Although the term has gained traction as a synonym of business intelligence, we look at Business Analytics as a specific use of statistics, mathematical models, analytical methods, and data mining to deliver data-driven insights and predictions. These insights help our clients make more informed and impactful decisions.
Business analytics can help answer a number of complex questions such as:
- Who are our valuable customers?
- What will next year’s sales be by product?
- How will an X% decrease in production time impact client satisfaction?
- What action will most impact profitability of our worst performing product?
We start by working with our clients to discover the specific opportunities and challenges that can be addressed with more sophisticated use of analytics.
Typical uses of analytic methods include:
- C-level: scenario analysis, simulation, prediction
- Finance: prediction, clustering, risk analysis, optimization
- Marketing: prediction, clustering, association, sequence analysis, queuing, regression
- Operations: prediction, statistical process control, risk analysis, optimization, queuing, scenario analysis, simulation, variance detection,
- Sales: prediction, variance detection, clustering, optimization
DecisionPath helps you leverage business analytics
Once specific analytical goals and methods have been identified, we develop and implement appropriate data structures – taking into account the type of data required as inputs to the various analytical techniques and the format required by the specific tool or tools our clients will be using.
We also help our clients sort out whether their analytics approach should be factored into a wider business intelligence approach based on a single version of the facts of the business. While this is preferable from a longer-term perspective, it also happens that analytics is being approached on an ad-hoc or experimental basis in some companies until the value is proven to business executives. In those cases, more ad-hoc approaches to data acquisition and management may be in play, though the cost of that can be felt by analysts who have to spend a lot of time just getting and cleaning data – rather than exploring and analyzing data.
|Data Exploration and Preparation||Data Exploration||Exploration is where you begin to form your hypothesis about the problem you are trying to solve. In this step, you understand, shape, and select your data in a way that you believe will be pertinent to the problem at hand.|
|Binning||Segment by user defined ranges or values and indicate how many segments (bins) to create.|
|Outlier Removal||Remove the outlier noise that can impact the pattern in the data.|
|Consolidating Values||Group similar values into one to make easier to analyze/mine.|
|Sampling||When the data set is too large to work with, due to limits of hardware and time, sampling is used. The sample data should be big enough to contain the significant information, but small enough to process quickly.|
|Descriptive Analysis||Oversampling||Used when there is not enough data for the analysis. When mining rate events, this helps by amplifying their presence in order to allow the data mining capabilities to perform better. Data mining capabilities perform poorly when there are infrequent events in the data. This will increase the size of the data set.|
|Clustering (Segmentation)||Used to indentify natural groupings of cases based on a set of attributes. Cases within the same group have more or less similar attribute values|
|Association (Market Basket Analysis)||Finding which attributes “go together.” If you buy “X”, what other products do you tend to buy?|
|Inferential Analysis||Sequence Analysis||Finds patterns in a series of events called a sequence. This sequence could occur during a single interaction, or over entire client lifecycle.|
|Classification (Decision Trees)||Classifying something based on other characteristics. For instance, we can use age, gender and location to determine someone’s likely income level.|
|Regression (Estimation)||This is similar to classification, except that instead of looking for patterns that describe a class, the goal is to find patterns to determine a numerical value such as the amount a client will spend at a casino given that client’s income and gender.|
|Prediction (Time Series)||Predicting or forecasting results that exist in the future.|
|Deviation Detection||Detecting deviation from normal behavior.|
(What-If and Goal-Seek)
|What-If: estimates the impact that changes in one attribute has over a different attribute. Specify a new value for an attribute, and see how this hypothetical change affects the outcome attribute.Goal-Seek: is used to determine what change is most likely to generate a desired outcome or goal. Specify the goal, choose the attribute to change and search all the possible values of the attribute to find the one that maximizes the likelihood of reaching the goal.|