Tableau ও Power BI পরিচিতি

এটি real procurement decision — million dollar implication।

Cost analysis:

• Tableau: ৫০ creator × $৭০/mo + ৫,০০০ viewer × $১৫/mo = ~$৭৮,৫০০/mo = $৯.৪ লাখ/বছর।
• Power BI Pro: ৫,০০০ × $১০/mo = ~$৬ লাখ/বছর।
• Power BI Premium Capacity: ৪-৮ × $৫,০০০/mo = $২.৪-৪.৮ লাখ/বছর + free viewer।
• Tableau ~$৩-৫ লাখ বেশি বছরে।

Existing infrastructure:

• Bank সাধারণত Microsoft shop — Office365, Active Directory, Azure।
• Power BI integration — single sign-on, Teams embed।
• Tableau আলাদা environment — IT additional support।

Talent availability:

• Bangladesh-এ Power BI talent ৩x বেশি।
• Hiring & retention সহজ।
• Tableau senior — premium pay।

Regulatory & data residency:

• Bangladesh Bank guidelines — sensitive data on-premise।
• Tableau Server: Linux/Windows on-premise — supported।
• Power BI Premium-এ data on-premise option (Premium Capacity)।
• দু'টোতেই possible — কিন্তু Power BI Premium Capacity-র setup easier।

Use cases:

• Branch performance dashboard — both tools fine।
• Customer 360 — DAX powerful, Tableau-র Story বেশি।
• Regulatory reporting — DAX-এর precision better।
• Marketing analytics — Tableau-র aesthetics better।

Recommendation: Power BI primary, Tableau supplementary:

1. Power BI primary deployment — সব business user।
2. Tableau ২-৩ Desktop license — analytics team-এর creative needs।
3. Migration plan — যদি Tableau ইতিমধ্যে থাকে, ১৮-২৪ months phased move।
4. Hybrid period: side-by-side, gradually retire Tableau।

Counter-argument — কখন Tableau primary:

• International HQ-এর mandate (e.g., Standard Chartered global Tableau)।
• Existing $20M Tableau investment — sunk cost।
• Brand/marketing-heavy use case dominant।
• Mac-only data team।

Implementation roadmap:

• Month 1-2: Power BI Premium Capacity setup, IT integration।
• Month 2-4: ১০ pilot dashboard build, top business users train।
• Month 4-6: Roll out 100 dashboard, training program।
• Month 6-12: Self-service enablement, governance framework।
• Year 2: Tableau-only use case identify ও migrate।

Risk mitigation:

• Pilot data marts — production-এ যাওয়ার আগে।
• Dashboard certification process — quality control।
• Audit log monitor — sensitive data access।
• DR plan — Power BI service downtime।

মূল উপলব্ধি: ROI calculation cost-only না — talent, ecosystem, future-proofing সব মিলে। Bangladesh banking-এ Power BI default rational choice; exception need explicit justification।

Career path-এর critical question — myth ও reality আলাদা।

BI tool যেখানে যথেষ্ট:

• Standard reporting: sales, finance, operations dashboard।
• Slice-and-dice analysis: filter, drill-down, pivot।
• Calculated metrics: KPI, YoY growth, ratio।
• Scheduled refresh: nightly data update।
• Geographic visualization: map dashboard।

BI tool-এর সীমাবদ্ধতা:

(১) Custom data transformation:

• Power Query/Tableau Prep — limited reach।
• Complex JSON parsing, regex extraction — Python pandas easier।
• Multi-table fuzzy join — BI tool struggle।

(২) Statistical modeling:

• Linear regression — built-in BI।
• Random Forest, XGBoost, neural net — Python অপরিহার্য।
• Time-series forecasting beyond Prophet — Python।

(৩) NLP & unstructured:

• Customer review sentiment — Python (transformers)।
• Text classification — pandas + scikit-learn।
• BI tool কেবল aggregated text count।

(৪) Custom visualization:

• Sankey diagram — Tableau-এ workaround।
• Network graph — Python-এ networkx।
• Animation timeline — D3.js বা Plotly।

(৫) Automation & scripting:

• Email alerts on threshold — Python script + cron।
• Multi-source data combine — pandas।
• Web scraping — BeautifulSoup, Selenium।

(৬) Integration with ML pipelines:

• Model training — Python sklearn, TensorFlow।
• Real-time inference API — FastAPI।
• BI tool dashboard শুধু output show করে।

Real-world team structure:

• Data Analyst: ৭০% BI tool, ৩০% Python। Insights ও dashboard।
• Data Scientist: ১০% BI, ৯০% Python/SQL। Modeling, A/B test।
• Data Engineer: ০% BI, ১০০% Python/SQL/Spark। Pipeline।
• BI Developer: ৯০% BI tool, ১০% SQL। Specialized।

Career trajectory:

• BI-only path: BI Developer → Senior → BI Manager। Salary cap কিছুটা।
• Python+BI path: Data Analyst → Data Scientist → Lead। উপরের ceiling।
• Hybrid most valuable — communication + modeling।

Bangladesh-এ market reality:

• Junior BI role — TK 30-50K (Power BI / Tableau)।
• Senior BI — TK 80K-1.2L।
• Senior Data Scientist (Python+ML) — TK 1.5-3L।
• Both skills — premium remote international rates ($30-100/hour)।

Recommendation:

• Year 1: SQL master, Power BI/Tableau professional।
• Year 2: Python + pandas + scikit-learn add।
• Year 3: Cloud (Azure/AWS) + ML add।
• "BI tool only" — short-term comfort, long-term ceiling।

মূল উপলব্ধি: BI tool ≠ Python replacement, complementary skill। Future-proof career-এর জন্য দু'টোই অপরিহার্য।

DAX-এর filter context — সবচেয়ে কঠিন এবং powerful concept। এটাই DAX mastery-র threshold।

Filter context কী:

• প্রতিটি cell-এ একটি filter context থাকে — visual-এর axis, slicer, page filter সব মিলে।
• Measure সেই context-এ evaluated।
• "Total Sales" — পুরো table-এর sum না, current cell-এর filter-এ sum।

Common pitfalls:

(১) SUM vs SUMX confusion:

// ভুল — পুরো sales-এর sum, line-by-line না
Bad Total = SUM(Sales[Quantity]) * SUM(Sales[Price])

// সঠিক — row-by-row evaluate
Good Total = SUMX(Sales, Sales[Quantity] * Sales[Price])

(২) ALL() ভুল ব্যবহার:

// "% of total" calculation
% Share = DIVIDE(
    [Total Sales],
    CALCULATE([Total Sales], ALL(Product))
)
// ALL(Product) — Product filter বাদ; অন্য filter (date) থাকে।
// ALL(Sales) লিখলে — সব filter বাদ, ভুল ফলাফল।

(৩) RELATED vs RELATEDTABLE:

• RELATED — many-to-one relationship-এ single value।
• RELATEDTABLE — one-to-many side-এ table return।
• উল্টো ব্যবহার = error।

(৪) Time intelligence ছাড়া Date table:

• SAMEPERIODLASTYEAR — proper Date table দরকার।
• Sales table-এর date column-এ direct বললে — ভুল।
• Solution: Date dimension table তৈরি, mark as date table।

(৫) Bidirectional filter overuse:

• Convenience-এ all relationship bidirectional — performance disaster।
• Ambiguous filter path — model break।
• Default single-direction, bidirectional শুধু absolutely needed-এ।

(৬) CALCULATE-এ filter-replace বনাম filter-add:

// "Mobile sales" — Product filter REPLACE
Mobile = CALCULATE([Total Sales], Product[Cat] = "Mobile")

// "Mobile + Fashion" — boolean filter
Top2 = CALCULATE([Total Sales],
    Product[Cat] IN {"Mobile","Fashion"})

// FILTER দিয়ে preserve
Mobile2 = CALCULATE([Total Sales],
    FILTER(Product, Product[Cat]="Mobile"))

(৭) Iterator-এ row context vs filter context:

• SUMX — row context creates।
• Inside SUMX-এ measure call → context transition (row → filter)।
• Subtle bug source।

Debugging strategies:

(ক) Visual decomposition:

• Complex measure → তিন-চারটি intermediate measure ভাগ।
• প্রতিটি card visual-এ display — value verify।

(খ) DAX Studio:

• Server timings — কোন measure slow।
• Query plan — DAX engine কী করছে।
• Free tool, must-have।

(গ) Performance Analyzer:

• Power BI Desktop-এ built-in।
• প্রতিটি visual-এর time, query।
• Bottleneck pinpoint।

(ঘ) ISFILTERED, ISCROSSFILTERED:

• Conditional logic — current filter state-এ ভিন্ন behavior।
• Total vs subtotal handling।

(ঙ) HASONEVALUE:

• Single context vs multiple — গুরুত্বপূর্ণ।
• Total row-এ HASONEVALUE = false।

Bangladesh real-world example:

• Bank-এর "branch-wise NPL" dashboard।
• Total NPL ratio mismatch — branch-level filter context issue।
• Solution: branch dimension থেকে calculation, transaction table-এ থেকে না।

Learning curve:

• Week 1: SUM, AVERAGE basic।
• Month 1: CALCULATE, filter context।
• Month 3: Time intelligence।
• Month 6: Advanced patterns।
• Year 1+: Mastery।

মূল উপলব্ধি: DAX = SQL-এর চেয়ে গভীর, programming language ভাবে শিখতে হবে। "DAX Patterns" book (Marco Russo) bible।

Portfolio interview-এ technical skill-এর চেয়ে গুরুত্বপূর্ণ। Data ও story বাছাই critical।

Data source ideas (publicly available):

(১) Bangladesh Bureau of Statistics (BBS):

• Population Census ২০২২।
• Labor Force Survey।
• Household Income & Expenditure Survey (HIES)।
• Industry production index।

(২) Bangladesh Bank:

• Remittance inflow district-wise।
• Banking sector NPL ratio।
• Foreign exchange reserves।
• MFS transaction volume (bKash, Nagad)।

(৩) Election Commission:

• Election results 2008, 2014, 2018, 2024।
• Voter turnout by district।
• Party-wise vote share trend।

(৪) IEDCR / Health:

• COVID-19 case timeline।
• Dengue outbreak by district/month।
• Vaccination coverage।

(৫) Open Data:

• data.gov.bd portal।
• World Bank Bangladesh dataset।
• Kaggle "Bangladesh" search।

(৬) Sports/Entertainment:

• BPL cricket statistics।
• Bangladesh national team performance।
• Box office of Bengali films।

Recommended portfolio (3 dashboards):

Dashboard 1: "Bangladesh-এর জনসংখ্যাগত পরিবর্তন ১৯৭১-২০২২"

• Animated map — population growth।
• Age pyramid — demographic dividend story।
• Urbanization trend।
• Highlight: Tableau map skills, animation, storytelling।

Dashboard 2: "MFS Revolution: bKash, Nagad, Rocket Comparison"

• Time-series transaction volume।
• Geographic penetration map।
• Per-capita usage by district।
• Interactive filter — date range, MFS provider।
• Highlight: business analysis, KPI design।

Dashboard 3: "BPL ১১ Year Analytics" (entertainment-friendly)

• Player batting average over season।
• Team win-loss record।
• Top run-scorer Mexico-style timeline।
• Highlight: data wrangling, narrative chart।

Visualization selection criteria:

• One showcase chart — your unique technique।
• One simple-but-effective — KPI grid।
• One interactive — filter, parameter, drill-down।
• One geographic — Bangladesh map।
• One time-series — trend।
• Cohesive theme — color, font, layout।

Stand-out tips:

(১) Bengali typography:

• Title Bengali, axis label English mixed।
• Show local cultural awareness।

(২) Insight-driven, not feature-driven:

• "৩টি pattern detected" — bullet point।
• "Recommendation" section — business value।
• Story not technique।

(৩) Documentation:

• "Methodology" section — data source, cleaning steps, calculations।
• GitHub repo with Tableau workbook + cleaned CSV।
• README — clear।

(৪) Mobile responsive:

• Phone view design — recruiter মেইলে link খুললে।
• Device-specific layout (Tableau Public supports)।

(৫) Local context resonates:

• "Padma bridge before/after" — economic dashboard।
• "July Movement protests" — social media analysis।
• Recruiter-এর সাথে instant connection।

Interview talking points:

• "Why this dataset?" — curiosity-driven।
• "Data quality issues faced?" — problem-solving।
• "Iterations?" — version 1 vs final।
• "Limitations & next steps?" — honesty।

Distribution strategy:

• LinkedIn post weekly — one chart with story।
• Tableau Public profile — pinned at top।
• Personal website-এ embed।
• Bangladesh data viz community share (FB groups)।

মূল উপলব্ধি: Portfolio = your work, not certificate। Real Bangladesh data + insightful story = interview gold।