NEW
Font size
S
M
L
XL
WorksheetsBA 1
Total questions: 69
Worksheet time: 35mins
Name
Class
Date
1.
Which item is listed as a constraint in the 'Constraints, assumptions, risks' slide?
a)
Engineering resources
b)
Decision statement
c)
Office lunch menu
d)
Metric dictionary
e)
Scope boundary
2.
The success criteria and guardrails slide includes a product requirement related to what?
a)
Latency under a stated threshold
b)
Making dashboards colorful
c)
Increasing the number of KPIs
d)
Removing all guardrails
e)
Avoiding any experiments
3.
In the Impact/Effort matrix, which category is explicitly mentioned as a ranking outcome?
a)
Quick wins
b)
Legal briefs
c)
Talent reviews
d)
Security patches
e)
Brand refresh
4.
One A/B testing basic mentioned on the slides is to avoid what practice?
a)
Peeking at results early
b)
Randomization
c)
Choosing a primary metric
d)
Estimating test duration
e)
Ensuring group independence
5.
In the rule-of-thumb discussion for A/B test sample size, required sample size depends on baseline conversion rate and what else?
a)
Minimum detectable change
b)
Assumption
c)
Number of product categories
d)
Server region count
e)
Office location
6.
Quick Exercise 3 asks students to pick one option for A/B testing and define what alongside KPIs?
a)
Guardrails
b)
New office policies
c)
Brand slogans
d)
Employee titles
e)
Cloud vendor contracts
7.
An anti-pattern described is analytics work not tied to a decision. What is the recommended fix question to ask?
a)
What decision will this inform?
b)
Assumption
c)
Which desk layout is best?
d)
Who should attend the meeting?
e)
Decision statement
8.
The 'Wrap-up & Next Steps' section is described as covering what?
a)
References, homework, and what is next
b)
Only advanced calculus exercises
c)
Only legal compliance policies
d)
Only database administration tasks
e)
Guardrail metric
9.
Which statement is included in the key takeaways?
a)
KPI Trees connect the North Star to actionable levers
b)
A/B tests always require zero data
c)
CRISP-DM removes the need for business goals
d)
Business analytics is only about storage
e)
Dashboards should avoid filters
10.
Which is listed under suggested references?
a)
Data Science for Business
b)
The Art of War
c)
The Iliad
d)
Gray's Anatomy
e)
Metric dictionary
11.
The homework asks students to build what for a familiar digital product?
a)
A 3-level KPI Tree
b)
Data quality check
c)
A distributed file system
d)
Constraint
e)
Guardrail metric
12.
Prep for the next lecture includes hands-on work with what?
a)
SQL and exploratory data analysis (EDA)
b)
Robotics and control theory
c)
3D modeling and animation
d)
Formal verification of compilers
e)
Stakeholder map
13.
In the Q&A prompts, one discussion question asks about the biggest barriers to what?
a)
KPI Trees in practice
b)
Deploying satellites
c)
Building compilers
d)
Writing legal contracts
e)
Designing office layouts
14.
The problem framing canvas template includes which element?
a)
Context and goals
b)
GPU driver versions
c)
Keyboard shortcuts
d)
Baseline
e)
Hiring interview questions
15.
The blank KPI Tree template includes which component to define at the top?
a)
North Star
b)
Data drift threshold
c)
AUC target
d)
Guardrail metric
e)
Server rack layout
16.
In the CRISP-DM checklist slide, which phase name is included?
a)
Data Preparation
b)
Public Relations
c)
Payroll Processing
d)
Brand Management
e)
Hardware Procurement
17.
Which analytics layer is primarily concerned with recommending actions to achieve a desired outcome?
a)
Descriptive analytics
b)
Diagnostic analytics
c)
Predictive analytics
d)
Prescriptive analytics
e)
Data governance
18.
Which pairing best matches the slides' quick distinction between BI and AI/ML?
a)
BI: automate prediction in products; AI/ML: dashboards and KPI monitoring
b)
BI: dashboards and KPI monitoring; AI/ML: learn from data to automate prediction or recommendation
c)
BI: only data cleaning; AI/ML: only SQL queries
d)
BI: only experimentation; AI/ML: only data warehousing
e)
BI: code deployment; AI/ML: stakeholder management
19.
According to the data value chain principle on the slides, what should you start from when planning analytics work?
a)
The largest available dataset
b)
The newest machine learning model
c)
The decision to be supported
d)
The most detailed dashboard layout
e)
The most complex feature engineering
20.
Which combination of skills is listed as core skills for analytics roles on the slides?
a)
SQL, statistics, EDA, visualization, modeling, data storytelling
b)
Graphic design, video editing, motion capture, animation, sound mixing
c)
A stakeholder map that identifies decision owners and affected users
d)
Legal drafting, litigation, negotiation, compliance audits, contract redlining
e)
Customer support scripting, call routing, ticket triage, escalation policy, workforce management
21.
Which example best fits the 'trustworthy AI mindset' elements listed on the slides?
a)
Fairness and auditability
b)
Constraint
c)
More meetings and longer documents
d)
Cheaper laptops and more monitors
e)
Shorter project names and fewer dashboards
22.
In the e-commerce case, which set of levers is explicitly tied to growing revenue without increasing ad budget?
a)
Conversion rate, average order value, retention or repeat rate
b)
Number of warehouses, number of offices, number of employees
c)
A metric dictionary that standardizes KPI definitions and calculations
d)
A stakeholder map that identifies decision owners and affected users
e)
GPU upgrades, RAM upgrades, and CPU upgrades
23.
CRISP-DM is described as iterative. What is a valid reason to loop back in the process according to the slides?
a)
A new insight appears or the goal changes
b)
A metric dictionary that standardizes KPI definitions and calculations
c)
The team wants more meetings
d)
The office changes its seating plan
e)
A stakeholder map that identifies decision owners and affected users
24.
Which artifact is most aligned with the CRISP-DM emphasis on documenting the model for handover?
a)
Model card
b)
Office seating chart
c)
Brand tone guide
d)
Product packaging mockup
e)
Employee holiday calendar
25.
In Business Understanding, which statement best describes a guardrail KPI as used in the slides?
a)
A metric that limits unacceptable side effects while pursuing the main goal
b)
A guardrail metric used to avoid harming another key outcome
c)
A metric that replaces the need for business objectives
d)
A metric defined only after deployment
e)
A metric that is unrelated to decisions
26.
Why does the Data Understanding phase include reconciling KPI definitions with real data, as illustrated by order date vs ship date and returns?
a)
To ensure reported KPIs match the operational meaning in data
b)
To reduce the number of stakeholders
c)
To make dashboards load faster
d)
To avoid running any EDA
e)
To choose a cloud provider
27.
Which analysis is best suited for finding where an e-commerce conversion process is failing, based on the slides' Data Understanding quick wins?
a)
Funnel analysis across key events from session to purchase
b)
Guardrail metric
c)
Employee satisfaction survey
d)
Hardware stress testing
e)
Network penetration testing
28.
When segmenting by channel and device (mobile vs desktop; organic vs paid), what is the intended benefit mentioned on the slides?
a)
Identify which channel or device segments differ in performance
b)
Reduce data storage costs
c)
Eliminate the need for KPIs
d)
Guarantee causality without experiments
e)
Prevent any missing data
29.
Detecting leakage and double-counting of purchases due to retries or timeouts primarily protects which outcome?
a)
Accurate conversion and revenue measurement
b)
Higher screen resolution
c)
Lower office rent
d)
More colorful charts
e)
Faster keyboard input
30.
Which practice from the Data Preparation phase directly helps prevent training a model on information from the future?
a)
Using a time-based split for train and validation
b)
Adding more dashboard annotations
c)
Evaluation
d)
Reducing the number of stakeholders
e)
Skipping feature engineering
31.
Why does the slides' 'Data Prep quality & Feature Store' content emphasize lineage and versioning when reusing features?
a)
To ensure features are consistent and reproducible across models and time
b)
To make slide images smaller
c)
To avoid defining event names
d)
To remove the need for monitoring
e)
To eliminate the need for stakeholders
32.
Which modeling choice best matches the examples on the slides for predicting whether a user will churn?
a)
Classification
b)
Regression
c)
Principal component analysis
d)
Association rules
e)
Data warehousing
33.
Why do the slides recommend starting with a simple baseline model plus rules before tuning complex models?
a)
To create a reference point for performance comparisons using metrics like AUC or MAE
b)
To avoid needing any evaluation
c)
To guarantee the highest possible accuracy
d)
To remove the need for data preparation
e)
To prevent deployment monitoring
34.
Tuning a decision threshold is suggested to account for what business consideration?
a)
Different error costs and business goals
b)
Stakeholder map
c)
Office seating capacity
d)
Server rack dimensions
e)
Slide transition timing
35.
Which pairing best reflects the two evaluation perspectives emphasized in the slides?
a)
Technical metrics and business impact metrics
b)
A baseline comparison used to judge whether a change adds value
c)
Office metrics and meeting metrics
d)
A stakeholder map that identifies decision owners and affected users
e)
Salary metrics and vacation metrics
36.
According to the slides, why is experimentation (A/B tests and holdouts) part of evaluation?
a)
To measure uplift and guardrail impact under controlled comparison
b)
To replace the need for defining a target variable
c)
To avoid specifying success criteria
d)
To reduce the need for monitoring drift
e)
To eliminate segmentation
37.
In deployment, what is the purpose of canary or blue-green strategies as listed on the slides?
a)
Reduce rollout risk by controlling exposure during release
b)
Increase the number of KPIs
c)
Avoid defining an SLA
d)
Replace data quality tests
e)
Guarantee zero latency
38.
Which pitfall best explains the phrase 'Great model, zero business impact' on the CRISP-DM pitfalls slide?
a)
Lack of adoption or lack of experimentation despite good technical performance
b)
Using too many charts
c)
Collecting too much data
d)
Choosing a wrong file format
e)
A stakeholder map that identifies decision owners and affected users
39.
Which KPI choice best fits the slides' guidance to avoid vanity metrics unless tied to decisions?
a)
A metric that is actionable and supports a specific decision
b)
A metric chosen only because it is easy to display
c)
A metric that cannot be measured reliably
d)
A metric that has no owner or lever
e)
A metric that is unrelated to business goals
40.
In a KPI Tree, why is it useful to attach owners and levers to each branch?
a)
It clarifies accountability and makes drivers actionable
b)
It increases the number of dashboards required
c)
It replaces the need for a North Star metric
d)
It prevents any trade-offs from occurring
e)
It guarantees causality without tests
41.
Why do the slides recommend defining measurement windows for CR and AOV by channel and device?
a)
To make comparisons consistent and reduce biased measurement across segments
b)
To increase the number of metrics without purpose
c)
To avoid tracking events
d)
To eliminate bot traffic automatically without rules
e)
To ensure all users behave identically
42.
Which dashboard feature supports interpreting KPI changes in the context of known business events, according to the slides?
a)
Annotating events such as sales or UI changes
b)
Removing filters from dashboards
c)
Hiding the North Star metric
d)
Avoiding any weekly reviews
e)
Displaying only raw tables without context
43.
The problem framing template uses the DOC structure. What does DOC represent on the slides?
a)
Decision, Options, Criteria
b)
Data, Outputs, Code
c)
Design, Operations, Compliance
d)
Diagnosis, Optimization, Calibration
e)
Deployment, Ownership, Cadence
44.
Which mapping from business question to analytics task matches the slides' example for revenue growth without increasing ads?
a)
Diagnostic: find funnel bottlenecks; Prescriptive: prioritize improvements with highest ROI
b)
A stakeholder map that identifies decision owners and affected users
c)
Clustering: rename events; Optimization: change office layout
d)
Regression: build a KPI tree; Classification: design a dashboard
e)
Modeling: skip preparation; Evaluation: skip experiments
45.
Why do the slides recommend excluding variables generated after the decision point when defining features for a predictive target?
a)
To avoid leakage that would inflate apparent model performance
b)
To increase the number of available features
c)
To make dashboards more colorful
d)
To avoid time-based splits
e)
To eliminate the need for monitoring
46.
In the success criteria examples, why are technical criteria like acceptable uncertainty and product criteria like latency included alongside business KPIs?
a)
They ensure the solution is reliable and usable in production, not just profitable on paper
b)
They replace the need for defining revenue impact
c)
A leading indicator that changes before the main outcome changes
d)
They eliminate the need for guardrails
e)
They remove the need for a decision owner
47.
In an Impact/Effort matrix, which option best describes a 'quick win' as used on the slides?
a)
High impact with relatively low implementation cost
b)
Low impact with high implementation cost
c)
High impact with high implementation cost only
d)
Low impact with low implementation cost only
e)
An item that should always be avoided
48.
Which A/B testing practice on the slides most directly helps avoid false conclusions from repeated checking mid-test?
a)
Avoid peeking at results before the test completes
b)
Use dashboards without primary metrics
c)
Change the randomization unit every day
d)
Skip estimating test duration
e)
Remove group independence assumptions
49.
A team wants to optimize mobile checkout to increase revenue. Based on the slides, which combination best aligns with decision-centered analytics planning?
a)
Start by collecting every possible dataset, then decide what to optimize
b)
Define the decision to prioritize checkout improvements, set a revenue KPI, and add a conversion-rate guardrail
c)
Build a complex model first, then look for a business use case
d)
Skip KPIs to avoid constraining the team
e)
A measurable success criterion linked directly to the decision outcome
50.
You are predicting 'purchase within 7 days since session' and include 'applied voucher' as a feature. Why is this problematic according to the slides, and what is the correct fix?
a)
It reduces model size; fix by adding more regularization
b)
It causes leakage because the voucher is generated after the decision point; fix by excluding post-decision variables
c)
It reduces AUC; fix by switching to clustering
d)
It increases latency; fix by removing monitoring
e)
It changes the randomization unit; fix by using sessions instead of users
51.
Which scenario best illustrates inconsistent KPI definitions across teams causing issues, as warned in the CRISP-DM pitfalls and Data Understanding checklist?
a)
Two teams use different definitions of 'transaction date' (order date vs ship date), producing conflicting revenue reports
b)
A guardrail metric that prevents improving one KPI by hurting another
c)
A stakeholder map that identifies decision owners and affected users
d)
Two teams use different meeting tools
e)
A phase that delivers the solution into operational use
52.
A KPI Tree has overlapping branches that both count the same driver, and the North Star seems to improve without business impact. Which root cause and remedy are most consistent with the slides?
a)
Root cause: too many stakeholders; remedy: remove RACI
b)
Root cause: double counting from overlapping branches; remedy: redesign branches to be mutually exclusive and add guardrails
c)
Root cause: missing dashboards; remedy: add more boards
d)
Root cause: slow laptops; remedy: upgrade hardware
e)
Root cause: too many events; remedy: stop instrumentation
53.
Which choice best explains why the slides recommend attaching owners and levers to KPI Tree branches AND setting update cadence and dashboard ownership?
a)
To reduce the number of KPIs regardless of decisions
b)
To ensure accountability and operational follow-through, so metrics inform weekly decisions and are maintained
c)
To eliminate the need for EDA
d)
To guarantee that models are unbiased
e)
To avoid any trade-offs between metrics
54.
An e-commerce team raises AOV using bundles, but conversion rate drops slightly. Based on the slides' use of guardrails, what is the correct evaluation framing?
a)
Ignore conversion because AOV is the chosen driver
b)
Compare options using primary revenue impact while enforcing a conversion-rate guardrail threshold
c)
Evaluate only technical metrics like AUC
d)
Delay evaluation until after deployment without monitoring
e)
Replace guardrails with vanity metrics
55.
During Data Understanding, you detect sudden spikes in purchases that later disappear due to returns and cancellations. Which slide principle best addresses this measurement issue?
a)
Use canary releases to reduce rollout risk
b)
Define canonical KPI definitions that account for returns and cancellations
c)
Skip reconciling KPI definitions with data
d)
Only track page_view events
e)
Avoid segmenting by device
56.
A model performs well on historical data but degrades after rollout due to changing user behavior. According to the deployment and evaluation content, what should be in place to detect and respond?
a)
Only a static EDA report
b)
Monitoring for drift with alerting and an auto-rollback plan, plus cohort stability checks
c)
A one-time dashboard snapshot with no updates
d)
More complex feature engineering without monitoring
e)
A requirement to avoid any experiments
57.
Why do the slides emphasize time-aware cross-validation and time-based train/validation splits for many business problems?
a)
They make models smaller and faster to run
b)
They help prevent leakage and better reflect future performance when data has time ordering
c)
They remove the need for defining success criteria
d)
They guarantee higher revenue uplift
e)
They replace the need for data quality checks
58.
A team wants to declare success based on improved AUC, but the business KPI does not move and adoption is low. Which combination of slide guidance best addresses this failure mode?
a)
Focus only on more hyperparameter tuning
b)
Add business evaluation, experimentation plans, and decision-centered problem framing to drive adoption and impact measurement
c)
Remove all stakeholders from the project
d)
Stop monitoring and end deployment early
e)
Switch from KPI trees to vanity metrics
59.
In the slides' A/B testing basics, why is choosing the randomization unit (user vs session) and ensuring group independence critical?
a)
It ensures dashboards can show heatmaps
b)
It helps ensure the comparison is valid and reduces contamination between groups
c)
It increases the sample size automatically
d)
It makes conversion rate larger
e)
It eliminates the need to estimate test duration
60.
You build a funnel using events but the event taxonomy is inconsistent across teams and some data arrives late. Which set of actions is most consistent with the Data Prep quality slide?
a)
Accept inconsistencies and rely on the model to fix them
b)
Standardize taxonomy and event naming, maintain a data dictionary, and account for late-arriving data
c)
Remove uniqueness tests to speed up pipelines
d)
Skip schema and freshness tests
e)
Stop collecting events and use only orders
61.
A company wants a KPI dashboard that supports weekly decisions. Which dashboard design elements from the slides best help interpret changes and drive action?
a)
Only a single table of raw events with no context
b)
North Star with driver views, funnel breakdowns by device/channel, and annotations for major events
c)
A lagging indicator that confirms results after actions occur
d)
A dashboard with no filters to avoid complexity
e)
A dashboard that refreshes rarely to reduce costs
62.
Which outcome best reflects the slides' warning about 'metrics lag decisions' as a KPI Tree pitfall?
a)
The team updates metrics daily and uses them in weekly reviews
b)
Metrics refresh slowly, so decisions are made before the numbers reflect reality
c)
The team defines guardrails for key trade-offs
d)
The team standardizes event naming
e)
The team uses time-based splits for modeling
63.
Suppose you optimize a leading indicator in a KPI Tree (like a funnel step), but a health metric (like payment errors) worsens. Based on the slides, what is the best interpretation?
a)
Health metrics are optional and can be ignored
b)
A guardrail is being violated, so the optimization may be unacceptable despite driver improvement
c)
Payment errors are a vanity metric
d)
This proves the model is unbiased
e)
This guarantees long-term revenue growth
64.
The slides list several CRISP-DM artifacts (charter, data dictionary, model card, runbook). Which combination best supports reliable long-term operation after deployment?
a)
Only a model card
b)
Only a project charter
c)
A rollout plan, monitoring dashboard, and operations runbook with continuous improvement plan
d)
Only an EDA report
e)
Only a KPI tree diagram
65.
A team chooses many KPIs including vanity metrics, and no one owns them. Which paired slide principles would most directly correct this?
a)
More KPIs plus less documentation
b)
Fewer but better KPIs plus clear owners and levers tied to decisions
c)
Avoid defining guardrails plus faster refresh
d)
Skip problem framing plus more dashboards
e)
Use only technical metrics plus no RACI
66.
Using the slides' sample-size guidance, what change would generally increase required per-group sample size the most?
a)
Looking for a smaller minimum detectable change in conversion rate
b)
Choosing a larger minimum detectable change in conversion rate
c)
Adding annotations to dashboards
d)
Using a shorter backlog window
e)
Standardizing event naming
67.
A project has constraints (resources, rollout time, SLA) and risks (policy changes, supply disruptions). Based on the slides, where should these be explicitly captured to reduce failure risk?
a)
Only in a dashboard heatmap
b)
In the problem framing canvas and Business Understanding outputs such as scope, timeline, and risks
c)
Only in model hyperparameters
d)
Only in deployment logs after release
e)
Only in the glossary
68.
A team runs an experiment but keeps checking results daily and stops early when the primary metric looks good. Which slide guidance is being violated, and what is the likely consequence?
a)
Violation: define more KPIs; consequence: slower dashboards
b)
Violation: avoid peeking and use careful sequential testing; consequence: higher chance of false positive conclusions
c)
Violation: set randomization unit; consequence: smaller sample size
d)
Violation: deploy with blue-green; consequence: higher latency
e)
Violation: build a KPI tree; consequence: weaker SQL queries
69.
A model is trained using a feature called 'price_sensitivity', but the production pipeline computes it differently after a code change. Which slide concept is most directly intended to prevent this issue?
a)
Using more vanity metrics
b)
Feature store with lineage and versioning to maintain feature-model consistency
c)
Avoiding any dashboards
d)
Skipping data tests for uniqueness
e)
Replacing time-based splits with random splits
Reset
