Key Takeaways: Chapter 31 — Marketing Analytics and Campaign Analysis

The Core Mental Model

Marketing analytics is not about measuring everything. It is about measuring the right things — and then connecting those measurements into a coherent story about whether your marketing is working, why, and what to do differently.

Every marketing decision is ultimately a capital allocation decision: given a fixed budget, which investments produce the best returns? Python gives you the tools to answer that question with data rather than instinct.

Metrics That Matter

Customer Acquisition Cost (CAC) — Total marketing and sales spend divided by new customers acquired. Only meaningful in context of what those customers are worth.

Customer Lifetime Value (LTV) — The total revenue expected from a customer relationship: Average Order Value × Purchase Frequency × Customer Lifespan. Use gross-margin-adjusted LTV for investment decisions.

LTV:CAC ratio — The fundamental health metric for customer acquisition. Below 1:1 is existential. 3:1 to 5:1 is healthy for growth-stage businesses. Above 5:1 suggests underinvestment in growth.

ROAS — Revenue generated per dollar of ad spend. Meaningful only relative to your gross margin. Break-even ROAS = 1 / Gross Margin %.

CTR — Meaningful only when compared against channel-specific benchmarks. A 2% CTR means very different things in search versus display versus email.

Conversion rate — Always specify the denominator. "Conversion rate" from a landing page, from an email click, and from a free trial are three different measurements.

Attribution

• First-touch answers "how do customers discover us?" Gives full credit to the first interaction.
• Last-touch answers "what closes deals?" Gives full credit to the final interaction before conversion.
• Multi-touch attempts to distribute credit fairly across the whole journey, but requires more data infrastructure and involves modeling assumptions.
• No attribution model is objectively correct. The best approach is to use multiple models and look for channels that perform well across all of them.

A/B Testing Principles

1. Determine sample size before you start. Use calculate_required_sample_size(). Commit to running until you hit it.
2. Change one variable at a time. Multiple simultaneous changes produce uninterpretable results.
3. Never peek. Looking at results before reaching target sample size inflates false positive rates.
4. Statistical significance ≠ business significance. A 0.01% improvement may be real and still not worth implementing.
5. Multiple testing requires correction. Running N tests simultaneously multiplies your expected false positive count by N. Apply Bonferroni correction or similar.
6. Watch for novelty effects. Run tests long enough to see through the initial curiosity bump.
7. Use chi-square for conversion rates; t-test for continuous metrics like average order value.

Funnel Analysis

• A raw conversion rate hides the structure of the problem. Break the funnel into stages.
• Focus optimization effort where absolute drop-off is largest, not where the rate looks worst. A 20% drop-off rate at a high-volume stage is more impactful to fix than a 40% drop-off at a low-volume stage.
• Always segment your funnel. Mobile vs. desktop, new vs. returning, channel by channel. Aggregate funnels hide the most actionable patterns.
• Revenue impact modeling lets you prioritize which funnel stage improvements to invest in first.

Channel Analysis

• CAC without LTV is meaningless. A high CAC on a high-LTV channel is fine. A low CAC on a low-LTV channel can still be a bad investment.
• LTV:CAC by channel is the key table. Build it. Update it quarterly.
• Efficiency ratio (revenue share / spend share) quickly identifies which channels punch above their weight.
• UTM parameters are foundational. Without them, you cannot attribute conversions to channels reliably. Tag everything.

Budget Optimization

• All marketing channels exhibit diminishing returns at sufficient scale.
• Budget allocation should be based on marginal return, not average return.
• Simple diminishing-returns models (power functions) are sufficient for most business contexts.
• The practical decision rule: systematically shift budget from channels with efficiency ratios below 1.0 toward channels above 1.0, in proportion to the efficiency gap.

The Most Common Mistakes

Key Python Tools

# Sample size calculation
from ab_test_analyzer import calculate_required_sample_size

# A/B test analysis
from ab_test_analyzer import analyze_conversion_ab_test, print_stakeholder_report

# Funnel analysis
from funnel_analyzer import FunnelAnalyzer

# Channel comparison
pandas.groupby().agg()   # the workhorse for all channel metrics
scipy.stats.chi2_contingency()   # conversion rate significance testing
scipy.stats.ttest_ind()          # continuous metric significance testing

# Budget optimization
from scipy.optimize import minimize   # constrained optimization

One Sentence Each

• CAC: What you paid to get one customer.
• LTV: What that customer is worth over their lifetime.
• ROAS: Are your ads generating more revenue than they cost?
• CTR: Are people engaging with what you put in front of them?
• A/B test: The only way to actually know whether a change helped.
• Funnel analysis: Find the leak before you pump in more water.
• Attribution: Decide which channel gets credit — knowing no model is perfect.