Before Christmas I sent out a couple of obscure tweets referring to some work I was doing with our friends at SugarCRM for the IBM Connect (#ibmconnect) and Lotusphere (#ls12) conferences in Orlando later this month. In this blog post I wanted to share some specifics about this work and explain why I believe a generalized social recommendation infrastructure delivers a more productive way to effectively integrate knowledge into business processes.
This project will be demoing within the IBM Innovation Lab, alongside some other really cool research and emerging technology projects, so for anyone attending make sure to stop by. We are at the Dolphin, Asia 3.
The business problem we are trying to solve is very simple. We want to deliver accurate and highly relevant recommendations to help the sales organization most effectively close an opportunity; from finding experts to identifying similar opportunities and locating valuable assets that will maximize their chances of success. That’s it! The technical solution is a wee bit more involved and will incorporate a combination of content analytics, socio-semantic modelling, and social network analytics.
To deliver the required level of accuracy we need to model the end-to-end business process associated with closing an opportunity. This allows us to take multiple dimensions of an opportunity into account when making our recommendation calculations. For example;
- The client, industry and location, meetings related to the opportunity, products and potential competitors, and topic mentions within any related collateral, such as meeting minutes, proposal documents, activities and tasks, etc.
- Since we are modelling the business process and not just the data model it allows us to incorporate temporal information, such as the status of an opportunity in time.
- Our social network approach to this problem allows us to integrate the people interactions, such as who was involved in an opportunity, in what capacity, how actively, at what phases of the engagement, etc. As we include more and more people interactions across applications, we can make better inferences about both the people and the activities they are involved in.
- Finally, we have added some social features, such as people recommendations, which allow us to further build out the social network and incorporate feedback from the people involved in the opportunities.
But talk is cheap, so lets put our money where our mouth is and describe what we are concretely doing to realize the objectives as outlined above.
- We are harvesting the social network from the SugarCRM system, extracting explicit relationships from the underlying schema, inferring new ones, and adding some explicit social ones — most immediately a “recommended” relationship.
- We are then using our IBM Connections social indexer to merge these relationships with those already part of the IBM Connections social platform.
- We then use our IBM Connections recommendation system in order to provide us recommendations (people, opportunity, content, etc.) that take into consideration the combined social network from both the Connections collaboration system and the Sugar CRM business application.
This combining of the two systems not only gives us improved recommendations that leverage an understanding of the broader set of interactions between people within the business, but it also allows us to take into consideration more than is explicit in both systems. For example; if I am looking to realize a successful project outcome I need expertise, however a larger part (in my opinion) is the chemistry that makes up a team. By capturing the outcome of prior opportunities we can use past successful teaming as a way to infer optimal combinations for future projects.
Success is ultimately about people and therefore we need a people-centric analytics infrastructure that takes into account the subtleties of human interaction.