Over the last two decades, organizations have built enterprise systems and automated business processes that create a significant volume of structured and unstructured data from internal systems, transactional systems and machine generated data (like GPS system output, RFID tags, Web logs, etc.). By itself, this data isn’t very valuable until it can be effectively transformed into information through deep data analysis, modeling, forecasting and simulations to provide actionable insight and support fact-based decisions and agile strategies
Business intelligence enables greater value around data collection, data integration and standard dashboards and reports through analysis of past performance. Recently, organizations have started to go beyond traditional BI and explore ways to manage business outcomes using descriptive, predictive and prescriptive analytics. (See related graphic number one, left.)
Analytics is combination of quantitative analysis with the art of reasoning using multiple techniques such as statistical analysis, data mining, machine learning, pattern recognition and visualization to drive business strategy and performance. The objective is to build confidence and trust in data and institutionalize analytics to transform data-driven organizations.
An analytics culture must be embedded within the enterprise strategy and business operations to ensure that insights are available at all levels and functions in order to support decision-making. Defining and monitoring key performance indicators regularly will help indicate when to take action and/or implement process changes for success.
How to Start
Global competition, stringent regulatory compliance, the lack of clear market differentiation, new delivery channels and compressed time to market are all factors driving the adoption of analytics.
Analytics should be part of the business strategy as a means to improve efficiency and increase profitability through actionable insights. Thus, it is necessary to first identify business objectives, and then align the analytics strategy to effectively achieve the desired improvements. Organizations must approach analytics as a long-term solution that will bring a steady incremental ROI over time as a result of doing things better, smarter and faster.
The high-level business objectives of the organization or the overall goals for the program can be perceived as benefits from multiple business areas as shown in related graphic number two (left).
Establishing the Foundation for Analytics
The process of delivering business results through analytics is one of continuous improvement. The starting point is to address specific organizational needs, define the business case and proceed with data preparation, data mining and evaluation processes in cyclical order. But the model building and fine-tuning the model are very iterative in nature. Any issues (such as non-interpretable results or unexpected findings) may require reassessment of the data preparation, data mining algorithms and model building process or going back to business case definition.
Building an analytical model is the heart of the whole process, and the goal is to identify patterns and relationships in data. Each analytical model is produced by a specific algorithm. Some business problems require more than one algorithm and model to solve. Analytical techniques represent a class of data mining problems that can be solved using statistical and mathematical algorithms.
- Technique: Classification
Applicability: This is the most commonly used technique for predicting specific categorical (discrete, unordered) labels such as response/no-response, high/medium/low-value customer, likely to buy/not buy.
Algorithms: Logistic regression (logistic model or logit model), Naive Bayes, support vector machine and decision tree.
- Technique: Regression
Applicability: This is a technique for predicting a continuous numerical outcome such as customer lifetime value, house value or process yield rates.
Algorithms: Multiple regression, support vector machine.
- Technique: Attribute importance
Applicability: This ranks attributes according to the strength of a relationship with the target attribute. Use cases include finding factors most associated with customers who respond to an offer and factors most associated with healthy patients.
Algorithms: Minimum description length (MDL).
- Technique: Anomaly detection
Applicability: This identifies unusual or suspicious cases based on deviation from the general behavior or model norm. Common examples include health care fraud, expense report fraud and tax compliance.
Algorithms: One-class support vector machine.
- Technique: Clustering
Applicability: Clusteringi is useful for exploring data. Data is clustered or grouped based on the principle of maximizing the intra-class similarity and minimizing the inter-class similarity. Common examples include customer segmentation and drug discovery.
Algorithms: Kohonen networks, enhanced k-means, orthogonal partitioning clustering, etc.
- Technique: Association
Applicability: This is helpful for finding unusual patterns and rules associated with frequent co-occurring items from large data sets. It is commonly used for market basket analysis, cross-sell, root cause analysis, and is ubeneficial for product bundling, in-store placement and defect analysis.
Algorithms: Apriori, generalized rule induction (GRI).
- Technique: Feature extraction
Applicability: This produces new attributes as a linear combination of existing attributes. It is applicable for text analytics, pattern recognition and latent semantic analysis.
Algorithms: Non-negative matrix factorization.
This process can be defined by using the following six basic steps and summary of activities as given in the third related graphic (at left):
Step 1: Define the objective.
- This step includes analyzing business requirements, cases/issues, defining the scope of the problem, determining the metrics by which the model will be evaluated, and setting the final objective for the analytics project.
- At this point it is necessary to develop the framework for the complete analytics process.
Step 2: Analyze the operation.
- Identify the value the organization delivers to its customers, the applications used to deliver the business and core processes (including management systems and metrics, operational and transactional processes and touchpoints with external parties).
Step 3: Prepare the data.
- Data collection: Identify input data sets (identify input data, sample from a larger data set, and partition the data set into training, validation and test data sets). Identify the questions (who, what, where, when, why and how) that will address pain points and create revenue, cost or margin value.
- Conduct data (as well as metadata) quality analysis and explore data sets statistically and graphically (plot the data, obtain descriptive statistics, identify important variables, perform association analysis and find outliers).
- Data consolidation and cleansing.
- Data transformation and input parameter specification: Prepare the data for analysis (create additional variables or transform existing variables for analysis, identify outliers, replace missing values, modify the way in which variables are used for the analysis, perform cluster analysis, analyze data).
Step 4: Develop the analytical model.
- Select the modeling technique based upon the business case and mining algorithm.
- Build the model. Determine the parameter settings and model a target variable by using a regression model, a decision tree, a neural network, or a user-defined model.
- Assess the model (rank the models).
- Output construction in form of visualization and interface.
- Compare competing models (build charts that plot the percentage of respondents, percentage of respondents captured, lift and profit).
Step 5: Deploy the model.
- Deploy the analytics model into the production environment to assess acceptance of the model and its results in the business community.
- Collect data from the business processes after deployment.
- Evaluate effectiveness (monitor, report and recommend).
Step 6: Embed analytics.
- Determine the approach to embed the analytics into operations.
- Define the to-be business process (optimization logic added to rules engine, new work flows and simulations).
- Embed analytics capabilities into the decision-making processes.
The knowledge base acts as a common pool of information where the business case model, data, metadata, data preparation rules, data mining algorithms, results and evaluation information is kept.
It is advisable to start with a business objective that has high impact area and is critical to the business. It is best to adopt a diagnostic approach to determine opportunities and gaps to link analytics to the key business drivers and deliver actionable intelligence quickly.
The nature of a business and high impact areas across divisions determine the degree of adoption of analytics within the organization. For example: banking, insurance and financial services extensively leverage analytics for pricing, risk selection, claims management and fraud detection, in comparison with the manufacturing and logistics industries where the focus is more on supply chain optimization.
Analytics allows organizations to gain deep insights into their business by detecting exceptions and identifying patterns to drive effective decision-making. Master data management, data warehousing/BIPM, business rules management and data integration technologies are all key components of the analytics strategy. To start on the fastest path-to-value, companies should follow a well-defined implementation roadmap toward a fully integrated solution.
Register or login for access to this item and much more
All Information Management content is archived after seven days.
Community members receive:
- All recent and archived articles
- Conference offers and updates
- A full menu of enewsletter options
- Web seminars, white papers, ebooks
Already have an account? Log In
Don't have an account? Register for Free Unlimited Access