11 active segments + 3 hibernating/lost. MECE: every customer belongs to exactly one. Sorted by historical revenue.
| Segment | Customers | Cust % | All-time Rev | Rev % | Avg Orders | Avg Rev/Cust | P(alive) | Avg 1-yr LTV | Avg 3-yr LTV | Avg 5-yr LTV |
|---|
LTV is a per-customer metric. "Per customer (avg)" shows average forward LTV for customers in each segment — a measure of segment quality. "Fleet total" shows the aggregate forward revenue from each segment (LTV × customer count) — a measure of where the dollars are. BG/NBD model for expected purchase count × Gamma-Gamma model for value per transaction. Longer horizons assume stable purchase rate; uncertainty widens at 3 and 5 years.
% of each acquisition cohort that placed a repeat order N months later. Spikes at month 6 and 12 indicate anniversary / seasonal repurchase behavior.
SKU-merged, categorized by product type. Same SKU appearing under multiple names is consolidated.
| Category | SKUs | Revenue | Rev % | Units | Orders |
|---|
"Repeat %" = of customers who bought this SKU, the % who bought it 2+ times. "Variants" = how many different product-name spellings exist on this SKU (a count of 4+ means the SKU was renamed multiple times in Shopify).
| # | SKU | Product | Category | Revenue | Units | Orders | Customers | Repeat % | Variants |
|---|
Reads as: "When an order contains a product in row category, X% of those orders also contain a product in the column category." Hover for the lift score — lift > 1 means the pair is bought together more often than chance; lift < 1 means avoided. Diagonal = % of orders with that category that contain 2+ SKUs from that same category.
For each of the top 10 SKUs by revenue, the products most frequently bought in the same order. Attach % = of orders containing the anchor, what % also contain the partner. Baseline % = how often the partner appears in any order. Lift = Attach / Baseline.
| SKU | Product | Category | Co-orders | Attach % | Baseline % | Lift |
|---|
Pairs of high-volume SKUs (both sides with at least 300 orders) ranked by lift. High-lift pairs are products that go together — natural bundle candidates and cross-sell targets. Includes both same-category pairs (e.g. two shades a customer commonly pairs) and cross-category pairs.
| Product A | Cat A | Product B | Cat B | Co-orders | A→B % | B→A % | Lift |
|---|
| Code | Orders |
|---|
| Score | Recency (days since last post-mig. order) | Frequency (total orders all-time) | Monetary (total revenue all-time) |
|---|---|---|---|
| 5 | 0–30 | 10+ | Top 10% (≥$384) |
| 4 | 31–90 | 6–9 | 75–90% ($156–$384) |
| 3 | 91–180 | 3–5 | 50–75% ($67–$156) |
| 2 | 181–365 | 2 | 25–50% ($36–$67) |
| 1 | 366+ or no post-mig. order | 1 | Bottom 25% (<$36) |
Thresholds chosen for cosmetics repurchase cadence (faster than luxury goods). Monetary bands are percentile-based on this customer file.
| Segment | Definition | Strategy |
|---|---|---|
| Champions | R≥4, F≥4, M≥4 (active) | Reward, retain, VIP perks |
| Loyal Customers | R≥4, F≥3 (active) | Cross-sell, ambassador program |
| Cannot Lose Them | R=3, F≥3, M≥4 (sliding high-value) | Urgent personalized re-engagement |
| At Risk | R=2, F≥3 (lapsing repeats) | Win-back with strong incentive |
| Hibernating Heroes | R=1 OR pre-mig-only AND F≥4, M≥4 | Premium reactivation campaign |
| Hibernating Regulars | R=1 OR pre-mig-only AND F≥3 | Standard win-back |
| Needs Attention | R=3, F≥2 (modest sliding) | Nurture email, gentle nudge |
| Promising | R≥4, F=2 (newer repeat buyers) | Build the habit, second-product offer |
| Recent First-Timers | R=4, F=1 (bought 31–90d ago) | Second-purchase flow |
| New Customers | R=5, F=1 (bought ≤30d ago) | Onboarding + welcome flow |
| About to Sleep | R=3 F=1, or R=2 F≤2 (cooling off) | Re-engagement before they lapse |
| Lost | R=1 / pre-mig-only AND F≤2 | Suppress or one-shot win-back |
BG/NBD (Beta-Geometric/Negative Binomial Distribution) predicts the expected number of repeat purchases in the next 12 months, and the probability each customer is still "alive" (vs. churned). Gamma-Gamma model predicts the expected monetary value per future transaction, independent of frequency.
Model fit: BG/NBD r=0.4625, α=216.1, a=2.19, b=11.2 · Gamma-Gamma p=10.13, q=2.36, v=9.57 · Correlation between frequency and monetary value = 0.033 (model independence assumption satisfied).
Training data: Post-migration transactions only (on/after 2024-11-01) — 18,411 orders across 11,720 customers, 18 months of history. Pre-migration customers (no post-cutoff orders) receive a forward LTV of $0 in the rollup, since the model cannot estimate dormancy length from synthetic migration dates.