Market Segmentation & Customer Targeting Questions

Approach overview: I’d build an RFM table from event and payment logs, score customers on each axis, cluster/bucket into segments, then prioritize outreach by value and churn risk.

Compute R, F, M:

• Recency (R): days since last meaningful event (purchase or active session) as of reference date.
• Frequency (F): count of purchases or qualifying events in the last 12 months (or relevant window).
• Monetary (M): sum of payments (revenue) over the same window, adjusted for refunds and lifetime value if available.

Scoring/bucketing:

• Start with quintile-based scores (1–5) for interpretability: higher R score = more recent, higher F/M = more frequent/valuable.
• Validate with k-means (k=4–6) on normalized R,F,M to detect non-linear groupings; prefer k-means if distributions are heavily skewed or have natural clusters.
• Combine: use quintile scores for dashboards and k-means clusters to refine anomalies.

Naming & prioritization:

• Create human-friendly labels (e.g., Champions: R5F5M5; At-Risk High-Value: R1-2, F4-5, M4-5; New Loyalists; Dormant Low-Value).
• Prioritize: At-Risk High-Value > Champions (retain) > New Loyalists (engage) > Dormant High-Potential > Dormant Low-Value.

Actionable playbook (At-Risk High-Value — recent drop in recency but historically high F/M):

• Trigger: R score drops from 4–5 to 1–2 in last 30 days.
• Message: Personalized email subject “We miss you, [Name] — here’s 20% off your favorite [category]” plus dynamic content showing top past purchases and social proof.
• Channel & cadence: Email + SMS for 48–72 hour follow-up; in-app push if active on app. Sequence: immediate personalized offer → reminder 3 days later → final “last chance” 7 days later.
• Measurement: A/B test offer vs. no-offer; track reactivation rate, incremental revenue, and cost per recovered customer. Iterate thresholds and messaging based on lift.