14. Visuals and Their Formatting
14.19 AI Visuals: Decomposition Tree, Key Influencers, Q&A, Narrative
Start with a problem
Rani asks why is Pune at the top? and wants to click down from the total to the reason without building many charts.
Explain simply
AI visuals help explore: the Decomposition tree splits a measure by the dimensions you choose, level by level; Key influencers looks for factors linked to an outcome; Q&A answers typed questions; Smart narrative writes a text summary.
Show the data
| Store | City | Total Sales (₹) |
|---|---|---|
| Kothrud | Pune | 650 |
| College Road | Nashik | 500 |
| Rajarampuri | Kolhapur | 450 |
| Gangapur | Nashik | 300 |
| Baner | Pune | 200 |
About this data: This table uses columns: Store, City, Total Sales (₹). This table uses columns: Level, Selected, Value (₹). This table uses columns: Wrong result, Reason, Correction. Values shown are the lesson examples already on this page (not a new dataset).
Fictional sample data. Practice file: visuals_orders_apr_may_2026.csv (source: src/practice/).
Demonstrate
- Insert Decomposition tree.
- Analyze:
Total Sales. Explain by:City,Store,Category. - Click + next to the total › choose City.
- Click + on Pune › Store, then + on Kothrud › Category.
Show the result
Your result should look like this (fictional sample):
| Level | Selected | Value (₹) |
|---|---|---|
| Total | — | 2,100 |
| City | Pune | 850 |
| Store | Kothrud | 650 |
| Category | Dairy | 400 |
Explain what happened
Each split re-evaluates Total Sales with the filters of the path so far (Pune, then Pune + Kothrud).
Explore further
- High value split lets Power BI pick the dimension with the largest value.
- Key influencers needs many rows – eight orders are far too few for meaningful results.
Detailed reference for this lesson:
Decomposition tree. Breaks a measure down across dimensions in any order the user chooses. Field wells: Analyze (a measure), Explain by (dimensions). Example: Analyze = Total Orders; Explain by = City, Area, Category, Payment Mode, Platform. Users click + to split, or choose High value / Low value (AI splits) to find the biggest contributor. Formatting: Tree (bar colours, density), Data bars, Analysis (enable AI splits).
Key influencers. Finds factors that drive an outcome. Field wells: Analyze (a column or measure), Explain by, Expand by. Example: Analyze = Is Late (Late / On time); Explain by = Order Hour, City, Category, Payment Mode, Basket Size. It tells you which factors are most associated with late deliveries. It shows associations, not causes, and it needs enough rows to be meaningful.
Q&A visual. Users type a question in natural language, e.g. "total orders by city for Blinkit", and get a visual. Field wells: Q&A (the question box). Improve results with synonyms (Modeling › Q&A setup › Synonyms: "shop" = Dark Store). Supported language is mainly English.
Smart narrative / Narrative. Generates text that summarises the page or a visual, with dynamic values. Newer versions offer a Narrative visual with Copilot where it is available and enabled. Module 20.4 shows how to write your own dynamic text.
Ravindra Bagale's Tip
Friends, many students trust AI visuals like Key Influencers as proof of cause and effect. They show associations in your data, not causes. Use them to find leads, then check the numbers yourself before presenting a conclusion. Keep this in mind!
Ravindra Bagale's Tip – मराठी
मित्रांनो, बरेच students Key Influencers सारख्या AI visuals ना कारण आणि परिणामाचा पुरावा मानतात. ते तुमच्या data मधले संबंध (associations) दाखवतात, कारणं नाही. धागेदोरे शोधण्यासाठी त्यांचा वापर करा, मग निष्कर्ष मांडण्याआधी numbers स्वतः चेक करा. लक्षात ठेवा!
Ravindra Bagale's Tip – हिंदी
दोस्तों, बहुत से students Key Influencers जैसे AI visuals को कारण और असर का सबूत मान लेते हैं. वे तुम्हारे data में संबंध (associations) दिखाते हैं, कारण नहीं. सुराग ढूँढने के लिए उनका इस्तेमाल करो, फिर निष्कर्ष बताने से पहले numbers ख़ुद check करो. ध्यान रखना!
Fix common mistakes
| Wrong result | Reason | Correction |
|---|---|---|
| Key influencers says not enough data | Small sample | Expected; use a larger dataset |
| Users trust the AI split as a cause | Decomposition shows where, not why | Confirm with business context |
| Q&A misreads field names | Technical column names | Rename fields and add synonyms |
Practise independently
Practice task
In the decomposition tree, follow Nashik › Store. Which store is larger?
Hints
- College Road: 304 + 307.
Solution
College Road ₹500, Gangapur ₹300.
Think like a professional
AI visuals speed up exploration; final pages still need visuals you can explain line by line. Mention this balance in interviews.
Revision card