On-Site Personalisation & Product Recommendations: Segment Rules, Click-Through, Attributed Revenue, Holdout Control Groups, Placement Tests & Catalogue Coverage
Measure on-site personalisation with spreadsheets instead of hunches. You will turn behaviour into segment rules, measure recommendation click-through and attributed revenue, test attribution windows, compare segments, set up holdout control groups and test whether a lift is real, separate attributed from incremental revenue, compare widget placements, read position bias, check catalogue coverage, diversity and popularity bias, handle cold starts, match widgets to segments, set frequency caps, rank with relevance and margin, forecast the rollout, check tracking quality and finish with a personalisation scorecard. Behaviour-based segments only, fictional data of a home-goods shop, 20 hands-on labs, every cell checked as you go.
- One-time payment
- Instant access right after payment
- Secure checkout with Stripe
About this course
In this course you measure on-site personalisation the way a careful analyst would, with spreadsheets you build yourself. You work for Brightloom Home, a fictional online shop for kitchen, bedding, lighting and storage products, and in 20 hands-on labs you turn visitor behaviour into segment rules, calculate recommendation click-through and attributed revenue, compare attribution windows, and build holdout control groups with a test for statistical significance. You separate attributed from incremental revenue, compare widget placements, read position bias, and check catalogue coverage, diversity and popularity bias. Later labs cover cold-start fallbacks, matching widgets to segments, frequency caps, margin-aware ranking, a rollout forecast with payback, sample ratio checks and a final personalisation scorecard. Segments are based on behaviour only. Every formula is checked by the Coach as you type. All companies, visitors and numbers are fictional.
Lessons
- Behaviour Segment Rules: Thresholds, IFS Labels, Segment Sizes and Shares of VisitorsFree
- Recommendation Click-Through: Widget CTR, the Weighted Overall Rate, the Simple-Average Trap and Ranking Locked
- Attributed Revenue: Click-to-Order Rate, Revenue per Click, Revenue per 1,000 Impressions and Share of Site Revenue Locked
- Attribution Windows: Click-to-Order Lag, Window Flags and Orders and Revenue by Window Length Locked
- Segment Performance: SUMIFS Totals, Weighted Segment CTR, Revenue per Session and the AVERAGEIFS Comparison Locked
- Holdout Control Groups: Revenue per Visitor, Incremental Revenue per Week and the Overall Lift Locked
- Is the Lift Real? Conversion Difference, Pooled Standard Error, z-Score, p-Value and Verdict for Five Holdout Tests Locked
- Attributed vs Incremental Revenue: Incremental Share, Net Incremental Revenue, Return on Cost and Red Flags Locked
- Widget Placement Tests: View Rate, CTR on Views, Revenue per Visitor and Lift against the Control Layout Locked
- Position Bias: CTR by Slot, Index against Slot 1, Cumulative Click Share and the Top-Three Share Locked
- Catalogue Coverage: Covered Products, Coverage by Category, Never-Recommended Items and Visual Flags Locked
- Recommendation Diversity: Impression Shares, HHI, Effective Number of Products, Top-Three Share and Category Mix Locked
- Popularity Bias and the Long Tail: Exposure Ratio, Head-Mid-Tail Buckets and Shares of Sales vs Impressions Locked
- Cold Start and Fallback Rules: Data-Depth Bands, the Personalisation Switch, CTR by Mode and the Share on Fallback Locked
- Segment-to-Widget Matching: Best Widget per Segment, Gain vs the Default, Extra Monthly Revenue and a Colour Scale Locked
- Frequency Capping and Fatigue: CTR by Exposure Number, Index against the First Exposure, the Cap and the Wasted Share Locked
- Margin-Aware Ranking: Expected Margin per 1,000 Impressions, Two Rankings and a Blended Score with a Relevance Weight Locked
- Rollout Forecast and Payback: Linear Trend with SLOPE and FORECAST.LINEAR, Net Monthly Contribution and Months to Recover the Setup Cost Locked
- Tracking Quality: Sample Ratio Mismatch with Expected Holdout Size, Chi-Square, p-Value and a Red Flag Locked
- Final Project: Personalisation Scorecard with CTR, Net Incremental Revenue, Return on Cost, Coverage Rules and One Decision per Widget Locked
Learner reviews
Be the first to review this course.