# Payment Split System Low-Level Design Source: https://www.techinterview.org/post/3233467853/lld-payment-split/ Updated: 2026-04-17 · techinterview.org ## Payment Split System — Low-Level Design A payment split system divides a single transaction among multiple parties: splitting a restaurant bill, charging multiple payment methods, or distributing marketplace revenue between a platform and sellers. This design appears in interviews at Stripe, Venmo, Airbnb, and Lyft. ## Core Data Model ``` PaymentIntent id BIGSERIAL PK initiator_user_id BIGINT NOT NULL total_amount_cents BIGINT NOT NULL currency TEXT NOT NULL -- 'usd', 'eur' status TEXT DEFAULT 'pending' -- pending, completed, failed, cancelled created_at TIMESTAMPTZ SplitParticipant id BIGSERIAL PK payment_intent_id BIGINT FK NOT NULL user_id BIGINT NOT NULL amount_cents BIGINT NOT NULL -- their share payment_method_id BIGINT -- their payment method status TEXT DEFAULT 'pending' -- pending, paid, failed, declined charged_at TIMESTAMPTZ charge_id TEXT -- external payment processor charge ID SplitTemplate id BIGSERIAL PK owner_user_id BIGINT NOT NULL name TEXT NOT NULL -- 'Roommates', 'Dinner group' participant_user_ids BIGINT[] split_type TEXT -- 'equal', 'percentage', 'fixed' created_at TIMESTAMPTZ ``` ## Calculating Split Amounts ``` def calculate_splits(total_cents, participants, split_type): if split_type == 'equal': base = total_cents // len(participants) remainder = total_cents % len(participants) amounts = [base] * len(participants) # Assign remainder cents to first N participants for i in range(remainder): amounts[i] += 1 return amounts if split_type == 'percentage': # participants = [{'user_id': 1, 'pct': 60}, {'user_id': 2, 'pct': 40}] amounts = [int(total_cents * p['pct'] / 100) for p in participants] # Fix rounding: add remainder to largest share diff = total_cents - sum(amounts) max_idx = max(range(len(amounts)), key=lambda i: amounts[i]) amounts[max_idx] += diff return amounts if split_type == 'fixed': # Validate: sum of fixed amounts must equal total if sum(p['amount_cents'] for p in participants) != total_cents: raise ValueError('Fixed amounts do not sum to total') return [p['amount_cents'] for p in participants] ``` Critical: always compute splits in integer cents. Float arithmetic loses pennies. When rounding produces a remainder, assign it deterministically (e.g., to the initiator or largest share). ## Charging Participants ``` def charge_participant(participant_id): participant = db.get_for_update(SplitParticipant, participant_id) if participant.status != 'pending': return # Already processed (idempotency) try: charge = stripe.charge.create( amount=participant.amount_cents, currency=participant.currency, payment_method=participant.payment_method_id, idempotency_key=f'split-{participant.id}', ) db.execute(""" UPDATE SplitParticipant SET status='paid', charged_at=NOW(), charge_id=%(charge_id)s WHERE id=%(id)s AND status='pending' """, {'charge_id': charge.id, 'id': participant_id}) except stripe.CardError as e: db.execute(""" UPDATE SplitParticipant SET status='failed' WHERE id=%(id)s """, {'id': participant_id}) notify_participant_payment_failed(participant_id, e.message) # Check if all participants are done check_and_finalize_payment_intent(participant.payment_intent_id) ``` ## Payment Intent Finalization ``` def check_and_finalize_payment_intent(intent_id): participants = db.query(""" SELECT status, COUNT(*) as cnt FROM SplitParticipant WHERE payment_intent_id=%(id)s GROUP BY status """, {'id': intent_id}) counts = {row.status: row.cnt for row in participants} total = sum(counts.values()) paid = counts.get('paid', 0) failed = counts.get('failed', 0) + counts.get('declined', 0) if paid + failed == total: # All settled new_status = 'completed' if failed == 0 else 'partial' db.execute(""" UPDATE PaymentIntent SET status=%(status)s WHERE id=%(id)s AND status='pending' """, {'status': new_status, 'id': intent_id}) notify_initiator_split_complete(intent_id, paid, failed) ``` ## Handling a Declined Share When one participant's card declines, the initiator has three options: - **Request re-payment:** Notify the participant to add a new payment method and retry. - **Cover the shortfall:** Charge the remaining amount to the initiator's card. - **Cancel the split:** Refund all participants who already paid. ``` def handle_declined_share(intent_id, declined_participant_id, initiator_choice): if initiator_choice == 'cover': intent = db.get(PaymentIntent, intent_id) declined_amount = db.get(SplitParticipant, declined_participant_id).amount_cents # Charge initiator for the shortfall charge_user(intent.initiator_user_id, declined_amount, idempotency_key=f'cover-{declined_participant_id}') elif initiator_choice == 'cancel': # Refund everyone who already paid paid_participants = db.query(""" SELECT * FROM SplitParticipant WHERE payment_intent_id=%(id)s AND status='paid' """, {'id': intent_id}) for p in paid_participants: stripe.refund.create(charge=p.charge_id, idempotency_key=f'refund-split-{p.id}') db.execute("UPDATE SplitParticipant SET status='refunded' WHERE id=%(id)s", {'id': p.id}) db.execute("UPDATE PaymentIntent SET status='cancelled' WHERE id=%(id)s", {'id': intent_id}) ``` ## Key Interview Points - **Integer arithmetic only:** Never use floats for money. Store amounts in the smallest currency unit (cents). Round remainders deterministically. - **Idempotency keys per participant charge:** Use split-{participant_id} as the Stripe idempotency key so retries never double-charge. - **Async charging:** Charge participants in parallel via a job queue — do not block the HTTP response waiting for all card charges to complete. - **Partial completion state:** A split where some pay and one declines is a valid terminal state. Track it explicitly rather than treating partial completion as failure. Payment split and multi-party charge design is discussed in [Stripe system design interview questions](https://www.techinterview.org/post/3233460268/stripe-interview-guide-2026-process-bug-bash-round-and-payment-systems/). Payment split and fare distribution design is covered in [Uber system design interview preparation](https://www.techinterview.org/post/3233460273/uber-interview-guide-2026-dispatch-systems-geospatial-algorithms-and-marketplace-engineering/). Payment split and marketplace payout design is discussed in [Airbnb system design interview guide](https://www.techinterview.org/post/3233460279/airbnb-interview-guide-2026-search-systems-trust-and-safety-and-full-stack-engineering/).