The science of strategic decision making based on recommended actions derived from past performance, current conditions, and desired outcomes
If you are the "lucky one" and all your data is clean and available in a "datamart" or "data warehouse" then it a good time to start generating value from it. However, if your data strategy is not yet defined but all your data is being collected in your information systems or elsewhere there are still opportunities to answer some of your burning questions by cleaning and integrating your data and discovering trends and patterns from it. Unfortunately analytics can only be performed on data that has been collected and sometimes findings may point to a need for business process setup or re-engineering before a question can be answered. Depending on your needs, analytics tries to answer following questions:
- What has been happening in my business? This Descriptive Analytics usually takes form of dashboards that track metrics that capture the pulse of your business. It does not inform you on why something has happened/been happening.
- Why is it happening? Want to understand why an important metric has been dropping for some time. This Diagnostic Analytics takes form of a one-off analysis to understand the drivers of a specific problem like "why is sales in a region dropping?". It usually follows a successful delivery of Descriptive Analytics.
- Can it be predicted? Want to understand if past data can be used to predict future events like probability of your customer staying or leaving, estimated loss due to flood etc. This Predictive Analytics usually follows successful implementation of Descriptive and Diagnostic Analytics.
- Can the predictions guide my decision-making? Want to use a successful predictive model to guide/automate decision making such as allocating optimal media investment based on predicted media performance, to achieve desired sales uplift. Prescriptive Analytics usually follows successful implementation of a predictive model.
A sample deliverable for a predictive analytics project:
Depending on the maturity of processes, a predictive model can be delivered within 3-6 months with maintenance cycles every 6-12 months depending on the type of business. Similar deliverables with varying timelines can be provided based on type of Analytics project.
- Project Management
- Understanding and agreement of problem
- Data Audit
- Project proposal, and agreement of scope and costing
- Data collection and Preparation
- Exploratory Data Analysis
- Model Building and Testing
- Model Deployment
- Model Performance Reporting and Maintenance