Reference no: EM133967925

Assignment: Business Intelligence Dashboard Development

Objective
The objective of this assignment is to design and implement a Business Intelligence (BI) dashboard that provides actionable insights based on real-world data. You will apply key BI concepts including data collection, cleansing, analysis, and visualization, using tools such as Python, R, Tableau, Power BI, or R Shiny.

Assignment Steps & Guidance

Step 1: Define Your Theme and Business/Research Question
Select a theme or business/research question that your dashboard will address.
Ensure that the theme aligns with your interests and provides meaningful insights.
Examples:
Sales performance analysis for an e-commerce business.
Sentiment analysis of customer reviews.
Financial risk assessment for a stock portfolio.
Predicting housing prices based on market trends.
Deliverable: Submit a short plan (max. 2 pages) detailing - submission deadline week 5 of the semester 30. October 2025:
Chosen theme and relevance

Key questions to be answered by the dashboard
Potential data sources
Overview of intended visualizations

Step 2: Find Suitable Data

Gather at least three different data sources from credible platforms such as:
Kaggle
Tableau Public datasets
GitHub repositories
in this repository you can find a long list of public APIs:
Government open data portals (e.g., data.gov, European Data Portal)
APIs for real-time streaming data (e.g., Twitter API, financial market APIs)
Ensure the data is relevant, up-to-date, and has enough features for meaningful analysis.

Step 3: Cleanse and Prepare the Data
Data Cleaning Guidelines:Handle missing values (impute or remove as necessary).
Convert data types where applicable (e.g., dates, numerical values).
Normalize and standardize numerical data if required.
Remove duplicate or irrelevant entries.
Merge datasets correctly if combining multiple sources.

Step 4: Develop a User Interface Logic for Your Dashboard
Think about usability and user experience (UX):Who are the end users?
What key insights should be immediately visible? Get dependable, budget-friendly assignment help-starting today!
Should filters, interactive elements, or drill-down capabilities be included?
Consider layout and information hierarchy for easy navigation.

Step 5: Develop Data Analysis and Visual Outputs
Use Python (e.g., Pandas, Matplotlib, Seaborn, Plotly, Dash), R (ggplot2, Shiny), or Tableau/Power BI.
Ensure the following requirements are met:At least 5 visual representations (e.g., bar charts, line graphs, heatmaps, scatter plots, geographic maps).
One correlation analysis (e.g., correlation matrix, scatter plots with trend lines).
Integration of at least one real-time or streaming data source (optional but bonus marks awarded).

Step 6: Submit Your Dashboard
Host your dashboard on a local or web platform:
Python Dash / Flask / Streamlit (deployed via Heroku, Render, or local server)
R Shiny app (deployed via Shinyapps.io)
Tableau Public / Power BI (shared via links or files)

Business Intelligence & Analytics - Dashboard

Primary user: Who's your decision-maker? What do they need at a glance?
Success metric: How will you know your dashboard "works" (e.g., fewer stockouts, higher CTR, shorter cycle time)?

1) Define your theme & key questions (aligns with Step 1)
Pick a theme you actually care about (motivation helps when data gets messy).
Draft 3-5 sharp questions. Examples:
"Which customer segments drive margin growth?"
"Which features best predict MPG?"
"How do petal measurements differ by species?"
Sketch your initial layout (paper is fine): KPIs up top, trends/diagnostics in the middle, drill-downs at the bottom.
Deliverable tip for the 2-page plan (due 30 Oct 2025):
Include: theme + why it matters, 3-5 questions, potential data sources, first-pass wireframe, and candidate charts.

2) Find data (aligns with Step 2)
Aim for 3 sources (internal + external is ideal). If you start with built-in data for prototyping (like iris), add external sets later.
Check: freshness, license, coverage, and granularity. Note fields that join the datasets (IDs, dates, geo, categories).

3) Clean & prepare (aligns with Step 3)

Sanity checks: duplicates, missingness map, type mismatches (dates as text!), outliers.
Tidy rules: one variable per column, one observation per row, one table per dataset.
Document your fixes: a short "data journal" with bullet points (what you did + why).
Name things well: snake_case, no cryptic codes; keep a tiny data dictionary.

4) Design the UX (aligns with Step 4)
Above the fold: 2-4 KPIs that answer your main question fast.
Left = filters; right = details: consistent placement speeds up use.
Interaction: don't overdo it-each control should change a specific insight.
Accessibility: sufficient color contrast; don't rely on color alone.
Latency budget: <200-500ms for filter updates feels "instant."

5) Build visuals & analysis (aligns with Step 5)
Start simple, iterate: bar/line/scatter, then add depth (facets, tooltips, trend lines).
Correlation view: heatmap or scatter with smooth line; discuss any surprising relationships.
At least 5 visuals: Mix overview (KPIs) + trend + distribution + relationship + breakdown.
Optional real-time: one tile fed by an API or a simulated stream to prove the interaction pattern.

6) Tell the story (aligns with Step 6 report)
Intro: problem, users, KPIs.
Data: sources, cleaning, limits.
Methods: how you analyzed (just enough detail).
Findings: 3-5 insights with evidence (chart callouts).
Action: what should the user do next?

7) Self-check before submission (mini-rubric)
Does the landing view answer the main question in 5 seconds?
Are filters clearly labeled and actually helpful?
Do the units/scales match (no % mixed with absolute values in one axis)?
Is there a correlation view and at least 5 meaningful visuals?
Can someone else reproduce your pipeline from your code + readme?

Color & chart quick-guide (from Unit 07)
Categorical: distinct hues (don't exceed -8-12 without grouping).
Sequential: single-hue light-> dark gradient (for magnitude).
Diverging: two-hue gradient with a neutral midpoint (above/below target).
Bars for categories, lines for ordered/time. Avoid implying trends with a line on nominal data.

R Shiny Starter App (uses iris and mtcars)
This app gives you:
5+ visuals (KPls, histogram/density, boxplot, scatter + trend, correlation heatmap).
Filters, dataset switcher, and a simulated streaming tile (for the "real-time" bonus pattern).
Clean, commented code you can extend with your own data.
How to run: paste into an app.R, then run shiny::runApp().