Blog > Data Science
Data Roadmap for Managers (Baseline Version)
Defining a road map may not be an easy task for every type of data-oriented project. As KalybeAI we define key standards for different phases (use case selection, project management, coding standards, etc.) of our works. In this blog, we share the basic parts of our standard document to be able to use as the roadmap for the managers in the field. The document has 8 questions in the basic version. The number of questions can be extended according to the different requirements, and it is crucial to answer all questions in detail.
1. What is your goal?
• Some of the high-level goals may be as follows.
o Reducing costs
o Increasing sales
o Boosting revenue
• Some of the sub-goals may be as follows.
o Generating more leads
o Preventing customer churns
o Optimizing the spare parts (stocks)
o Forecasting customer demands
o Price prediction and optimization
2. What is the expected output?
• You may want to see just reports (Customers who are likely to churn, Customers who are likely to purchase, etc)
• System generates a number (the result of a prediction), and you may want to use it to take some actions
3. Do you have data for your goal (s)? (Yes/No/Need more)
• I have data as the result of different processes (in databases).
• We are collecting data in an automatic way (automation systems/machines).
• My data only consists of excel files.
4. What are the data entities in your data?
• Master Data (Customers, Products, Employees, etc.)
• Transaction Data (Events, Actions, Orders, Sales, Purchases, etc.)
5. What are the features, types, and characteristics of your entities?
• You can prepare a Data Definition Table (DDT) or data dictionary.
6. What is the size of your data (in MBs, GBs, TBs, PBs)?
• What are the size of files that contain your data (DB, excel, etc)?
• What are the counts of entities? (You can also group your entities according to years, months, some other info and re-calculate using these groups. So, you might want to concentrate on entities with the most data)
7. Can you measure the high-level quality of your data?
• The quality metrics of the data should be defined in a formal way including some stats like a number of inconsistencies, the ratio of data in an appropriate format, duplicate rows, etc.
8. Which type of analysis do you need?