Marketplaces · Buyers, sellers & disputes
Two sides. One trust problem.Every ticket.
Buyers ask where their order is. Sellers ask where their payout is. Both ask who's right when something goes wrong. Three queues, one operation.
“The seller hasn't shipped my order. It's been six days.”
AI collects → human resolves
Working…
- “The seller hasn't shipped my order. It's been six days.”
- “When does my payout land?”
- “Why was my listing removed?”
- “I paid and the item never existed. This is a scam.”
- “The buyer says the item was damaged, but I packed it perfectly.”
- “Can I change the delivery address?”
- “How do I set up shipping rates for international orders?”
- “I received a chargeback notice for an order that was delivered.”
What creates volume
- Transactions completed
- New seller onboarding
- Payout cycles
- Dispute & return policy
- Fraud pressure & promotions
Operating workflow
One contact, three possible owners.
A single "something's wrong with my order" contact is triaged into a buyer, seller or platform path. Each path has its own tools, its own tone and its own escalation.
Trigger
Buyer: "The seller hasn't shipped my order."
Buyer
- Order & payment status→
- Delivery exception→
- Buyer protection claim→
- Refund or replacement
Seller
- Listing, payout & account→
- Policy or performance flag→
- Seller specialist→
- Decision recorded
Platform
- Signal detected→
- Trust & safety desk→
- Action taken→
- Feedback to product
- 1
Side identified
Buyer · order #M-71822
- 2
Transaction retrieved
Order, payment, seller record, message history
- 3
Status resolved by AI
Awaiting dispatch · seller SLA · reminder sent
- 4
Threshold crossed?
Dispatch SLA breached · claim eligible
- 5
Routed to the right queue
Buyer, seller or trust & safety
- 6
Case closed in Zendesk
Both parties notified · reason tagged
Resource allocation
Where AI adds speed. Where people add judgment.
AI opportunities
- Order & delivery status
- Payout schedules & fee questions
- Listing how-to
- Address & contact updates
- Policy text & eligibility checks
- Account setup guidance
- Evidence collection for claims
- Seller reminders & nudges
Human-led interactions
- Buyer–seller disputes
- Fraud & scam reports
- Chargeback representment
- Listing removals & appeals
- Account suspensions
- High-value transactions
- Seller performance conversations
- Policy exceptions
Where service meets the business
A dispute handled well keeps both sides on the platform.
Where judgment, evidence and speed change the outcome of a transaction — and who stays.
- 01
Damaged-item dispute opened
- 02
Agent sees both sides' photos and carrier scans
- 03
Partial refund ruled within 24h
- 04
Buyer keeps buying · seller rating protected
Outcomes depend on your policies, take rate and category mix. Dispute rulings follow your rules, not ours.
When volume spikes
Contacts follow transactions. Disputes lag them by a week.
Contacts per week around a promotional peak. Status and how-to questions spike with orders and are absorbed by AI; disputes and fraud reports arrive later and always reach people.
Promotions double buyer status contacts within days; disputes and chargebacks peak 7–14 days later and need the trust desk staffed for it.
What we report
The KPIs that matter to a marketplaces operation.
First Response Time
48 sec
−58%
Dispute Resolution Time
1.6 days
−44%
Buyer CSAT
4.6 / 5
+0.3
Seller CSAT
4.4 / 5
+0.5
Dispute Rate
1.8%
Chargeback Win Rate
61%
+14 pt
Contacts per 1,000 Transactions
115
Fraud Case Time-to-Action
3.2 h
−52%
Systems we work inside
- Zendesk
- Salesforce Service Cloud
- Intercom
- Stripe
- Adyen
- PayPal
- Sift
- Internal admin tools
We operate in your systems, not beside them. Every interaction is logged, tagged and visible where your team already works.
How the operation is structured
01AI front line
Order, payment and payout status, listing how-to and policy checks on chat, email and SMS — for both sides, in your brand voice.
02Buyer queue
Delivery exceptions, protection claims and refunds, with the full transaction and seller history in view.
03Seller queue
Appeals, performance flags, payout holds and account questions handled by people who know your seller policies.
04Trust & safety desk
Fraud, disputes and enforcement — human-only, working from Sift signals and the complete transaction graph.
05QA & analytics
Contact reasons, dispute causes and fraud patterns reported back to product, policy and marketplace ops.
Marketplace calculator
What does support cost per 1,000 transactions?
Your transaction volume, both sides' contact rates and your dispute rate — sized into buyer, seller and trust queues before you talk to anyone.
Other industries
Evaluate your customer-service operation.
See what your current model costs, what a managed Human + AI operation would look like, and where the gaps are — before you talk to anyone.
