This story is all my thoughts that was well-structured and able to share with those of you — who are looking for guidance in Business Intelligence career path. It’s more about my personal experience than typical articles that you can search on the Internet.
Just take a look and comment your thoughts :)
1. Five Pillars Of Business Intelligence
- Business Acumen: a combination of knowledge and skill informed by experience: knowledge about domains, the skill to apply that knowledge. The ability to take a ‘big picture’ view of a situation, to weigh it up quickly, make a logical, sound decision confidently, and influence others to agree with you in order to have a positive impact towards achieving the objectives of the organisation.
- Data: be aware of data sources and how data is captured from each one in order to evaluate the data quality level and ensure the proper transformation and business rules are assigned before it can be stored and used in analysis
- Coding Languages: Business Intelligence requires coding for processing data to produce useful insights. BI analysts typically spend a larger amount of time coding in SQL and Python or R. SQL or Structured Query Language is a common querying language used to speak to databases to retrieve the required information. Python and R are programming languages used for data transformation and processing.
- Visualization: The value of data is truly realized when it can be presented to business users in a meaningful way. Data’ users need the ability to see high-level information on dashboards, and then be able to drill down or drill out from summaries to details. Information should also be presented efficiently via effective graphical displays.
- Communication: a critical soft skill to be able to describe the data, explain analysis of that data, and then offer actionable insghits. This involves describing complex technical information to non-BI professionals. Therefore, people in business intelligence need to be able to communicate clearly and effectively.
An example of BI process
2. Required skills and knowledge
- Business Skills: Skills in writing and verbal communication, critical thinking, problem-solving, teamwork
- Data Analysis Skills: SQL, Excel, data analysis skills (quantitative analysis, benchmarks, correlation measurement…) and visualization
- Programming Skills: Good programming skills in languages like Python, R
3. How to approach problems with data analytics
When receiving questions from Stakeholders, BI analysts need to answer by data and here is the way to do it
3.1 Clarify the question you need to answer, try to find the root causes
3.2 Identify metrics necessary to answer the question
3.3 Determine what data is available and what is not available
3.4 Acquire the data that is not available
3.5 Solve the problem by data
3.1. Clarify question by MECE framework
MECE stands for “Mutually Exclusive — Collectively Exhaustive”, is a popular mantra at McKinsey.
Read more here!
MECE is a structured problem-solving approach that forcing to list down all possible options without double counting.
There are 3 popular MECE frameworks: Issue Tree, Decision Tree, Hypothesis Tree
- Issue Tree is helpful for solving a complex problem as it facilitates splitting them up into smaller. A common type of case in which a MECE issue tree is used to analyse user behaviors by multiple segments (ages, gender, occupation…)
- Decision Tree is a tree-shaped graphical representation of decisions and potential outcomes of those decisions and is used to determine the comparative advantages and disadvantages of each decision. A common type of case in which a MECE decision tree is used to estimate Business Value of New Initiatives and then rank the priority.
- Hypothesis Tree is quite similar to an issue tree. However, a hypothesis tree organizes a problem around hypotheses, and often offers a more direct approach than an issue tree.
Let’s talk more about Hypothesis Tree.
When faced with complex problems and you are uncertain about the causes and solutions, or have no a particular direction of analysis, it is essential to develop a hypothesis tree and deep dive into them. Following 3 steps below:
- Understand the problem thoroughly by constantly asking WHY and HOW questions
- Break down problems and make assumptions piece by piece. Write down all your assumptions and MECE
- Decide what’s important to validate, then collect data and assess data correlation, consistency, and accuracy.
And…
The more curious you are about your assumed problem, the further you will be able to take your business with the lessons you have learned by exploring it
Another technique that can be used to elucidate business questions is Drill Down Technique.
Drill down is a simple technique for breaking complex problems down into progressively smaller parts. Drill Down technique is usually combined with other methods like MECE or 5-Why Analysis
3.2 Identify metrics to answer questions
In this step, please make sure you understand the problem to define a clear set of goals: What do you want to obtain from data? Just a performance update? Insights for next steps? Lessons learned?
Additionally, based on your domain knowledge, you decide which metrics to measure and how to calculate.
For example, you have to measure the performance of a feature for improvements in next stages. That is enough clear to go further: you need to collect metrics from various perspectives such as the number of users, user journey, user conversion rates and make recommendations to improve its performance. It is important to keep business objectives in mind when you think about the goal of the analysis work.
3.3 Determine the necessary data
The root of this step is to understand the data of your business and have well-defined goals and a clear sense of purpose (step 3.2).
If your necessary data is available, you’re in luck. Before use, please take the steps below to make sure you fully understand your data and are ready for analysis.
3.3.1 Data Exploration Process
Step # 01: Understand the context in which the data is produced
Step # 02: Understand the raw data you’ll be using
Understanding data is essential. It is important to work together with Data Engineers team so that you understand where this data comes from, if there is any error in the data collection, what transformations it has undergone and what columns mean to be. This avoids confusion when performing calculations and counting.
Step # 03: If using calculated or ETL data, you need to have a good understanding of the calculation logic in order to properly understand the metrics
Step # 04: Start building simple and reliable your data
It is very important that everyone must trust the results presented, so do an exploratory analysis first, create some charts, and check if the data is correct. It is often necessary to take small steps to answer a complex question, then gradually increase the depth of the analysis.
Step # 05: Validate your results with logic check or cross check with other data sources
Step # 06: Answer hypothesis with the data you have collected
3.3.2 Pitfalls of bad data quality
If you search the term “pitfalls of bad data quality” in Google, you can find 10.9 million results mentioning the cost of bad data. Obviously, data quality is far more important than data quantity.
Just a copied of an random article on the Internet because I have no more ideas to share :)
The consequences of poor data quality
Loss in Revenue
Lost each year due to lost productivity from poor quality data.
Inaccurate Analysis
If you’re conducting data analysis or predictive analytics with incomplete and incorrect data, you run the very real risk of being led down the wrong path. With duplications, missing fields and other anomalies in your data you’ll be wasting resources such as, for example, implementing a sales campaign based on poor data analysis.
Damaged Reputation and Fines
If you contact the same person or business multiple times unnecessarily, or are sending emails to a large number of dead addresses, you will likely cultivate a poor reputation both within the physical and digital world and appear inefficient to your actual and potential customers.
3.3.3 Basic data quality check
Some tips to quality check your data
- Validate data sources by checking record counts, data types, and the latest load date
- Check Data Summary Statistics & Distributions to find outliers, and make sure all values are the right data type
- Apply Critical Thinking to really understand your data
- Work with Subject Matter Experts to ensure your analysis is correct
- Compare with Similar Analysis to check on final results, especially if you have never worked with this data before.
3.4 Acquire the data that is not available
Finally, when you can’t get necessary data because it doesn’t exist, consider going out and collecting it. Depending on what data you need, you might conduct a survey, request new data tracking, buy it from 3rd parties. This is usually a last solution because it can be time-consuming and expensive, but it gives you the chance to get data that is tailored to your own needs, and keep it to yourself.
4. Critical thinking beyond analytic skills
In the real business world, the critical thinking behind the analysis is more urgent and relatively hard to train compared to hard skills.
Description, assumptions and hypothesis are essential to BI to understand why we need to do the analysis and what we should provide to stakeholders. We should step out more to dig the reason that leads to the results and provide our suggestions based on the data we collect and analyze from a diversified perspective.
BI should understand more about business, especially the markets, products, and financial performance, try to think from a business perspective and put yourself in customer shoes to create values more than skills for the business and world.
