Turning Raw Glucose and Wearable Data into Meaningful Insights: A Step-by-Step Data Analysis Process
Updated: Jul 21
"Raw data doesn't create value on its own. The value comes from cleaning it, transforming it, analyzing it, and presenting it in a way that helps people make better decisions."
When I first started learning data analysis, I thought the most important part was building dashboards. But after working on a healthcare analytics project using Continuous Glucose Monitoring (CGM) and wearable device dataset, I realized that dashboards are only the final step. The real work begins with raw data.
In this project, I worked with raw data containing glucose readings, heart rate, EDA, Temperature wearable device data. I focused on designing the database creation, data Ingestion, preparing and transforming the data using PostgreSQL, performing analysis to uncover patterns, and building interactive Power BI dashboards to communicate meaningful insights.
This blog walks through the data analysis process I followed.
The Data Analysis Process
Every successful project begins with understanding the available data and preparing it for analysis. Since this project started with raw files, my workflow focused on preparing and analyzing the data before creating visualizations.
Workflow
Raw Data → Data Ingestion → Data Cleaning → Data Transformation → Exploratory Data Analysis (EDA) →Data Analysis → Data Visualization → Business Insights
Each stage builds on the previous one, ensuring that the final insights are based on accurate and reliable data.
Step 1: Understanding the Raw Dataset
This project started with raw healthcare datasets containing:
Patient information
Continuous Glucose Monitoring (CGM) readings
Wearable device measurements
Heart rate data
Activity levels
Calories burned
Food consumption data
At the beginning, the data was not ready for analysis. Raw datasets often contain missing values, duplicate records, inconsistent formats, and unstructured information.
Before answering any healthcare questions, I needed to understand the data structure and prepare it for analysis.
Step 2: Data Ingestion and Database Creation
The first step was bringing the raw data into a structured database environment using PostgreSQL.
I created database tables, defined relationships between datasets using primary keys and foreign keys, and designed an Entity Relationship (ER) diagram.
This helped organize different healthcare data sources and created a strong foundation for analysis.
A well-structured database makes it easier to access, manage, and analyze data efficiently.
Step 3: Data Cleaning Using SQL
Raw data often contains missing values, duplicates, inconsistent formats, and other quality issues that can affect analysis.
Using SQL, I cleaned the dataset by:
Removing duplicate records
Handling missing or null values
Standardizing date and time formats
Correcting inconsistent values
Validating data quality
Cleaning the data ensured that later analysis was based on reliable and consistent information.
Step 4: Data Transformation Using SQL
Once the data was cleaned, it needed to be transformed into a format suitable for analysis.
The transformation process included:
Joining multiple tables
Creating calculated columns
Categorizing HbA1c values
Aggregating daily patient metrics
Preparing summary tables for reporting
These transformations made it easier to answer analytical questions and build visualizations efficiently.
Step 5: Exploratory Data Analysis(EDA)
After cleaning and transforming the healthcare data, the next step was to explore the dataset and understand patient health patterns. I performed Exploratory Data Analysis (EDA) using PostgreSQL to identify trends, relationships, and unusual patterns within glucose monitoring, wearable device, and lifestyle data.
I used advanced SQL techniques such as Window Functions, CASE Statements, Stored Procedures, and User-Defined Functions to perform calculations, compare patient metrics, and answer important healthcare-related questions.
The analysis included:
Patient Health Profile Analysis
Analyzed the distribution of patients across different HbA1c categories to understand overall glucose levels.
Compared patient groups based on health indicators to identify variations in metabolic health.
Glucose Pattern Analysis
Examined daily glucose trends to understand how glucose levels changed over time.
Identified patients with the highest glucose readings to detect potential risk patterns.
Calculated average glucose levels by patient to compare individual health profiles.
Analyzed glucose spikes and variability to understand fluctuations in glucose control.
Wearable Device Data Analysis
Analyzed heart rate patterns to understand patient activity and physiological changes.
Evaluated calories burned by patients to study activity levels and lifestyle patterns.
Lifestyle and Nutrition Analysis
Identified frequently consumed food items to understand dietary patterns.
Analyzed relationships between food intake, activity levels, and glucose changes.
Health Score Calculation
Created a daily health score by combining multiple health metrics such as glucose levels, heart rate, activity, temperature, and calories.
Used this score to evaluate overall daily health status.
Through EDA, I transformed raw healthcare measurements into meaningful insights by identifying trends, detecting abnormal patterns, and understanding relationships between different health factors.
Step 6 : Data Analysis and Insights
After exploring the dataset through EDA, the next step was performing deeper analysis to answer specific healthcare questions.
The analysis focused on:
Identifying patients with higher glucose levels
Comparing health metrics across patients
Understanding glucose variability patterns
Evaluating lifestyle factors affecting health indicators
Measuring daily health performance
The goal was not just to calculate numbers, but to understand what those numbers represented and how different health factors were connected.
Step 7: Data Visualization
After completing the analysis, I imported the processed data into Power BI to create an interactive dashboard.
The dashboard included visualizations such as:
KPI Cards
Clustered Column Charts
Line Charts
Pie Charts
Bar Charts
Trend Analysis
Interactive filters and slicers allow users to explore patient health metrics from different perspectives, making the results more accessible and easier to interpret.
Key Insights
The analysis revealed several meaningful healthcare insights, including:
Glucose patterns varied significantly across patients.
HbA1c categories provided a clear overview of patient health status.
Daily glucose trends highlighted periods of higher variability.
Wearable device metrics added valuable context when interpreting glucose data.
Interactive dashboards made it easier to compare multiple health indicators simultaneously.
These insights demonstrate how properly prepared data can support better understanding of patient health.
Technologies Used
SQL – Data cleaning, transformation, and analysis
Power BI – Dashboard development and visualization
What I Learned
Working on this project changed the way I think about data analysis.
Creating dashboards is important, but dashboards are only as valuable as the data behind them.
I learned that successful data analysis requires:
Understanding the data
Cleaning and preparing it carefully
Transforming it into a usable format
Answering meaningful questions
Presenting insights clearly through visualization
Each stage builds on the previous one, and skipping any step can affect the quality of the final insights.
Final Thoughts
Turning raw glucose and wearable device data into meaningful insights is more than just writing SQL queries or creating Power BI dashboards. It is a structured process that starts with raw data and ends with information that people can understand and use.
This project strengthened my SQL skills, improved my data visualization abilities in Power BI, and reinforced the importance of following a structured data analysis process.
Whether you're working with healthcare, finance, retail, or any other industry, the same principle applies:
Clean data leads to reliable analysis, and reliable analysis leads to meaningful insights.
