Enrich the order evidence for review

Question for this chapter

Which customer context should be attached before a refund request is judged?

Why this matters now

Amount alone makes every expensive order look risky. Tier alone can hide repeated refunds from a valued customer. Read the order, customer, and history in one row.

Try it

Confirm the source datasets under refund_ops:

• raw_orders — 12 rows
• raw_customers — 10 rows
• refund_history — 10 rows

Then open the order_enrichment pipeline and select Run. The join_order_customer code connects orders and customers on customer_id, then left-joins each customer's refund count and amount.

Success looks like this

enriched_orders contains 12 rows. Inspect these columns together:

order_id, customer_id, customer_tier, total_amount,
past_refund_count, past_refund_total, status

For example, ORD-2025-0001 is a $1,299.99 order from gold customer CUS-001 with two past refunds. That is evidence for the next score, not an approval or rejection.

Interpret the result

Enrichment does not make a decision. It gives reviewers a shared bundle so they begin from the same facts.

Next decision

You have assembled the review evidence. Next, classify the 12 orders by risk while using the result only to order the review queue.