Consumer Product Optimization Framework

Core Philosophy: The Power Law of Growth

Most founders fail by trying to improve averages. Consumer products—especially those driven by performance marketing—do not follow a normal distribution; they follow a Power Law:

• 90% of your growth will come from 10% of your creatives
• 90% of your revenue will come from 10% of your users
• 90% of your conversion lift will come from 10% of your funnel steps

Stop optimizing the median. Find the 10x signal, double down, and cut the rest ruthlessly.

Part 1: Dimension Locking (The Variance Killer)

Before you spend $1 on ads, you must constrain your degrees of freedom. Every unlocked dimension creates exponential noise that masks your signal.

The Rule of Constraints

Limit your initial testing to a "Single Profitable Vector."

The Math of Complexity

You don't have the traffic, time, or team to optimize 120 things. Unlocked dimensions create noise that masks signal.

"If you run a bar, you see who walks in. If three burly woodsmen walk in, you don't offer them espresso martinis. In digital marketing, you are blind. Locking dimensions is how you force the 'woodsmen' (your target audience) to reveal themselves so you can see what they actually want."

Part 1B: Data Confidence Scoring (G/Y/R)

Before making any decision, score each data source. Entertainment products follow power law distributions—a small segment of users drives most retention, most revenue, most engagement. But you can't find the power law segments if your data is lying to you.

The Framework

Data Confidence Map Template

For each data source in the company's stack:

Real-World Example

Part 1C: Two-Way Verification (Non-Negotiable)

The Process

Diagnosis Framework (When Delta > 1%)

Real-World Example

Part 1D: Clean Cohort Start Date

Early tracking is always messy. Teams waste weeks arguing "but the old data shows..." Stop this by declaring an explicit boundary.

The Framework

How to Pick the Date

Choose the LATEST of:

1. First app version after event schema stabilized
2. Week after attribution + analytics fully configured
3. Date when known identity bugs were fixed
4. Date when key events started firing consistently

Part 1E: Identity Stability Check

Mobile analytics often have multiple user identifiers that drift apart. This causes inflated unique user counts, broken funnel joins, and inconsistent cohort definitions across reports.

The Check

Common Root Causes

Real-World Example

Part 1F: Environment Filter Validation

Mobile analytics often have TWO env indicators that behave differently:

The Check

Real-World Example

Part 1G: Power Law Segment Identification

Once data is trustworthy (G-tagged), find the power law. In entertainment products:

The Analysis

For each behavioral segment, compute retention lift:

Real-World Example

The power law: "Both" users are the high-retention population. Maximize flow into this segment before optimizing anything else.

Key Questions

1. Is this causal or selection? Does the behavior create retention, or do high-intent users just do this naturally?
2. What's the intervention? How do we move users from low-value to high-value segments?
3. What's the ceiling? What % of installs can realistically reach the high-value segment?

Part 1H: Intent Signal Identification

Not all early behaviors predict retention equally. Some are intent signals—they reveal pre-existing user motivation rather than creating it. Finding these signals lets you:

• Route high-intent users to faster conversion paths
• Avoid wasting effort trying to "convert" low-intent users
• Accurately attribute outcomes to causes vs selection

Intent vs Engagement Signals

Analysis Template

Real-World Example

Tactical Implications

Part 1I: Causal Model Testing Framework

You found a power law segment: users who do X retain at 2x the rate. But is it causal or selection?

The Testing Framework

Step 1: Check for intent signals

If users who do X also show strong intent signals (push permission, fast time-to-action, etc.), selection is likely.

Step 2: Forced path A/B

Step 3: Prompt holdout

Ship the intervention. Hold out 10%. Compare outcomes for users who:

• Were prompted and did X
• Were prompted and didn't do X
• Weren't prompted but did X naturally

Part 1J: Paywall Timing Analysis

The Framework

Analysis Checklist

1. Segment paywall views by context — What was the user doing when they saw it?
2. Compute CVR by context — Which contexts have highest conversion?
3. Check coverage — Are high-intent users reaching high-CVR contexts?
4. Identify leaks — Are users dropping off before reaching blocked moments?

Real-World Example

Common Patterns

Part 1K: Checkout Funnel Decomposition

Real-World Example

Analysis Checklist

1. Add checkout_start event if not present
2. Compute both conversion rates separately
3. Prioritize the bigger leak — Fix 10%→15% (1.5x) before optimizing 93%→95%
4. Segment by context — Where are checkout starters coming from?

Common Fixes by Leak Location

Part 1L: Platform Economics Segmentation

Real-World Example

Analysis Template

Platform Strategy Matrix

Part 2: Statistical Discipline & The Velocity Equation

Time is an Output, Not an Input

The Code Velocity Equation

Your measurement cadence must match your team's ability to ship:

1. If you ship daily: Measure in 48-hour windows
2. If signal is 3x better than baseline: Cut the test early—you don't need 2,000 installs to prove that a forest fire is hot
3. If time to significance > 14 days: Your optimization target is too small or your spend is too low—move to a higher-leverage part of the funnel

Part 3: Creative as Targeting (The New Meta)

In the era of automated ad platforms (Advantage+, App-level bidding), the creative is the targeting. The algorithm looks at who engages with your ad and finds more people like them.

The Outlier Strategy

1. Launch 5-10 "Polarizing" Creatives: Don't test small variations (e.g., button color). Test radically different hooks
2. Kill the Losers Fast: If an ad has 0 conversions and low CTR after 2x your target CPA spend, kill it
3. The 80/20 Creative Rule: Once you find a "Winner" (high CTR + high conversion), stop testing new concepts. Spend 80% of your time iterating on that winner to extend its life

"If you can turn on even $1,000 of profitable spend today that generates $2,000, it has real compounding effects. Because we're so early in the timeline, the deeper we get, the bigger improvements need to be to meaningfully move the needle."

Early optimization compounds. Late optimization is marginal.

Part 4: End-to-End Funnel Alignment

The Ad is the Product

The biggest point of failure in consumer apps is a "Broken Chain." Marketing buys an audience, and Product shows them something generic.

Siloed Teams Break This Chain

• Marketing optimizes CTR in isolation
• Product optimizes engagement in isolation
• Neither knows who the other is sending/receiving

Solution: Cross-functional visibility. Your growth/marketing lead needs to be in the room with acquisition, seeing raw data, constraining execution.

Part 5: The "Duolingo Rule" (Best Practice > Innovation)

Don't Reinvent the Wheel

Unless your core innovation is a new payment model, do not A/B test your paywall or onboarding flow against your "gut."

"If Duolingo, Calm, or Strava does it, 100 people have already spent $10M testing it. Their 'standard' is your 'baseline.'"

The "Standard" Stack

• Paywall: Annual vs. Monthly toggle, clear "Try for Free" button, social proof/testimonials
• Onboarding: 5-7 questions to build "sunk cost" and personalization
• Subscription: Use RevenueCat or Glassfy—do not build your own receipt validation logic

When to Deviate from Best Practice

Only after you've:

1. Implemented best practice
2. Measured baseline performance
3. Identified specific hypothesis for improvement
4. Have statistical power to test deviation (1,000+ conversions/mo)

Part 6: Set-It-and-Forget-It (Locking Profitable Vectors)

When you find a combination (Ad A + Onboarding B + US iOS) that is ROAS positive: Lock it.

The Lock Protocol

1. Stop tweaking the UI
2. Stop changing the price
3. Document the winning configuration (audience, creative, funnel)
4. Set a "Scale Trigger": If ROAS > X for 7 days, increase spend by 20%
5. Move the Team: Take your optimization energy and find the next vector

Even if it's only $100/day profitable → that's $100/day extending runway while you find the next win.

Part 7: The Prioritization Matrix

Stop working on things that don't move the needle.

When to Use "Best Practice, Don't Test"

For anything where:

• Industry has established standard
• Your sample size won't reach significance for months
• Improvement ceiling is modest

→ Just implement best practice and move on. Test higher-leverage items.

Part 8: Practical Application Checklists

Part 9: Common Founder Traps (The "Red Flags")

Part 10: Dimension Unlock Triggers

Dimension locking is for resource conservation, not permanent strategy. Build explicit unlock criteria:

Summary: The Optimization Loop

Framework Version 2.0 | Transcend Official Guide | January 2026