Explain the campaign decision with evidence

Return to the re-engagement question and explain both the groups to activate and the work to hold.

Question for this chapter

Which customers should you re-engage, and what should not happen yet?

Complete the solution

Initial problem	Applied solution	Evidence left behind	What changed
Source orders contained missing values.	Removed seven missing amounts and five missing customer IDs, then replaced two unknown products.	88 clean rows, 2 `PRD-UNKNOWN` rows	Valid purchases and unknown products are distinct.
Customer state was too vague.	Classified with recency, frequency, and monetary questions.	22 customer rows	Each label has an explainable rule.
All four groups were assumed to exist.	Counted the actual 7, 11, 4, and 0 distribution.	Zero `at_risk` customers	The empty campaign is on hold.
Product-level personalization was desired.	Materialized the purchase graph and isolated unknown products in one node.	22 customers, 88 orders, 9 products	Only known products become relationship evidence.

State the final decision

JaffleMart will prepare recognition and retention for seven champion customers, repeat-purchase suggestions for eleven loyal customers, and second-purchase onboarding for four new customers. It will not send an at_risk campaign because that group is empty. Known-product purchase paths can support personalization, while PRD-UNKNOWN relations remain excluded until resolved.

Completion check

Reduced 100 source orders to 88 cleaned rows.
Can explain when to remove a missing value and when to replace it with PRD-UNKNOWN.
Can explain recency, frequency, and monetary value in plain language.
Confirmed the 7, 11, 4, and 0 segment distribution.
Drafted actions for the three groups that exist.
Confirmed seven purchase paths for CUST-017.
Excluded unknown-product relations from product personalization.

Apply next

A real campaign also needs customer consent, reachable channels, recent-contact frequency, margin, and experiment groups. When that data is unavailable, record it as a prerequisite for the next decision instead of filling the gap with an estimate.