Start and end
Where midnight orders go
GMV separate from net sales
Refunds return to the session
Two accounts must not double count
Define a session before ranking hosts
A stream runs from 23:00 to 01:00. A platform day cuts GMV at midnight. A session board that uses the start date keeps both hours on the first day. A board that uses the payment date splits the second hour onto the next day. Rankings change with the cut. Before a review, write the session id, the start, the end, and whether a midnight order stays with the session that opened or moves to the payment date.
GMV and net sales after refunds are different. A session with high GMV and a wave of refunds three days later will keep looking better if the rank never leaves GMV. The example is about attribution, not an account’s results.
A session, an order, and a payment are not one row
A session contains many orders. An order may be paid in parts, or may come from a short video or a product card rather than the session cart. Counting every payment with that room id pulls in off-session sales. Counting only payments within five minutes of a click drops later payments. The window length is a definition, not a technical detail. Kimball’s grain rule says the ranked row should be one session, not one payment.
If two accounts stream the same product, one order enters one session unless the business splits commission and the amounts still add back to the store. Double counting makes the hosts’ GMV exceed store GMV.
Send refunds back to the original session
A refund can arrive three days later, or next month. Posting the negative amount on the refund day’s session can create negative sales on a day with no stream, while the original session’s GMV never falls. A campaign review should move the refund back to the original session and keep the refund date. You can see session quality and the lag separately.
Say whether a partial return reverses units or money. If a returned gift reverses units but not the GMV already counted, session net sales will not match the order. The refund rate’s numerator is refunds assigned back to the session. The denominator is that session’s GMV. State both windows.
Two ranking paths
Path one sums GMV on the opening session. It suits the same-night review of traffic and checkout. It does not suit a bonus. Path two uses net sales on the original session after a stated refund window. It suits a rank that is closed only when the window ends. Until then, mark the rank as open. Do not pay a bonus from path one.
Do not split a midnight session by payment date and then compare the pieces with other full sessions. Compare the whole session, or say clearly that the board is a daily slice, not a session rank.
Counterexamples, acceptance, and the boundary
Counterexamples: a midnight session split across two days and then called the best single session; the same order counted on two accounts; a refund posted on the refund day so an empty day shows negative GMV; a bonus paid on net sales before the window closes. Prepare one midnight session, one late off-session payment, one partial refund after three days, and one order duplicated across two accounts. Check that session GMV does not exceed store GMV, and that refunds return to the original session.
BI0.AI can be a candidate place to take livestream orders into an operating review. This article does not show that the current release already posts refunds back to a session or handles two accounts. Use the room and order fields in the platform export.
Public references
Take the next step with BI0.AI
Talk through a real business scenario and see how governed AI BI can fit your team.
Explore BI0.AI