems that hardly anyone has ever experienced. Who will be the first to conquer this one? Many companies—including SAP and Oracle—are trying to figure it out. But they typically try to fall back on algorithms and filters. The software needed is closer to what the CIA and the NSA use to parse billions of phone calls and E-mail messages while trying to figure out plots. It’s much closer to artificial intelligence than cryptography. Military satellite technology eventually came to consumers in the form of GPS. How long will it take for AI to visit the local retail chain, where software will peruse the world to find out the best assortment to be displayed tomorrow?
-Marc
January 21st, 2010 at 5:58 pm
This information is hugely valuable. We analyze it for real use and action with retailers everyday. There is so much real-action insights in social media…
Some examples:
*Cries for help! – are cusotmers complaining about something online that you can answer? We find it, analyze it and route it to people to get involved in the conversation.
*Deep sentiment analysis – What do people like? What don’t they like? What is contributing to them deciding to buy or not buy?
*Product quality and issues – Is there a product or a product feature that is having an issue? Impacting product quality? Something we should alert customer service about?
Thanks for writing this!
-Michelle de Haaff
CMO Attensity Group
January 29th, 2010 at 1:23 am
Until the semantic web comes to be (if it ever does), there are ways retailers can collect and process unstructured user data. For example ratings and tags, particularly semantic tags (controlled vocabularies) do provide some level of structure. The rest can be mined using probability models, the way search engines do it. While it’s easy to overlook, Google is powered by a probability engine, not semantics.
January 29th, 2010 at 7:13 am
Fabien’s right, but that’s a lot easier for data you can influence (such as your own Facebook area or comments within your site) as opposed to things that are entirely out of your control, such as other people’s discussion forums or social pages. For that matter, social sites you’ve never heard of two continents away. For data that is unstructured and out there in many places, well, that requires yet a different strategy. But there’s gold out there–free for the taking–for those creative enough to find ways to get at it.