Solve Dataset_Metadata_and_Business_Rules questions in the dabstep payment processing dataset. Use this skill for questions about dataset column meanings, fee factor directionality, business rules encoded in fee structures, "Not Applicable" detection, and metadata interpretation for the dabstep payment processing domain. Trigger whenever questions reference payment fees, fraud indicators, boolean fee factors, volume tiers, ACI codes, or any field from the dabstep dataset.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
A direct command skips the review prompt. Inspect the source before running it.
Solve Dataset_Metadata_and_Business_Rules questions in the dabstep payment processing dataset. Use this skill for questions about dataset column meanings, fee factor directionality, business rules encoded in fee structures, "Not Applicable" detection, and metadata interpretation for the dabstep payment processing domain. Trigger whenever questions reference payment fees, fraud indicators, boolean fee factors, volume tiers, ACI codes, or any field from the dabstep dataset.
Dabstep: Dataset Metadata & Business Rules
Dataset Overview
Seven files describe a payment processing system:
File
Contents
payments.csv
Transaction records (fraud flag, amounts, card types, countries)
payments-readme.md
Column-by-column documentation for payments.csv
fees.json
1000 fee rules with conditions and rates
manual.md
Domain definitions, fee factor descriptions, business rules
"What column indicates X?" / "What does field Y mean?"
Check payments-readme.md first (columns), then manual.md (definitions).
Example: fraud column → has_fraudulent_dispute (from payments-readme.md: "Indicator of fraudulent dispute from issuing bank (Boolean)")
2. Boolean Fee Factor Questions
"Which boolean factor is cheaper when True/False?"
Rely on manual.md explicit statements, not raw data averages. Raw averages are confounded by other correlated variables (card scheme, MCC, etc.) and can be misleading.
Boolean factors in fees.json: is_credit, intracountry
Cheaper when True: intracountry (domestic transactions are cheaper)
Cheaper when False: is_credit (non-credit/debit is cheaper)
3. General Factor Directionality
"Which factors lead to cheaper fees when decreased/increased?"
Decreasing → more expensive: capture_delay (faster = more expensive), monthly_volume, intracountry
4. Volume/Threshold Analysis
"At what volume do fees stop decreasing?" / "Highest volume where fees are not cheaper?"
Analyze actual fee data from fees.json:
import json, pandas as pd
withopen('fees.json') as f:
fees = json.load(f)
df = pd.DataFrame(fees)
for vol in ['<100k', '100k-1m', '1m-5m', '>5m']:
s = df[df['monthly_volume'] == vol]
print(vol, s['rate'].mean(), s['fixed_amount'].mean())
The actual data shows >5m has the highest average rate — fees do not become cheaper at >5m. The >5m tier is the highest volume at which fees are not cheaper.
5. "Not Applicable" Detection
"Does X fee exist?" / "What is the value of Y?"
A question should be answered "Not Applicable" when the concept genuinely does not exist in any dataset file. Verify by:
Searching manual.md for the concept
Searching all keys in fees.json
Checking payments.csv columns
Searching payments-readme.md
Example: "excessive retry fee" — manual mentions excessive retrying causes downgrades but defines no specific retry fee amount → answer is "Not Applicable"
Common Pitfalls
Don't trust raw averages for boolean factors. The raw average rate for is_credit=True can appear lower than is_credit=False due to confounding (different MCCs and card schemes have both different base rates and different credit ratios). The manual's explicit statement is authoritative.
intracountry in fees.json uses 0.0/1.0 floats, not Python booleans. 1.0 = True (domestic), 0.0 = False (international).
null in fees.json means "applies to all values" — do not treat null rules as lower-priority; they apply when no more-specific rule matches.
Read the full manual.md — it has 10+ sections. The output may be truncated; if so, use grep or targeted searches to find specific definitions.
Answer Format Rules
Column names: exact string as in payments-readme.md (e.g., has_fraudulent_dispute)
Multiple factors: comma-separated, alphabetical or as listed (e.g., monthly_fraud_level, is_credit)
Volume tiers: exact strings from data (e.g., >5m, 1m-5m)
Fraud levels: exact strings from data (e.g., <7.2%, >8.3%)
