Typical DMSA Analytics Projects
INTRODUCTION: DMSA High Level Approach to Data Analytics
- Initially, the DMSA team of Statisticians and Business Strategists meet with your organisation to understand your organisation and then customise our analysis to meet your strategic objectives
- We integrate your data into a single comprehensive view using ‘clean’ and ‘sanitised’ data
- Profile existing client base
- We apply granular customer segmentation and predictive modelling to predict trends and customer behaviours.
- Use understanding of the existing client base to target new clients / areas
- Implement models on client’s system.
Typical DMSA Analytics Projects
1. Data Cleaning of Company Database
DMSA’s Analysts are Statisticians that also have strong programming skills for the management of large data sets
Their programming skills involve:
- Extracting data from any data sources like SCADA, Oracle, SAP, SAS, Microsoft suite, or raw sources with flat files, databases and the like.
- Transforming and restructuring data
- Much effort is devoted to data assessment and data integrity testing. Once these initial data management phases are completed systematic errors are identified and presented to management for the next phase – rectification.
- Data cleaning on out of range, erroneous and missing data is often required. This phase is performed in discussion with the client as certain data errors requires a client’s insight into business processes.
- Missing data often requires applying the relevant imputation techniques or weighting of the data
- Insights gained from the data cleaning exercise then informs the type of analysis that can be applied in different situations.
2. Analysis of Marketing Campaign and Sales Data
- Evaluate impact of sales promotions
- Evaluate and improve existing Sales Forecasting systems
- Focus on and identify parts of the marketing budget and marketing activities that yield optimal returns.
- DMSA Sales Forecasting uses sophisticated time series analysis to extrapolate past behaviour into the future.
- Basket Analysis to predict buying behaviour
- Leads modelling for campaigns: we apply predictive models to determine where clients are most likely to interact and buy the product.
- To improve insight and performance, we link Customer transactional data to Survey data from Customer and Employees
3. Collections Optimisation
- Link dwellings / firms to Geocoded Data
- Do impact tests on past collection interventions
- Use segmentation to determine uptake per suburb for example: by payment options
- Predict default individuals, Suburbs, Segments of interest.
4. Risk Modelling
- Use Predictive Modelling to develop lapse and churn models that can identify high and low risk customer groups
- Segment customer data according to risk profiles
5. Forensic Investigations
- Identify suspicious activities, e.g. suspicious insurance claims
- Identify “ghost” employees
- Review procurement on supplier databases to determine payroll and supply chain fraud
- Identify collusion of staff and customers to facilitate fraudulent activity
6. Credit Analytics
- Apply predictive models to determine credit-worthiness of potential clients
- Determine the probability for lapsing
- Produce ratio and trend analysis and create projections