5% of orders create a message. Which 5%?
Contact rate by product across 6,200 marketplace orders in 90 days: the handful of SKUs that generate a third of the work, what customers wrote about each, and what that tells you to fix.
The headline number and why it hides things
Over 90 days one UK furniture seller shipped 6,205 marketplace orders and had 496 customer conversations. Five per cent. It sounds like a customer-service number. It is a product and logistics number wearing a customer-service hat: the 5% is not spread evenly, and the pieces it is made of point at different fixes.
Two cuts do most of the work: contact rate by marketplace, and contact rate by product SKU across marketplaces.
By marketplace: 1% to 16%
The spread is not about the customers. Marketplace A routes every question to the seller and its customers expect a conversation about delivery; Marketplace E’s own support team answers most questions before the seller sees them. The same sideboard sells on both. Compare the operators by what they do with a message, not by the rate alone.
By product: the SKUs that make the work
Group by SKU across marketplaces (the same product has a different listing id on each), keep products with five or more orders, and the picture changes from “customer service” to “which products”.
Across the whole range the median product had a contact rate near 4%. The eight above sit at 29–60%. They are all large flat-pack items, and the reasons below say why.
What they wrote about
Each conversation is classified once, from the customer’s first message. For the high-contact products the reasons cluster:
- Delivery delay and item not received on the two TV units and the dining table. Those ship by two-man courier on a booked slot; when the slot moves, the customer writes. That is a carrier and communication problem. Which carrier, exactly.
- Missing parts and faulty item on the floating TV unit and the sideboards. Six quality complaints on 26 orders of one unit in 90 days is a packaging or QC problem, and it shows up as refunds later. What each one costs.
- Cancellation on the 38-order TV unit: customers ordering, then cancelling before dispatch when the delivery estimate appeared. That is a listing problem: the promised date was wrong on one marketplace.
What to do with a contact-rate table
- Sort by contacted, not by rate. The 60% product had three conversations; the 29% product had eleven. Fix the one that makes the work.
- Read the reason column before the product page. Delivery reasons point at the carrier or the promised date; missing parts and faults point at the product or its packaging; cancellations point at the listing.
- Compare the same SKU across marketplaces. The floating TV unit had 6 of 9 orders contacted on one marketplace and 0 of 5 on another. Same box, different courier.
- Recount monthly. A packaging change shows in this table within a month; it never shows in the overall 5%.
- 5% overall hides a 1%–16% spread by marketplace and 29%–60% on a handful of products.
- Group by SKU across marketplaces; listing ids differ per operator.
- Sort by number of conversations, then read the reasons: delivery → carrier, faults → packaging, cancellations → listing.
- Recount monthly; the fixes show up here long before they show in the total.
Questions people ask
What is a good customer contact rate per order?
Across five UK marketplaces and 90 days, one furniture seller saw 5% of orders lead to a customer message overall, with 1% to 16% by marketplace. For bulky, delivered goods anything under 5% is good; a single product above 30% has a specific problem worth finding.
How do you calculate contact rate per order?
Orders with at least one customer message in the period, divided by all orders in the period, on the same marketplace. Count threads, not messages, and exclude operator notices and operator-relayed system messages, or the rate is inflated by things the customer never wrote.
Why is contact rate so different between marketplaces?
Partly the customer base and partly the operator. A marketplace whose customer service team answers questions before they reach you shows a low rate; one that routes everything to the seller shows a high one. The product mix per marketplace matters too, so compare rates on the same SKU across marketplaces before drawing conclusions.