It's a simple idea that makes sense to anyone who has ever looked to their sales activity to try to produce their revenue forecast, or looked at weather data to project flight delays, but one of the biggest challenges in conducting analysis in enterprises today arises from the gap between structured and unstructured information silos.

The structured data technology stack has been optimized for years to enable discovery of the "Why Axis" within the various structured data systems in an enterprise (ERP, CRM, etc.) through data warehouses, data marts and OLAP cubes; however, what happens if the "why" isn't in your structured data?

If you look at your sales activity volume versus your sales outcomes, you might seem some correlations, but there might still be unexplained anomalies, or you may think the correlation is not good enough for you to make critical decisions based on your analysis.  Where do you turn for more information?  The answer is in the other 80% of the content in your organization: your unstructured information.

Let's say you had complete access to your unstructured and structured information so that you could compare data and metadata from each.  In the case of sales forecasts, your unstructured content might tell you things like:

• The number of successful and unsuccessful sales calls per sales rep based on the sentiment contained within the call summary

• The most common key words or phrases used during successful sales email exchanges versus unsuccessful exchanges.

Another example, featured in an Attivio demo, involves understanding baseball player performance by correlating public image over time (positive/negative news articles or articles about "scandals") against player statistical performance.

The key technologies that enable these types of correlations between structured and unstructured content in the Attivio Active Intelligence Engine (AIE) are:

• Unstructured content metadata extraction and aggregation: Create structure around unstructured content

• Entity extraction: Extract people, places, locations, concepts to understand relationships and correlate data with content

• Sentiment analysis: Determine positive/negative sentiment directed at a person, place, location or other entity

• Relevancy and fuzzy matching - Provide "fuzzy" information correlations using linguistics techniques such as synonym expansion, edit distance and acronym detection and expansion

With the addition of unstructured content to your analysis, the often elusive "Why Axis" within unstructured information can be unlocked to provide a complete view of trends and drivers for better and more precise decision making.

Author Bio

Rik Tamm-Daniels is the Vice President of Engineering at Attivio overseeing Attivio's product development and Agile development process.  He has a BS and MS of Computer Systems Engineering from Boston University and has worked at numerous technology start-ups in Boston and Chicago over the past 10 years in hands-on development, architecture design and senior management roles.

TrackBack URI for this entry

Comments (0)