Module 1 – Decision making, data

and analytics

Business Logic
Dr. Tuan Truong
Learning outcomes

01 02 03
Identifying and Classifying Explaining and
clearly articulating decisions: using the
a business Strategic – Tactical framework:
decision – Operational Decision →
Information → Data
What decisions does a business have to make
• Choose a familiar business (café / e-commerce / store / app, etc.)
• List 5 decisions they have to make in one month.
• For each decision: “What decision?” + “Who made it?” (Decision |
Who)

• Note: The decision must be a “choice between options,” not an “operation”—

remember the decision point in BPMN?
Why start with
"decision"?
• Businesses have dashboards, but
they lack clear decision-making.
• Good data = faster/more
accurate/measurable decision-
making.
• Know how to ask the right
questions before searching for
data.
 how good data creates business advantage
•Relevant to a decision
•Timely enough to act
•Accurate enough to trust
•Consistent (same definition across teams)
•Granular enough to diagnose (segment, channel,
cohort)
•Actionable (leads to a clear next step)

“Top 5 underperformers show normal traffic but low

conversion during evenings → adjust staffing + retrain floor
staff.” – Retail operation insights

Decision → Question → Metric → Data → Insight → Action

 Decision-first question E-commerce marketing budget decision
•Decision: We need to decide whether to increase the Google Search
budget for next week.
•Decision: We need to decide whether to ________. •Options:
•Options: A / B / C A) Increase budget +20%
•Success metric: We optimize for __________. B) Keep budget unchanged
•Constraints: Must keep __________ above/below C) Decrease budget -20% / pause low-performing campaigns
•Success metric: We optimize for profit per day (or contribution
__________.
margin), not just revenue.
•Time window: We measure impact over •Constraints: Must keep ROAS ≥ 2.0 and gross margin ≥ 25%; do not
__________. exceed daily inventory capacity for top sellers.
•Segments: We will check by __________. •Time window: Measure impact over 7 days (next week), with a daily
•Trigger thresholds: If metric is __________ then do check for early warning.
__________. •Segments: Break down by campaign (brand vs non-brand), device
•Owner: __________ decides; __________ executes. (mobile vs desktop), and new vs returning customers.
•Trigger thresholds:
•If ROAS < 2.0 for 2 consecutive days → reduce spend on those
campaigns by 20% and shift budget to best performers.
•If profit/day increases ≥ 10% while constraints hold → scale
budget by +20%.
•Owner: Growth/Marketing Lead decides; Performance Marketer
executes changes; Data Analyst validates tracking and reports results.
What is business
decision
• Decision = choosing one option from
several options to achieve a goal
(within constraints)
• 4 core components:
• Goal: What do you want to
achieve?
• Options: What options are
available?
• Criteria/Metrics: What criteria
are used to make the choice?
• Constraints (budget, time, legal,
etc.)
3 types of decision
• Strategic: Long-term direction (quarter-
year), changing the way
• Tactical: Optimizing medium-term plans
(week-month), allocating resources
• Operational: Daily operations, execution
& exception handling
• Example – Restaurant Decisions
• Strategic: Open more branches? Segment
positioning?
• Tactical: This month's promotion?
Seasonal menu?
• Operational: Table priority? Handling late
orders? Shift scheduling?
Stakeholders Unclear owner → “Many people
& Decision offer suggestions, but no one
takes responsibility”
Owner
Decision Owner = the person
who finalizes the decision and
is responsible for the outcome

Stakeholders = those who

influence / implement /
receive the output
 RACI

• RACI - Task-Focused Responsibilities

• Responsible: The doer(s) who
perform the work.
• Accountable: The one person
ultimately answerable for task
completion (must be only one).
• Consulted: Those who provide
input (two-way communication).
• Informed: Those who need to be
kept in the loop (one-way
communication).
DACI

• DACI - Decision-Focused Ownership

• Driver: Manages the decision
process, gathers info, and pushes for
a decision.
• Approver: The ultimate decision-
maker with final authority (the
"Decision Owner").
• Contributors: Subject matter experts
who provide input and
recommendations.
• Informed: Stakeholders kept updated
on the decision and its outcomes
RACI vs DACI
• Focus: RACI for tasks/workflows; DACI
for decisions/approvals.
• Decision Ownership: DACI has a clear
Approver, making it great for clarity on
who decides; RACI's Accountable role is
about overall success, not just the final
'yes/no'.
• Best For: RACI in large projects with
many tasks; DACI in situations needing
quick, clear decisions, like product
development or strategy
Decision making process
• Intelligence - Discovering,
identifying, and
understanding the problems
occurring in the organization
• Design - Identifying and
exploring solutions to the
problem
• Choice - Choosing among
solution alternatives
• Implementation - Making
chosen alternative work and
continuing to monitor how
well solution is working
• Monitoring and feedback →
learning & adjusting
Simon’s Decision-Making Process
Remember the question: what major should I apply to?
Where does "data" appear in the process?
• Intelligence: Anomalous signals (drop/spike)
• Design: Hypothesis & verification
• Choice: Trade-off simulation / Expected impact
• Implementation: Monitoring, guardrails
• Feedback: Impact measurement, learning from experience
 Basic framework - Decision → Information → Data
• Decision → Information → Data
• Decision: What decision needs to
be made?
• Information: What information is
needed to make the decision? (3–5
sentences)
• Data: What data is needed to
answer the question?
(field/table/source)
15 minutes Identify a decision
exercise
Who is/are decision owner

Who are stakeholders

What information we need to give

decision

What types of data we need, in what

departments or functions
Evolution of Computerized Decision Support to
Analytics/Data Science

Figure 1 Evolution of Decision Support, Business Intelligence, Analytics, and AI

How we transform data to information?
• transform data into information through a process called data
processing or data transformation
• Collection: Gather raw data from various sources (databases, sensors,
forms).
• Processing/Transformation: Clean (fix errors, remove duplicates), structure
(organize), and enrich (add context) the data.
• Analysis: Apply techniques to find patterns, trends, and relationships.
• Interpretation & Presentation: Contextualize findings and present them
clearly (charts, reports) for understanding and action
Business
Analytics
Figure 1.21 Three
Types of Analytics
 • Descriptive or reporting analytics
• Answering the question of what
happened
Descriptive • Retrospective analysis of historic data
• Enablers
Analytics • OLAP/DW
• Data visualization
• Dashboards and Scorecards
• Descriptive statistics
 Descriptive Analytic Examples
Category Techniques / Methods Decision question Typical outputs Example use-case
KPI reports, dashboards, “Are we on track vs target, and where Daily/weekly KPI Daily revenue, conversion rate,
Reporting
scorecards are we off?” views, alerts AOV by channel
OLAP / Roll-up, drill-down, slice-and- “Which segment/product/channel is Pivot views by Sales by region → by store → by
Aggregation dice driving the change?” dimension SKU
Descriptive Mean/median, variance, “How big is the change and how Summary stats, Median delivery time vs average
statistics percentiles, growth rates, shares unusual is it?” distribution snapshots (spot skew/outliers)
Distribution checks,
“Is the data trustworthy, and what Data quality + pattern Find refund spikes linked to a
EDA missing/outlier profiling,
patterns stand out?” findings product batch
correlations
Funnel Step conversion rates, drop-off “At which step are we losing customers Identify checkout step causing
Funnel drop-off report
analysis by stage most?” abandonment
Cohort Cohort retention/repurchase “Did performance improve/worsen for Retention Retention by signup month
analysis curves newer cohorts?” tables/curves after pricing change
Rule-based RFM, demographic/behavior “Which customer groups behave “High-value repeats” vs “one-
Customer segments
segmentation splits differently and need different actions?” time discount buyers”
80/20 contribution, A/B/C item “Where should we focus to get the Top 20% products generate
Pareto / ABC Concentration insights
classes biggest impact?” 75% of margin
Simple
Control charts, z-scores, moving “Is this spike/dip a real issue requiring Alerts on abnormal Alert when conversion drops
anomaly
averages action?” shifts beyond normal seasonality
detection
Predictive Analytics
• Aims to determine what is likely to happen in the future
(foreseeing the future events)
• Looking at the past data to predict the future
• Enablers
• Data mining
• Text mining/Web mining
• Forecasting (i.e., time series)
 Predictive analytics techniques

What it
Category Techniques / Methods Decision question Example use-case
predicts
“How much will demand/sales be if Continuous Predict weekly demand per
Regression Linear regression, Ridge/Lasso
drivers change?” values product
Logistic regression, trees, RF, “Which cases are most likely to be X vs Classes / Predict churn / fraud / lead
Classification
XGBoost, SVM, NN Y?” probabilities qualification
Time-series ARIMA/SARIMA, ETS, Prophet- “What will future performance look like if Future values Forecast daily orders with
forecasting style, LSTM nothing changes?” over time seasonality
“When is an event likely to occur (and for Predict time until subscription
Survival analysis Cox model, survival curves Time-to-event
whom)?” cancellation
Propensity Purchase/upsell/click “Who is most likely to respond if we do Likelihood of
Who will buy within 7 days?
modeling propensity nothing special?” action
Uplift / treatment “Who will change behavior because of an Incremental Who buys because of a
Uplift models, causal forests
effect intervention?” impact coupon?
Unsupervised “What hidden groups exist that should be Cluster customers, then
k-means, hierarchical, DBSCAN Group patterns
segmentation treated differently?” predict churn by cluster
Advanced anomaly Isolation Forest, One-Class “Which events are suspicious/unexpected Unusual Detect fraud-like transaction
detection SVM, autoencoders and need investigation?” behavior patterns
 • Aims to determine the best possible decision
• Also called decision analytics or normative
analytics
• Uses both descriptive and predictive to create
the alternatives, and then determines the

Prescriptive
best one
• Enablers

Analytics • Optimization
• Simulation
• Multi-Criteria Decision Modeling
• Heuristic Programming
• The outcomes (i.e., decisions) can be
provided to decision maker to use or
implemented automatically
 Prescriptive analytics techniques
What it
Category Techniques / Methods Decision question Example use-case
recommends

Best Allocate ad budget to maximize profit with

Optimization LP, IP/MIP, nonlinear optimization “What is the best plan given constraints?”
allocation/plan CAC + inventory constraints

Stochastic / robust Robust optimization, chance “What plan is best given uncertainty and risk Risk-aware best
Inventory ordering with uncertain demand
optimization constraints limits?” plan

Monte Carlo, discrete-event “What outcomes are likely under different Outcome Simulate warehouse throughput to choose
Simulation
simulation choices?” distributions staffing

What-if / scenario Sensitivity analysis, driver-based Scenario Profit impact if price +5% and conversion
“If we change X, what happens to Y?”
planning planning comparisons −1%

Decision trees, expected value, “Which option has the best expected (risk- Best choice among Launch now vs delay based on risk-adjusted
Decision analysis
risk/utility adjusted) payoff?” options payoff

Reinforcement “What action should we take each time to

Bandits, RL policy learning Adaptive policy Dynamic pricing / personalized offers
learning maximize long-term value?”

Recommendation Collaborative filtering, learning-to- “What is the next-best offer/item for this Next-best Recommend products optimizing conversion
systems rank, contextual bandits user now?” action/item + margin

Vehicle routing, assignment, “How do we schedule/route resources to Optimal Optimize delivery routes to cut time and fuel
Scheduling & routing
workforce scheduling minimize cost/time?” schedules/routes cost
Analytics
Examples in
Selected
Domains
Example of Analytics
Applications in a Retail
Value Chain
Analytic
Application Business Question Business Value
Inventory 1. Which products have high 1. Forecast the consumption of fast-

Analytics
Optimization demand? moving products and order them with
2. Which products are slow moving sufficient inventory
or becoming obsolete? to avoid a stock-out scenario.
2. Perform fast inventory turnover of
slow-moving products by combining them
with one in high demand. Examples in
Price Elasticity 1. How much net margin do I have
on the product?
2. How much discount can I give
1. Markdown prices for each product can
be optimized to reduce the margin dollar
loss.
Retail Value
Chain
on this product? 2. Optimized price for the bundle of
products is identified to save the margin
dollar.

Market Basket 1. What products should I combine 1. The affinity analysis identifies the Table 1.1 Examples of
Analysis to create a bundle offer? hidden correlations between the Analytics Applications in the
2. Should I combine products products, which can help in following
based on slow-moving and fast- values a) Strategize the product bundle
Retail Value Chain
moving characteristics? offering based on focus on inventory or
3. Should I create a bundle from margin. b) Increase cross-sell or up-sell
the same category or different by creating bundle from different
category line? categories or the same categories,
respectively.
Homework
• Choose a real-business and one decision
• Identify the following
• Goal: What do you want to achieve?
• Options: What options are available?
• Criteria/Metrics: What criteria are used to make the
choice?
• Constraints (budget, time, legal, etc.)
• Create decision making process - Simon (4 stages)
• Complete the flow: Decision -> Information -> Data

Each group bring 1 A0 + colored stick notes in the next class section