Module 6

Prioritization & Tradeoffs

📖 4 Lessons⏱ 60-90 min 🧪 Lab: RICE Scoring with AI Scenario Modeling✅ Quiz: 7 questions

Learning Objectives

Apply RICE and other prioritization frameworks
Use AI for scenario modeling and tradeoff analysis
Calibrate confidence scores with evidence
Evaluate when AI recommendations should be overridden

Lesson 6.1

Prioritization Is About Saying No

A PM's job is not to say yes to everything. It's to choose the highest-leverage opportunities, manage constraints explicitly, and make tradeoffs visible and defensible.

Cognitive Biases That Distort Prioritization

Bias	How It Manifests	Countermeasure
Anchoring	First number discussed becomes the reference for all estimates	Get independent estimates before group discussion
Sunk cost	"We've already spent 3 sprints on this, we can't stop now"	Evaluate on future value, not past investment
Availability bias	Overweight recent, vivid, or emotionally charged inputs	Force systematic review of all data
Completion bias	Prefer finishing something over starting higher-value work	Ask: "If this weren't started, would I prioritise it today?"
HiPPO effect	Highest Paid Person's Opinion dominates	Use RICE to make criteria explicit before opinions
Optimism bias	Underestimate effort, overestimate impact	Use historical data for calibration
Survivorship bias	Focus on successful features, ignore killed ones	Review past initiative outcomes before estimating new ones

⚠️

AI-specific bias: AI models amplify anchoring and availability bias because they reflect your framing. If you describe an initiative enthusiastically, the AI will score it higher. Always present initiatives neutrally when scoring.

Lesson 6.2

RICE Framework

Formula

RICE = (Reach × Impact × Confidence) / Effort

Factor	Question	Scoring
Reach	How many users per quarter?	Number (count)
Impact	How much does each user benefit?	3=massive, 2=high, 1=medium, 0.5=low, 0.25=minimal
Confidence	How certain are we of these estimates?	100%=high, 80%=medium, 50%=low
Effort	How many person-months?	Number (person-months)

Lesson 6.3

Effort Estimation: Getting the Denominator Right

Effort is the most error-prone variable. Humans consistently underestimate effort by 2-3x. Three estimation techniques:

Technique 1 — Analogy

Compare to something your team has already built. "RBAC is similar to the team permissions we built in Q1. That took 12 person-months. RBAC adds 4 more roles + audit logging = ~1.5x. Estimate: 18 person-months." Anchors to real data instead of wishful thinking.

Technique 2 — Decomposition

Break into components and estimate each: UI flow (4pw) + progress tracking (2pw) + contextual help (3pw) + data source simplification (6pw) + QA (2pw) + 20% buffer (3.4pw) = ~20pw total. Forces you to think through full scope.

Technique 3 — Three-Point

Best (3 months) + Most Likely (5 months) + Worst (9 months). Expected = (3 + 4×5 + 9) / 6 = 5.3 months. Makes uncertainty explicit.

🔴

Never use AI effort estimates for commitment. AI does not know your team's velocity, technical debt, or skill distribution. Always calibrate against a similar completed project and add 20-50% buffer.

Lesson 6.4

Confidence Calibration

Confidence Level	What It Should Mean	Required Evidence
100% (High)	We have direct data supporting all four RICE factors	User research + behavioral data + tech feasibility confirmed
80% (Medium)	Data for some factors, educated estimates for others	Some user research + proxy data + engineering has reviewed scope
50% (Low)	Guessing based on analogies or assumptions	No direct data; extrapolating from other products

💡

Confidence Threshold Rule: If an initiative scores high on RICE but has 50% confidence on 2+ factors, do not commit to building it. Commit to a 1-2 week research spike that reduces uncertainty. Re-score after with improved evidence.

Lab 6

Prioritization with AI Scenario Modeling

Scenario: FlowScale has 8 initiatives proposed for Q3. You have capacity for 3. Prioritize using RICE with AI scenario modeling.

8 Initiatives for Q3

1. Guided Onboarding Wizard (from Module 5 PRD)
2. Role-Based Access Controls (RBAC) — enterprise permissions
3. Industry-Specific Templates — 5 vertical template packs
4. AI Workflow Suggestions — suggest next step based on usage
5. Bulk User Import — CSV upload for enterprise onboarding
6. Public API Status Page — transparency for merchants
7. Webhook Retry Logic — automatic retry for failed deliveries
8. Workflow Sharing — share workflows across team members

Step 1

Create prompts/rice-scorer.md. The prompt must require: explicit math for Reach, explanation for Impact score, evidence justification for Confidence level (do not give all initiatives the same confidence), and decomposition for Effort. Output: ranked table + top 3 for Q3 at 4 engineer-months capacity + what would need to change for excluded initiatives.

Step 2 and 3

Run RICE scoring. Then identify:

1 initiative where you would override the RICE score — explain why
1 assumption the AI made that you believe is wrong
1 second-order effect the AI missed

Step 4

Write deliverables/W06-prioritization.md with: AI ranking, your adjustments with reasoning, final Q3 plan (3 initiatives), "What we are not doing" section, and "What would change this" section.

Deliverables

RICE scorer prompt
Raw AI output (scores and ranking)
Override analysis (3 specific items)
Prioritization memo

How to Verify Completion

All 8 initiatives scored with explicit math visible for each RICE component
Confidence levels differ across initiatives — not all scored 80%
Your override analysis cites specific evidence, not just preference
"What we are not doing" section explicitly names at least 4 excluded initiatives with reasoning

Done when: Your CEO could read the prioritization memo, understand why you picked the 3 initiatives you did, and what evidence would cause a re-prioritization.

Quiz

Module 6 — Knowledge Check

0 / 7 answered

Question 1 of 7

In RICE scoring, what does Confidence represent?

A How confident the AI model is in its estimate

B How certain you are in the quality of your Reach, Impact, and Effort estimates based on your evidence

C How confident stakeholders are that the initiative should be built

D The probability that engineering will complete the initiative on time

Correct. Confidence reflects the quality of your evidence for the other three factors. 100% = direct data for all. 80% = some data, some estimates. 50% = mostly guessing. High AI confidence in its output does not equal your 100% — it reflects your evidence quality.

Question 2 of 7

What is the HiPPO effect and how does RICE mitigate it?

A HiPPO refers to large, slow-moving projects; RICE speeds them up

B HiPPO is a memory management protocol; RICE optimizes it

C HiPPO = Highest Paid Person's Opinion dominating decisions; RICE makes scoring criteria explicit before opinions are shared, reducing authority bias

D HiPPO is a framework for enterprise features; RICE is for consumer products

Correct. When you define and apply RICE criteria before the prioritization discussion, it is harder for seniority alone to override data. The CEO still has a voice, but that voice now needs to address specific numbers rather than just assert preference.

Question 3 of 7

Initiative A scores RICE = 120. Initiative B scores RICE = 85, but B has 100% confidence while A has 50% confidence on 3 factors. What should you do?

A Build Initiative A first — higher RICE score wins

B Build both in parallel to save time

C Do not commit to Initiative A — run a research spike to reduce uncertainty first; Initiative B with solid evidence may be the safer bet

D Reject Initiative B — lower RICE score means it's not worth pursuing

Correct. The Confidence Threshold Rule says: if an initiative has 50% confidence on 2+ factors, do not commit. Initiative A's high score is built on shaky evidence. A 1-2 week research spike could either confirm the score or reveal it should be much lower.

Question 4 of 7

Why should you never use AI effort estimates for delivery commitments?

A Because AI always overestimates effort to seem impressive

B Because effort estimates require executive approval

C Because AI estimates in person-hours, not story points

D Because AI does not know your team's velocity, technical debt, skill distribution, or current sprint commitments

Correct. AI effort estimates assume ideal conditions with a generic team. Your team has specific strengths, existing technical debt, ongoing commitments, and quirks that only you and your engineering lead know. Use AI estimates for relative comparison only, never as commitments.

Question 5 of 7

The Guided Onboarding Wizard reaches 1,200 users/quarter with Impact=2, Confidence=80%, Effort=5 person-months. What is its RICE score?

A 192

B 384

C 480

D 300

Correct. RICE = (1200 × 2 × 0.80) / 5 = 1920 / 5 = 384. Always show your working when presenting RICE scores to stakeholders so they can challenge the assumptions, not just the final number.

Question 6 of 7

A PM says: "We've already spent 3 sprints on the AI Workflow Suggestions feature. We have to finish it." Which cognitive bias does this demonstrate?

A Anchoring bias — the 3-sprint number is dominating the decision

B Sunk cost fallacy — past investment does not justify future investment if the initiative no longer delivers value

C Completion bias — the PM wants to finish things before starting new work

D Availability bias — the recent 3 sprints are top of mind

Correct. Sunk cost fallacy. Past investment that cannot be recovered should not influence future decisions. The correct question is: "If we were evaluating this initiative fresh today with no prior investment, would we prioritise it over the alternatives?"

Question 7 of 7

What does a "What would change this" section in a prioritization memo accomplish?

A It shows the PM has considered all possible future scenarios

B It is a legal disclaimer protecting the PM from blame if priorities shift

C It makes the decision falsifiable — specifying what evidence would cause you to reprioritize, which builds credibility and demonstrates systematic thinking

D It gives executives a list of demands to prevent the decision from changing

Correct. A falsifiable decision is a strong decision. "We'll maintain this priority unless RBAC demand from enterprise accounts exceeds 3 churned accounts in Q3" is specific, measurable, and shows you're thinking rigorously. It also prevents re-litigating decisions based on vague feelings.

Module 6 Score

←

Module 5 — PRDs

→

Module 7 — Roadmapping