Batch
Score a file of orders
Upload a CSV of orders and get every row back with a risk score, a decision and what that decision is expected to cost. Same model, same thresholds, same arithmetic as the single-order console.
Upload
One row per order, one column per checkout field. Up to 5,000 orders per file.
What the file needs to contain
A header row and these 25 columns, in any order. Add order_id if you want your own reference echoed back. Any other column is rejected rather than ignored — a misspelled header should be an error, not a silently dropped signal. Leave prior_return_rate and days_since_last_order blank on a customer’s first order; blank means “no history”, and 0 would tell the model they have a spotless one.
Customer
- account_age_days
- prior_orders
- prior_rto_count
- prior_return_rate
- days_since_last_order
- is_guest_checkout
- phone_verified
- email_domain_type
Order
- order_value
- item_count
- category
- discount_pct
- coupon_used
- payment_method
- hour_of_day
- day_of_week
Address
- pincode
- city_tier
- address_char_len
- address_has_landmark
- shipping_billing_city_match
- address_edited_at_checkout
Velocity
- orders_from_pincode_24h
- orders_from_phone_7d
- distinct_addresses_for_phone
The sample file is 60 orders drawn from the validation months of the synthetic dataset, with the labels removed — enough to see all three bands without preparing anything.
Scored orders
Every row that validates, with its band and its expected cost.
No file scored yet.
Drop a CSV above, or download the sample file and upload that to see how it works.