Weekly demand plan, plant-to-depot sourcing, freight benchmarks and dealer health for the Telangana and Andhra Pradesh network, with the Odisha unit as a scenario. Fed from SAP SD, the EDMS freight costing and truck GPS.
Line width is weekly tonnes. Hover a depot to see where it is sourced from.
Freight and handling saving, all depots
–
Annualised from the last 13 weeks of plans and dispatches, against the targets agreed with Sales and Logistics.
A regression on trend, seasonality, rainfall and the scheme calendar, compared against today's rule of thumb (same week last year). The last 13 weeks are held out to test both honestly.
Target stock = safety stock for forecast error and transit variability (95% service) plus half a four-day replenishment cycle, plus one day kept for walk-in dealer pickups. Stock value is taken at Rs 4,200 per tonne.
A minimum-cost network flow solves the whole week at once: which plant serves which depot, by road or rail, within plant capacity and siding limits. Today's plan is modelled as a planner giving each depot a home plant and using rail only on the longest lanes.
Largest cost movements first.
A diesel-indexed rate model fitted on 18 months of road freight invoices sets a fair rate for every lane. Lanes and transporters that pay above it are where renegotiation starts.
Each dot is one invoiced trip. Red dots are more than 8% above the model's fair rate.
Score weights cost against benchmark 40%, on-time delivery 40%, detention 20%.
A churn score flags dealers who are drifting before volume is lost. The pricing assistant sits inside the existing approval step: it suggests a price, the approver still decides.
Score built from volume trend, order frequency, payment days, scheme use, price gap to competitors and multi-brand stocking. Weights are illustrative until trained on real churn history.
A chat assistant that answers in the dealer's own language and looks up live order, dispatch and scheme data before it replies. Dealers use it on WhatsApp. This console shows the same conversation with the data lookups behind every reply.
Shows which system each answer came from, so every reply is auditable.