In a nutshell
Slow responses, inconsistent compliance with platform deadlines, and unscalable support workflows don’t just create friction — they directly degrade seller performance scores, reduce visibility, and cost revenue. This article maps the operational reality of multi-marketplace support and what a scalable approach actually looks like in practice.
⏱ Time to Read: appr. 9 min
The optimization gap
Listings, ad spend, pricing, promotions — these are the areas most marketplace managers tune regularly. What tends to get far less attention is the inbox.
eDesk, a customer support platform built specifically for e-commerce sellers, handles more than 3 million customer conversations per month across thousands of seller accounts. One of the most consistent patterns they observe across that customer base is the gap between how much sellers invest in the commercial side of their operations and how little attention the support side receives — right until something breaks.
“Most marketplace sellers optimize for everything except inbox,” noted Gareth Cummings, CEO of eDesk, during the webinar How to Win More Sales on Marketplaces Without Losing Revenue in the Inbox, hosted by Marketplace Universe. “Ad spend, listings, pricing — these get tuned every week. Inbox operations? Often untouched.”
As a seller scales across more marketplaces and drives more traffic, inbox volume rises. Service level agreements (SLAs) start slipping. Message workflows that worked for three channels stop working for ten. Response consistency degrades. The compounding effect of those operational failures shows up directly in seller performance metrics — and that is where the revenue impact becomes hard to ignore.
👉 Marketplace Universe Insight: The median eDesk seller operates across 17 channels (marketplace × country). 85% of sellers on the platform operate between 3 and 30 channels simultaneously. At that scale, inbox management is not a minor operational detail — it is a core business function.
Why seller performance is an operational question
Every major marketplace — Amazon, Zalando, bol, eBay, Kaufland and others — maintains a seller performance score. The exact methodology varies, but the inputs are remarkably consistent: response times, SLA compliance, review scores, claims and escalation rates.
These scores directly determine a seller’s visibility on the platform. Miss SLAs consistently and products rank lower. Miss them badly enough and a seller gets delisted entirely — with products sitting in fulfillment, no longer reachable by the customers they already paid to acquire.
Public holidays, weekends, and time zone differences can easily cause an SLA to be missed without a single agent making an error. Getting back on track requires actively demonstrating to the marketplace that the problem has been fixed. Meanwhile, the products are offline.
Pre-sales: the most underrated conversion driver
Much of the discussion around marketplace customer service focuses on post-purchase: returns, tracking queries, complaints. The pre-purchase window gets less attention — and according to eDesk’s platform data, this is where the biggest conversion opportunity sits.
Pre-sale questions look mundane: “Will this arrive before Friday?” “Is this compatible with X?” “Do you ship to Austria?” But buyers asking these questions are close to a purchasing decision. They are high-intent. They need an answer, not a follow-up.
eDesk’s data shows that sellers who respond to pre-sale questions within 15 minutes see a 40 to 50 percent conversion uplift compared to those who respond the same day or later.
“You’re in the mode to buy. It could be the evening or the weekend. If you can answer quickly, typically you’ll make that purchase. If it comes back the next day, you might have moved on. You’ve lost that sense of urgency.”
— Gareth Cummings, CEO, eDesk
👉 Marketplace Universe Insight: 92% of webinar attendees said same-day response was their current standard for pre-sale questions. According to eDesk’s platform data, same-day is not fast enough to capture the conversion uplift — the threshold that makes a material difference is 15 minutes.
Hitting that window consistently across 10, 20, or 30 channels simultaneously is not straightforward. That is where prioritization logic and workflow structure become operationally critical.
The Pertemba case: what scaling to 160 marketplaces actually required
Pertemba is what Irene Epp, its Operations Support Manager, describes as “a bit of a unicorn.” The company does not sell its own branded products — it brings brands onto e-commerce marketplaces and manages the entire selling and service operation on their behalf, including all customer communications across all channels and languages.
The numbers alone tell a story:
- 160 active marketplace channels
- 1,000+ customer emails per day
- ~18 people in the entire customer service team
The operation started on Gmail, with messages processed sequentially and routing done by hand. Scaling to the current scope was not possible on that model.
“If you would need to log in and out of marketplaces — even if it takes only a minute — a minute times 130 is already a lot of time gone. Really not possible to do.” — Irene Epp, as pointed out on Marketplace Universe’s podcast Let’s Talk Marketplace, episode 145 (#LTM145)
Want to know more? Then check out episode 145 of our podcast “Let’s talk Marketplace”! In it, Irene Epp and Garth Cummings dive deeper into automation and how to handle channel expansion while keeping customers happy:
The shift to eDesk’s platform changed the structural logic entirely. SLA timers are set at the channel level, pre-sale tickets can be filtered into a dedicated queue, and AI-assisted translation handles incoming and outgoing messages across all languages — eliminating the need for language-specific hires at scale.
“We can prioritize our marketplaces on need. We can prioritize emails on need. For pre-sale questions, we set a filter and a standard SLA — I want it answered within an hour, that’s the maximum. With eDesk, changing the SLAs for specific tickets means they pop up on top for you,” Epp explained during the webinar.
What actually changed operationally:
| Before | After | |
| Routing | Manual | Intelligent, SLA-based |
| Templates | ~1,000 across all language/scenario combinations | ~140 standardized, AI translation handles languages |
| Second-line team | 12 people | 7 people |
| SLA compliance | — | 97.6%, including peak season |
👉 Marketplace Universe Insight: Around 90% of all incoming queries fall into a handful of standard categories — where’s my order, I want to cancel, what’s my tracking number, how do I return this. Automation handles the volume. Human agents handle everything else.
Multi-lingual support is no longer optional
For European marketplace sellers, multi-lingual support is a structural requirement. Marketplaces in Germany, France, the Netherlands, Spain, Poland and elsewhere expect — and in many cases require — responses in the local language. Getting the tone wrong carries real consequences — Pertemba once had a strongly negative customer reaction simply from an automated message using the informal “du” instead of the formal “Sie” in German.
The Japanese market adds another layer entirely. Epp described receiving what read in translation as a polite message — only for Japanese-speaking colleagues to point out the customer was actually furious. “What they put in writing is not always what they are saying to you. For a Japanese person, they are so polite when they are mad. As a European or English speaker, you wouldn’t pick up on that culture.”
The practical solution at Pertemba is to cross-train all agents across all marketplaces and let AI-assisted translation handle the language layer in both directions.
“Even if one agent is sick or on holiday, we don’t have that hole and the scramble to not hit the SLA because we don’t have that skill in our team. That’s why we could expand to 160 channels — because yes, we have some Spanish speakers, Japanese speakers, German and French. But that’s it. We don’t need to hire for every language.”
— Irene Epp, Let’s Talk Marketplace #LTM145
Customer support as a business intelligence function
Irene Epp describes the customer service team as “the watchdog of the company” — the part of the business where problems surface first. A SKU with unusually high return rates, a carrier struggling with deliveries in a specific region, customers in a given market suddenly asking the same question: all of it arrives in the inbox before it shows up anywhere else.
Cummings offered a concrete example from eDesk’s customer base. A seller had a top-performing SKU by sales volume — but when they surfaced the support data, the picture shifted: that product was generating high returns, negative reviews, and suppressed rankings. The seller had been driving sales for a product that was quietly undermining their ability to sell on the platform.
“If you went and spoke to your customer support team today and asked them what are the top three things that are happening, they will tell you straight away — they’re dealing with this every day. And you probably will be surprised in terms of what you would hear,” Cummings noted during the webinar.
Scaling support without breaking the team
The operational argument for scalable tooling is not just efficiency — it is retention. Every agent who leaves takes knowledge of workflows, marketplace rules, escalation paths, and communication patterns with them.
“You can set your limits as high as you want, but you will find you’re replacing one agent after another. Because at some point they either burn out or they’ve had enough. It’s a stressful job. You get abuse. It’s not even the best-paid one. Keeping your team happy is very important — because every team member that leaves is knowledge that leaves.”
— Irene Epp, Marketplace Universe webinar
Epp’s model is to keep the team at around 80–85% capacity for most of the year, preserving headroom to absorb peak periods. At Pertemba, the core team has remained stable for three years across continued channel growth — a result she attributes in part to tooling that removes a significant portion of the manual cognitive load from agents.
A structured internal knowledge base, integrated into eDesk and searchable by any agent covering an unfamiliar channel, sits at the center of that stability. The more comprehensive it is, the more effectively both the team and the platform’s AI automation can draw on it.
Where to start: the one metric that tells you everything
Both Cummings and Epp converge on a simple starting point in the Let’s Talk Marketplace podcast: measure your actual response times.
“You’d be surprised sometimes — if you’re not tracking it and then you discover it can be days or longer. So that’s the first metric you should look at: how quickly are we getting back to our customers. And that will tell you a lot,” Cummings said.
Response time is the metric that reveals whether the structure is working. Once that is visible, the path toward fixing it becomes clear.
Key Learnings
- SLA compliance is a revenue question. Response times, reviews, and escalation handling feed directly into marketplace visibility scores — and missing SLAs due to public holidays or weekends alone can take products offline. Leaving support operations unstructured while investing in listings and ads creates a compounding risk.
- Around 90% of marketplace support queries are standard and automatable — order status, cancellations, tracking, returns. Automation handles the volume; human agents handle everything else.
- Responding to pre-sale questions within 15 minutes can drive a 40–50% conversion uplift, according to eDesk’s platform data. Same-day response — the current standard for most sellers — is not fast enough.
- Multi-lingual support is a structural requirement in Europe. AI-assisted translation, properly configured, allows teams to cover far more markets than language-specific headcount would permit.
- Customer support is business intelligence. When the team has enough capacity to pay attention, the inbox surfaces product problems, carrier failures, and assortment issues before they show up anywhere else in the business.
- Start by measuring response time. If actual response times across channels are not being tracked today, that is the first metric to surface. It will tell you more than almost anything else about the state of the support operation.
Missed the live webinar? Watch the full recording here. For the full Pertemba story in their own words, listen to the Let’s Talk Marketplace podcast episode 👉 Automation at Scale: How to handle 1,000+ Daily Support Tickets #LTM145.
eDesk is a customer support platform built for e-commerce sellers, handling more than 3 million customer conversations monthly across thousands of seller accounts globally. Learn more or start a free two-week trial at edesk.com.
15.06.2026 – Written by Ricarda Eichler, Journalist and Author for OHN