There are 4 types of analytics in data world:
1) Descriptive Analytics → What happened?
- It summarizes past data
- Example: “Sales increased 15% last month”
Used visuals: tables, cards, bar charts
2) Diagnostic Analytics → Why did it happen?
- It explains the reason behind the result
- Example: “Sales increased because ads performed well”
Tools: Decomposition tree, Key Influencers
3) Predictive Analytics → What will happen?
- It forecasts the future
- Example: “We predict a 20% drop in churn next month”
Tools: ML models, forecasting, AutoML
4) Prescriptive Analytics → What should we do?
- It suggests the best action based on predictions
-
Example:
- “Offer a discount to high-risk customers”
- “Increase ad budget for Segment A”
- “Send follow-up emails to low-engagement users”
💡 Prescriptive = ML + recommended action
Descriptive — What happened?
Purpose: summarize past data. Tools: cards, tables, bar/line charts.
Example exam cues: “show totals”, “last month’s sales”.
Diagnostic — Why did it happen?
Purpose: explain root cause. Tools: Key Influencers, Decomposition Tree, drillthrough.
Example cues: “identify drivers”, “explain drop in revenue”.
Predictive — What will happen?
Purpose: forecast future events using ML. Tools: AutoML, forecasting, classification/regression models.
Example cues: “predict”, “probability”, “forecast”.
Prescriptive — What should we do?
Purpose: recommend actions (next-best-action / optimization). Tools: recommendation engines, optimization models, ML + action mapping.
Example cues: “recommend”, “suggest action”, “optimize”, “next best offer”.
Exam rule of thumb: If the question asks what action to take, think prescriptive → ML + action.
What is prescriptive analytics?
Prescriptive analytics is the step beyond predicting. It answers: “Given what we know and what’s likely to happen — what should we do now?”
Core elements:_ An ML or optimization model_ that produces predictions or scores.
Decision logic that maps predictions to actions (rules, optimization objective, cost/benefit).
Integration into reports or systems to display or enact the recommended actions.
Typical output: a column or visual that says “Offer 20% discount” or “Route to specialist” — not just “churn probability = 0.85”.
*Embedding an Azure Machine Learning model into Power BI: how it works *
- Train model (Azure ML, Fabric AutoML, Python/R).
- Deploy model as a web service (REST endpoint).
- Call endpoint from a dataflow / Power Query / Fabric pipeline or call it outside and bring scored data in.
- Use outputs in Power BI as calculated columns or visuals.
- Map predictions to actions with DAX, Power Query logic, or an operationalization layer (e.g., if churn_prob > 0.8 → “Offer 30% off”).
- Present recommendations in report visuals (tables, cards, conditional formatting).
Why this is preferred:
- Dynamic (updates with new data)
- Automated and scalable
- Can incorporate business rules and thresholds
Static text boxes & manual updates — when they appear (and why they’re wrong for prescriptive)
Static text boxes: useful for titles, descriptions, instructions—never for dynamic recommendations.
Manual updates: feasible for tiny teams, but not production, not scalable, and usually wrong on exams when compared with ML/automation options.
Exam shortcut: Options including “static text”, “manual monthly updates”, or “historical visuals only” are almost always incorrect if the question asks for suggestions/recommendations/next actions.
When to use forward-looking suggestions vs historical-only
Use forward-looking (predictive/prescriptive) when: the question mentions recommend, suggest, next best action, optimize, forecast.
Use historical-only (descriptive/diagnostic) when: the question asks report, summarize, view past trends, explain reasons.
Be strict: “suggest” = prescriptive → prefers an ML-based, automated approach in DP-600 questions.
Question (demo): You need to add a feature to a Power BI report that suggests targeted marketing actions based on customer segment performance. Which technique integrates prescriptive analytics effectively?
Options:
A. Embedding an Azure Machine Learning model within the report to analyze segments and suggest actions.
B. Using static text boxes to provide generic marketing advice.
C. Manually updating the report each month with new recommendations.
D. Relying solely on historical data visualizations without forward-looking suggestions.
Step 1 — identify keywords: “suggests targeted marketing actions” → suggest = prescriptive; “targeted” = personalization; “based on performance” = data-driven.
Step 2 — map to analytics type: That’s prescriptive (what action to take).
Step 3 — choose solution: Only A (embedding Azure ML) supplies automated, personalized, data-driven recommendations.
Why others wrong: B is static; C is manual and not scalable; D is descriptive only.
Answer: A. Embedding an Azure ML model.
Exam tip: Always ask yourself “does this option produce an actionable output per user/segment that changes dynamically?” If yes — ML/endpoint/automation is right.
“What happened?” → Descriptive
“Why did it happen?” → Diagnostic
“What will happen?” → Predictive
“What should we do?” → Prescriptive
If question asks for recommendations / next best action / optimization, pick options mentioning ML / AutoML / Azure ML / optimization / recommendation engine.
If an option includes static text, manual process, monthly update, it’s almost always wrong for predictive/prescriptive needs.
For diagnostic questions, favor Key Influencers or Decomposition Tree.
For predictive questions, favor AutoML / time-series forecasting / classification/regression.
*Implementation mapping * — common patterns you should know
- Churn management: Predictive model → score customers → map to “Offer discount / Call / No action”.
- Recommendation engine (retail): Collaborative filtering or content-based recommenders → store top N recommendations per user → display in Power BI.
- Inventory optimization: Forecast demand (time series) → run optimization for reorder quantities → provide reordering actions.
- Predictive maintenance: Predict failure probability → if > threshold → schedule maintenance action.
- Fraud detection: Predict risk per transaction → rules map to block / monitor / allow.
⭐ What is "Embedding an Azure Machine Learning Model"?
Embedding means connecting Power BI to a deployed Azure ML model so it can:
- Score data
- Predict outcomes
- Output recommended actions
- Generate prescriptive insights
How it works (simple flow):
- Build ML model
- Deploy it as a web service (API endpoint)
- Call that model from Power BI
- The model outputs a recommendation like:
- “Recommend Email Campaign B for Customer X”
- “Upsell Product Y to Segment Z”
This is dynamic, scalable, and automated.
⭐ What is a "Static Text Box"?
A static text box is simply text typed manually inside a Power BI report.
Example texts:
- “Increase marketing spend”
- “Improve engagement”
❌ It is not connected to data
❌ It cannot change dynamically
❌ It is not analytics
Used only for:
- titles
- explanations
- notes
- labels
Never used for prescriptive analytics.
⭐ What is "Manual Updating"?
You manually type new recommendations every month like:
“This month focus on Segment B”
Problems:
- not scalable
- error-prone
- slow
- not data-driven
- changes do not react to real-time data
❌ This is not analytics
❌ Completely ignored in AI/ML questions
Only useful for:
- temporary reports
- low-impact dashboards
- small teams
Exam trick:
Manual work is always the WRONG answer in DP-600 for anything AI/ML/autonomous.
⭐ When do we need "Forward-Looking Suggestions"?
Forward-looking = predictive + prescriptive
Used when:
- you must take action
- business impact depends on future outcome
- you want to optimize decisions
Examples:
- churn prevention
- retention strategies
- budget optimization
- marketing targeting
- fraud prevention
Exam Tip
If question says:
- “suggest”
- “action”
- “recommend”
- “next best step”
- “optimize”
➡️ Answer is ALWAYS ML model.
⭐ When do we rely only on historical data?
When the question focuses on:
- understanding past performance
- monitoring KPIs
- root cause analysis
- business reporting
Not for making decisions.
Historical-only = descriptive or diagnostic
Exam clue:
If options contain things like “historical only,” it will ALWAYS be the wrong answer for prescriptive/predictive scenarios.
🔥 PART 2 — Question Breakdown (Step by Step)
❓ Original Question
You need a feature that suggests targeted marketing actions based on customer segment performance.
Step 1 — Identify keywords
🟩 “suggests actions” → prescriptive
🟩 “marketing actions” → decision-making
🟩 “based on segment performance” → ML/AI scoring
➡️ So the correct option must include:
- ML model
- automated recommendations
- dynamic logic
🔥 OPTION-BY-OPTION BREAKDOWN
A. Embedding an Azure Machine Learning model
✔️ ML Model → Dynamic
✔️ Can generate recommendations
✔️ Prescriptive
✔️ Auto decision-making
✔️ Scalable
➡️ Correct
B. Using static text boxes
❌ Not data-driven
❌ Not dynamic
❌ Only generic advice
❌ Cannot target segments
➡️ Wrong
C. Manually updating the report
❌ Not scalable
❌ Not analytics
❌ Prone to errors
❌ Exam always rejects manual work
➡️ Wrong
D. Relying solely on historical data
❌ Only descriptive/diagnostic
❌ No suggestion capability
❌ No future insight
➡️ Wrong
⭐ Tips & Tricks for DP-600 MCQs (VERY IMPORTANT)
Here is the cheat sheet to instantly identify correct answers:
✔️ If question mentions recommendations, answer = ML model
Keywords:
- next best action
- suggest
- recommend
- optimize
- prescriptive
✔️ If question mentions future, answer = AutoML, ML, forecasting
Keywords:
- prediction
- forecast
- future trend
- probability
✔️ If question mentions understanding the past, answer = visuals
Keywords:
- historical
- summarize
- monitor
- trend analysis
✔️ If option contains manual, static, text box, it's ALWAYS wrong
🔥 PART 3 — Notes (Documentation Style, Clear Examples)
📘 Prescriptive Analytics in Fabric / Power BI
Prescriptive analytics uses machine learning or decision algorithms to recommend actions.
Tools Used:
- Azure ML models
- Fabric ML models
- AutoML
- Python/R notebooks
- ML scoring in dataflows
Example:
Dataset: Customer segments
Model Output:
- Segment A → Recommend email offer
- Segment B → Recommend discount
- Segment C → No action
Power BI shows recommendation column dynamically.
🔥 PART 4 — 10 Similar Questions (With Answers + Explanations)
Q1
You want Power BI to select the best offer for each customer. How do you implement this?
A. Use an Azure ML recommendation model
B. Add a table showing purchase history
C. Write static text suggestions
D. Update the report manually each week
Answer: A
Q2
You need a feature that tells sales teams whom to call first based on predicted closing probability.
A. Deploy a predictive model with scoring rules
B. Use a bar chart by region
C. Place a slicer for sales rep
D. Show the last quarter’s performance
Answer: A
Q3
You want to reduce churn by recommending retention actions.
A. Build a churn ML model + action mapping
B. Show all historical churn in a chart
C. Add commentary in a static textbox
D. Update recommendations manually
Answer: A
Q4
A report must decide whether to reorder inventory based on predicted demand.
A. Use a demand forecasting model
B. Show a line chart of stock
C. Add a table with past orders
D. Provide generic reorder advice
Answer: A
Q5
Marketing wants a "next best campaign" suggestion dashboard.
A. Integrate an ML model that outputs recommended campaigns
B. Use a KPI visual
C. Display last year engagement
D. Add comments explaining campaign types
Answer: A
Q6
You want insights on why sales dropped last quarter.
A. Use decomposition tree (diagnostic)
B. Use ML recommendation model
C. Add manual comments
D. Use a line chart only
Answer: A
Q7
You need a fully automated system that suggests actions daily.
A. ML model + automation
B. Static text updated monthly
C. Manual tags
D. Historical-only dashboards
Answer: A
Q8
Which feature predicts future sales?
A. AutoML or forecasting
B. Static text
C. Manual trend writing
D. Card visual
Answer: A
Q9
Which is NOT prescriptive analytics?
A. Optimization model
B. Next best action
C. Static business advice in text box
D. Recommendation engine
Answer: C
Q10
Which delivers dynamic insights?
A. ML model
B. Static text
C. Manual report editing
D. Hard-coded comments
Answer: A
Question 1 of 10
You need to add a feature to a Power BI report that recommends the best retention action for customers predicted to churn. Which method implements prescriptive analytics correctly?
A. Integrating an Azure ML churn model that outputs recommended actions for each customer.
B. Adding a static text box with general retention advice.
C. Showing a historical churn trend without predictions.
D. Updating the report manually every month with retention suggestions.
✅ Correct Answer: A. Integrating an Azure ML churn model that outputs recommended actions for each customer.
This implements prescriptive analytics because the ML model can predict who is likely to churn and then recommend specific actions (e.g., discount, call, email). It is automated, data-driven, and scalable.
❌ B. Static text box
Static text is generic and NOT connected to data. It cannot provide personalized retention advice → NOT prescriptive.
❌ C. Historical trend
Historical churn charts are descriptive, not prescriptive. They explain the past but cannot suggest actions.
❌ D. Manual updates
Manual work is not scalable, prone to errors, and NOT analytics-driven.
Question 2 of 10
A sales dashboard must suggest which leads the team should prioritize each day. Which approach delivers true prescriptive analytics?
A. Using an ML scoring model to rank leads by probability of closing.
B. Showing a line chart of last month’s sales.
C. Adding a generic “Follow up with leads” text box.
D. Having analysts manually rank leads weekly.
✅ Correct Answer: A. Using an ML scoring model to rank leads.
ML scoring dynamically identifies high-value leads and automates prioritization → prescriptive.
❌ B. Historical chart
Only descriptive, no decision-making.
❌ C. Generic text advice
Not personalized, not dynamic.
❌ D. Manual ranking
Not scalable, not data-driven.
Question 3 of 10
A retail analytics report should recommend whether each product should be reordered or discontinued based on demand forecasts. What should you implement?
A. A demand forecasting ML model that outputs a recommended stock action.
B. A table that lists last year’s product sales.
C. A static note saying “Monitor inventory frequently.”
D. Monthly manual updates from supply chain staff.
✅ Correct Answer: A. Demand forecasting ML model
It provides future prediction + recommended action, which is exactly prescriptive analytics.
❌ B
Only historical → descriptive.
❌ C
Static → no analytics.
❌ D
Manual → not scalable, not intelligent.
Question 4 of 10
You want Power BI to automatically decide the best promotional offer for each customer segment. Which solution is correct?
A. Deploying a recommendation ML model and integrating it into the report.
B. Adding a bar chart showing which promotions performed best last year.
C. Writing generic promotional tips in a text box.
D. Updating the promotion recommendations manually.
✅ Correct: A
This uses ML to determine the best action, which is the core of prescriptive analytics.
❌ B
Descriptive only.
❌ C
Static & not dynamic.
❌ D
Manual → wrong for DP-600.
Question 5 of 10
A financial analytics report must advise departments on how to reduce expenses. How should this be implemented?
A. Using an ML optimization model that outputs recommended cost-saving actions.
B. Displaying historical spending trends.
C. Adding a static text box with general budgeting tips.
D. Manually adding cost-saving suggestions each quarter.
✅ Correct: A
Optimization models + ML = true prescriptive analytics.
❌ B
Only describes past spending.
❌ C
Static → no intelligence.
❌ D
Manual → not scalable.
Question 6 of 10
Your customer engagement dashboard must recommend the “next best action” for users with low engagement. What method should be used?
A. Build and integrate a next-best-action ML model.
B. Show a table of users with their engagement score.
C. Add a text box describing how to improve engagement.
D. Let analysts manually email suggestions to teams weekly.
✅ Correct: A
Prescriptive analytics = ML-based recommendations.
❌ B
Descriptive.
❌ C
Static generic advice.
❌ D
Manual process → wrong.
Question 7 of 10
A manufacturing report must suggest machine maintenance actions before a failure occurs. Which approach is correct?
A. Use a predictive maintenance model connected to Power BI to suggest preventive actions.
B. Show last month’s machine downtime data.
C. Use a static text explaining maintenance guidelines.
D. Rely on manual monthly updates from technicians.
✅ Correct: A
This is prescriptive: prediction → recommended action.
❌ B
Historical only.
❌ C
Static text.
❌ D
Manual → not prescriptive.
Question 8 of 10
A subscription business wants the report to recommend actions for customers predicted not to renew. What method should be used?
A. Use an ML renewal prediction model and map scores to recommended retention actions.
B. Add a chart showing renewal rates last year.
C. Write static renewal advice.
D. Update the file manually each month.
✅ Correct: A
Prediction + Action = prescriptive.
❌ B
Descriptive.
❌ C
Static.
❌ D
Not scalable.
Question 9 of 10
Your dashboard must automatically suggest the best discount for each product based on profitability and forecasted demand. Which solution is correct?
A. Integrate an ML model that calculates optimal discount levels.
B. Show last year’s discount performance.
C. Add a generic text box explaining discount strategies.
D. Let managers manually pick discount levels.
✅ Correct: A
ML-driven optimal action → prescriptive.
❌ B
Historical.
❌ C
Static.
❌ D
Manual → wrong.
Question 10 of 10
A Power BI report must recommend which suppliers to prioritize based on reliability predictions. What should be implemented?
A. An ML model that predicts supplier reliability and outputs recommended choices.
B. A matrix of historical supplier performance.
C. A text box summarizing procurement guidelines.
D. Manual supplier ranking updated monthly.
✅ Correct: A
Predict → recommend → prescriptive.
❌ B
Descriptive only.
❌ C
Static info.
❌ D
Manual = always wrong for prescriptive questions.
Top comments (0)