Asking Effective Questions:
We all learn the tools like SQL or Python to analyze dataset but one thing that is more important before we start coding is what is it that we want to accomplish with the analysis.
That is where asking effective questions come into play. In real life too, we are used to ask questions for everything- some are good questions that can bring meaningful answers which we can use for decision making and some are just questions without giving us any valuable direction.
For example, if you go to ice-cream shop and ask if they have any tropical flavor ice cream, the answer would be yes or no. That may not necessarily gives the exact ice cream that you want.
Or someone going through a depression asking why are they feeling depressed, may not give them the actual useful answer that they are looking for to come out of depression.
A better question may be, which are the tasks that I do after which I feel depressed or is there a certain time of the day when I feel depressed more than any other time or am I feeling more depressed in some weather condition than others or are there any activities that make me feel happy so I can do them more often?
In data analytics once you understand your dataset, then you will come out with all such questions. A good practice would be to list all questions that come to your mind and then later on decide which ones are useful for your analysis.
In the dataset given below, what potential questions that you can ask that will help you with analysis.
This is a data of online sales orders placed by customers through app for a grocery company.
Column fields:
user_id customer id
order number indicates what is the number of this order that user_id has placed
order_dow is which day of the week order is placed
order_hour_of_day is what hour of the day order is placed
days_since_prior_order is days since last order was placed
Questions:
- Which customer has placed maximum no of orders
- Which is the busiest day of the week when max orders are placed
- Which is the busiest hour of the day when max orders are placed
- List customers who place orders at least once a week
- List customers who did not order for 30 days or more
- List top 10 customers
- How many orders are received every day of the week
These are very basic questions but may help the company with following tasks
Insights:
- Who are our most precious customers so we can run loyalty program for them
- Is there a specific day when we need to keep max staffing?
- Is there a period of the day when we need to keep max staffing
- Can we look at app load during the day so it runs without any failures
- Are there customers who order once and then did not order so we can target marketing campaign for them?
- For warehouse and packaging capacity how much daily volume we need to prepare for
- Who are our most regular/loyal customers so we can run targeted marketing campaign for them
- Are there customers who like to place orders only during certain time of the day?
- Are there customers who order only during weekends?
In case, if you thin more questions can be asked from this dataset, please feel free to list them and I will add them here for the benefit of everyone.
As you can see from this example that from what looks like a mundane data otherwise, now companies can use it to make decisions to improve operational efficiency and customer satisfaction using the historical data
Now if we can ask the company to provide more data regarding what products are bought in each orders at what price and some customer data about location, age, gender, family status etc. we can use these combinations and derive more insights which can help the company with specific targeted list of buying behavior of different groups of customers.
This can be very valuable as instead of generic marketing, now using these insights, companies can get more response from their marketing effort. And that is why data analysts are important resource for companies.
Everyday, tons of data is collected by companies from their daily operations and they can take the help of data analyst to get multiple benefits from it.
What can data analysis do for the company?
- Improve processes
- Identify opportunities and trends
- Launch new products
- Serve customers
- Make thoughtful decisions
Example of Data Analytics process
Most companies use employee satisfaction surveys every year to gauge employee engagement and also to fine tune some policies related to human resource that can increase employee retention.
Our attempt here is to provide an example of such a process to be conducted in a company whose primary issue is that higher employee turn over in the first year after employees join them.
A group of analysts have been given the task to gather data, analyze the results and present it to stakeholders.
They identified the whole project in to different stages
stage-1: Define goal of the program
Team asked some basic questions to stakeholders to decide the goal and accomplishment benchmarks.
- Do you have historical data collected from new employees who joined in the last 5 years?
- Do you have top 3 reasons in mind(based on experience or exit interviews) why new employees leave the company within a year?
- What is your goal of employee retention(in percentage) in the next year?
- What needs to happen to label this exercise as success(any percentage improvement in retention)?
- Do you know any managers within the company whose departments have higher retention and do you know if they are doing anything differently?
- Team defined timelines of the program with a final date of completion
Stage-2: Data collection
- Team prepared a questionnaire for new employees and asked managers of all departments to try and get the maximum participation from employees.
- Team also established rules for who would have access to this data once it is available.
- Team also decided that only few members will have access to raw data collected from employees. Others will have access to a modified data after some sensitive information like salary etc. removed- this is to protect the privacy of the data. If employees were told that survey would be anonymous, then employee name would be removed from the data before sharing it to stakeholders.
No comments:
Post a Comment