Filters & LOD

4. Filters — Tableau's Processing Order

Filters are applied in a specific priority order. This is a frequent interview question.

Context Filter — The Most Asked Interview Question

Problem: You want "Top 10 customers in Delhi." If you apply a normal Top 10 filter alongside a Delhi filter, Tableau finds Top 10 from ALL India first, THEN filters for Delhi — giving wrong results.

Solution: Make the Delhi filter a Context Filter. Tableau will first isolate Delhi data, then find the Top 10 within that subset.

WITHOUT Context Filter:
Step 1: TOP 10 from ALL India → Maybe none from Delhi
Step 2: Filter Delhi → 0 results! ❌

WITH Context Filter on City='Delhi':
Step 1: Filter Delhi first → All Delhi customers
Step 2: TOP 10 from Delhi → Correct 10 ✅

🧠 Interview mein ek line: "Context filter creates a temporary table with the filtered data, and all subsequent filters work within that subset."

---

5. Calculated Fields & Table Calculations

5.1 Calculated Fields

Custom formulas you create in Tableau. They are computed at the data source level.

Common Examples:

// Profit Margin
[Profit] / [Sales] * 100

// Customer Age Group
IF [Age] < 25 THEN 'Young'
ELSEIF [Age] < 40 THEN 'Middle'
ELSE 'Senior'
END

// Days Since Last Order
DATEDIFF('day', [Last_Order_Date], TODAY())

// Year-over-Year Growth
(SUM([Sales]) - LOOKUP(SUM([Sales]), -1)) / LOOKUP(SUM([Sales]), -1) * 100

5.2 Table Calculations

Computed after the main query based on what's visible in the view. They work on the result set, not the raw data.

Calculation	What It Does	Use Case
Running Total	Cumulative sum across rows	"Total revenue accumulated over months"
Percent of Total	Each value as % of the sum	"What % does each region contribute?"
Difference	Change from previous value	"How much did sales change month-to-month?"
Percent Difference	% change from previous value	"Monthly growth rate"
Rank	Rank values	"Rank products by revenue"
Moving Average	Average of N surrounding values	"3-month moving average to smooth trends"

🧠 Interview ke liye important: "Table calculations are computed on the result set, not the underlying data. That's why they depend on the Table (across) or Pane (down) direction — called addressing and partitioning."

5.3 Sets and Groups

Feature	What It Is	Example
Group	Manually combine dimension members	Group Delhi, Mumbai, Bangalore into "Metro"
Set	Dynamic or fixed subset based on condition	Top 20% customers by revenue (updates automatically)

---

6. LOD Expressions — Level of Detail

LOD expressions let you compute aggregations at a different granularity than what's shown in the visualization.

LOD Type	Syntax	Behavior
FIXED	{FIXED [City] : SUM([Sales])}	Always computes at the specified level, ignores the view
INCLUDE	{INCLUDE [Customer] : AVG([Sales])}	Adds the specified dimension to the view's granularity
EXCLUDE	{EXCLUDE [Month] : SUM([Sales])}	Removes the specified dimension from the view's granularity

Example — FIXED LOD:

Scenario: Your view shows product-level sales. You want each product row to also display its CATEGORY total for comparison.

Calculated Field: Category Total
{FIXED [Category] : SUM([Sales])}

Result:
Product: Laptop  | Sales: 45000 | Category Total: 185000
Product: Monitor | Sales: 20000 | Category Total: 185000
Product: Mouse   | Sales: 500   | Category Total: 185000
                                  ↑ Same for all Tech products

Example — INCLUDE LOD:

Scenario: View shows City-level data. You want the average order value per customer per city (not the city-level average).

Calculated Field: Avg Order per Customer
{INCLUDE [Customer_ID] : AVG([Order_Amount])}

This first calculates AVG order for EACH customer, 
then aggregates those customer-level averages up to the City view.
Without LOD, you'd just get the city's overall average — which
treats a customer with 1 order and 100 orders equally.

🧠 Yaad kaise rakho: FIXED = "is level pe fix karo, view ko ignore karo." INCLUDE = "ek aur level add karo." EXCLUDE = "ek level hatao."

---

7. Dashboard Design Best Practices

Rule	Why	Example
KPIs at the top	First glance should show key numbers	Total Revenue, Growth %, Customer Count as big scorecards
Follow Z-pattern	Users read left→right, top→bottom	Important charts top-left, supporting charts bottom-right
Maximum 3-4 charts	Information overload confuses stakeholders	Revenue trend + Category split + Top customers + KPI cards
Consistent colors	Same color = same meaning throughout	Green = good, Red = bad, Blue = neutral
Interactive filters	Let users explore on their own	Region, Date Range, Product Category dropdowns
Descriptive titles	"Revenue" ❌, "Monthly Revenue Trend (FY 2023-24)" ✅	Include metric name + time period
Mobile-friendly	Executives check dashboards on phones	Create device-specific layouts

Dashboard Actions

Action Type	What It Does	Example
Filter	Clicking a mark filters other sheets	Click a bar in City chart → filters Product chart
Highlight	Clicking a mark highlights related marks	Click "Technology" → highlights in all charts
URL	Opens a webpage on click	C