OTIF---Time-to-Ready

📊 Research on the time of order co-ordination (time_to_ready)

🎯 Purpose of the study

To investigate the metric time_to_ready (time to order reconciliation between customer and supplier), identify its patterns, dependencies and impact on key metrics: OTIF, share of overdue orders (late_%) and delivery times.

---

📏 Metric Description

time_to_ready is the time (in days) from order creation to supplier approval.
This is a critical step in the supply chain, affecting further operations: shipment planning, delivery dates, order fulfilment on time.

---

🧭 Approach

• Analysis of time_to_ready distribution
• Clustering of orders by processing time
• Hypothesis testing of relationships with other metrics
• Automatic search for anomalous objects:
  • Distribution Centres
  • Customers
  • Customer Catalogues
  • Legal Entities
• Dependency visualisation
• Predictive model building using CatBoost, interpretation via SHAP

---

🧪 Hypotheses

1. ** As time_to_ready increases, late_% within a cluster grows.** ✅ Tested using χ² test for trends.

2. Median time_to_ready varies for at least one day of the week. ✅ Checked using the Kraskell-Wallis test.

---

🔍 Results

• A statistically significant relationship was found between approval times and the proportion of delinquencies (late_%).
• Identified days of the week where approval times are systematically higher/lower.
• Created automated scripts to find ‘bad’ groups (RAs, customers, etc.) with high time_to_ready.
• Built CatBoostRegressor model to predict time_to_ready, finding key influencers via SHAP values.
• Found the time_to_ready significance threshold at which the risk of order delay increases significantly.

---