# How to reduce support ticket volume in ecommerce **Published:** April 13, 2021 | **Updated:** October 29, 2026 | **Authors:** Gladly Editorial | **Categories:** Best practices > Reduce ecommerce support ticket volume: find what drives conversations, prevent order status questions, fix returns, and let AI resolve the routine ones. --- Most ecommerce support teams can predict their busiest week by looking at the sales calendar. A promotion lands, orders double, and a few days later the inbox does too, because every order carries some chance that the customer will need to ask about it. If nothing about how you handle those questions changes, support volume grows in a straight line with revenue, and so does headcount. To bend that line, you'll find out what's driving the volume, stop counting the same person more than once, prevent the questions you can predict, fix the product and policy problems behind the ones you can't, and let AI resolve the routine questions that are left. The last section covers how to measure the reduction so you know customers are getting answers and not quietly giving up. ## Why ticket volume climbs with order volume Ticket volume is mostly a function of orders. Each order creates a handful of moments where a customer might reach out: right after checkout to change something, while the package is in transit, when it arrives in the wrong size, and when they decide to send it back. Multiply those moments by a holiday peak and you get the familiar pattern of backlogs, overtime, and seasonal hires who need training right before the busiest week of the year. Order volume is the number you're trying to grow, so the levers sit further down: how many of those orders turn into a question, and how many of those questions need a person to answer them. Every tactic below works on one of those two numbers. ## Find the contact reasons behind the volume Before you change anything, pull a sample of recent conversations and tag each one by why the customer reached out. Most ecommerce brands find the same categories near the top: - Order status and delivery timing (often called WISMO, for "where is my order") - Returns, exchanges, and refunds - Order changes and cancellations - Sizing, fit, and product questions - Damaged, missing, or wrong items - Discount codes, account access, and payment problems Rank the categories by volume and by how much effort each one takes. A category with high volume and a simple answer is where self-service and AI help most. A category with lower volume and long, messy conversations usually points to a policy or product problem that automation won't fix. Then pick the metric you'll track. Total conversation count rises whenever sales do, so a great sales month can look like a support failure. Contacts per order (sometimes called the order-to-contact ratio) divides the number of customers who reached out by the number of orders in the same period. If that ratio falls while orders grow, you're reducing volume in the way that matters to your budget and your customers. ## Count customers, not tickets Some of the volume on your dashboard is the same customer counted several times. A customer emails about a late order, gets no reply by lunch, opens a chat, and then follows up on Instagram. A ticket-based help desk can log that as three tickets, each one can land with a different team member, and the customer ends up with three threads and three partial answers. Reported volume and handle time both climb, and the customer ends up angrier than when they started. Gladly organizes support around people. Email, chat, SMS, and voice all land on one conversation timeline tied to the customer's profile, so a follow-up joins the conversation that's already open and the team member who picks it up can see everything that's happened. What you end up counting is customers who need help, which is the number you staff for. [Deckers](/customers/deckers/), the company behind UGG and HOKA, moved its support team off a ticket-based system and onto Gladly, and it has since reduced its order-to-contact ratio by 29%. Our guide to [text-based customer service](/blog/text-based-customer-service/) covers how Deckers' team uses SMS as part of that setup. KÜHL saw the same effect after moving to one customer profile: weekend email volume dropped from more than 100 to about 40. ## Prevent questions before they're asked Order status questions are the easiest to prevent, because customers mostly ask them when they haven't heard anything. An order confirmation with a realistic delivery window and a tracking link that works on the first tap heads off most of them. A short message when a carrier misses its date handles a lot of the rest, and it lands better coming from you than after the customer has refreshed the tracking page four times. The same idea works beyond shipping. If a style runs small, say so on the product page and again in the order confirmation. If a backordered item will ship separately, tell the customer at checkout, before they open the box and find half their order missing. Our post on [what WISMO is and how to reduce WISMO tickets](/blog/what-is-wismo-and-how-to-reduce-wismo-tickets/) goes deeper on order status messaging. ## Fix the product and policy causes Some questions keep coming no matter how fast you answer them, because the cause sits outside support. The contact reason tags from earlier tell you where to look. Returns are usually the first stop. If customers have to email to start a return, every return becomes a conversation. A self-service returns flow with clear rules on return windows, item condition, and refund timing takes most of those off your team's plate, and the conversations that remain are the exceptions that need a person's judgment. Sizing and fit questions point back to the product page. A size chart buried in a tab, fit notes that don't say how a style compares to your bestseller, and missing measurements all turn into pre-purchase questions, and often into returns a few weeks later. FAQs and help content cover the long tail, as long as they answer the questions customers ask in the words they use. Pull the exact phrasing from your tagged conversations and write the answers around it. Our post on [FAQ updates that reduce support volume](/blog/5-faq-updates-that-reduce-support-volume-for-dtc-brands-vx/) has specific examples, and these [knowledge base best practices](/blog/customer-centric-knowledge-base-best-practices/) cover how to keep help content accurate as policies change. When answers are easy to find, customers use them: 60% of Deckers' customers self-serve. Send what you learn back to the teams who own the cause. A spike in "arrived damaged" is a packaging or carrier problem, and a run of questions about one discount code is something marketing can fix in the next email. ## Resolve routine questions with AI After prevention and fixes, you're left with a large set of questions that have clear answers but still need someone to look something up: where an order is, whether it can still be changed, how to start an exchange. AI handles this work well when it can see the customer's orders and act in the systems where those answers live. A lot of guides call this ticket deflection and judge it by how many customers never reach a person. Resolution is the better test, because it asks whether the customer got what they came for. An AI that points someone to a help article they've already read only delays the question, and that customer usually comes back through another channel in a worse mood. Gladly AI connects to your commerce and logistics tools, so it can look up a customer's order history and take action on routine requests. When a question needs a person, it hands the conversation to your team with the full history attached, so the customer doesn't have to repeat anything. AI and team members work from the same customer profile and the same conversation, so a handoff doesn't start a second conversation either. V Shred reduced support contacts by 35% year over year, with an AI resolution rate of 35%. At [KÜHL](/customers/kuhl/), the AI answers common, repetitive questions and routes anything more complex to a team member, and the brand saw a 44% reduction in WISMO emails. For more on how this works, see our guides to [AI customer service for ecommerce](/blog/ai-customer-service-ecommerce/) and [agentic AI in customer service](/blog/agentic-ai-customer-service/). ## Measure the reduction without hurting CSAT Volume can drop for bad reasons. A contact form that's hard to find, a phone number pulled from the site, or an AI that loops customers back to the same article will all lower your volume, and the customers who gave up will show up later in reviews, chargebacks, and churn. Track these numbers together: - **Contacts per order:** the ratio from earlier, so you can compare a slow month with a peak one. - **Resolution rate:** the share of conversations resolved, split between AI and team members and broken out by contact reason. - **CSAT:** measured separately for AI-resolved conversations and for conversations with your team, so a drop in one doesn't hide in the average. - **Repeat contacts:** customers who come back about the same issue within a few days, which is the clearest sign that something marked resolved wasn't. If contacts per order and repeat contacts fall while CSAT holds steady, the reduction is real. [Rothy's](/customers/rothys/) is a useful benchmark here, with 31% of conversations resolved by AI and an AI CSAT above 93%. Our post on [how to tell if your AI is working](/blog/how-to-tell-if-your-ai-is-working/) covers which metrics to watch after launch. If the goal behind lower volume is lower spend, these guides to [reducing the cost of customer service](/blog/reducing-the-cost-of-customer-service/) and [optimizing support to lower costs](/blog/how-to-reduce-customer-service-costs/) cover the budget side. The easiest place to start is a week of tagged conversations, which will tell you which of these fixes to make first. --- *This content is provided by Gladly. Visit [gladly.com](https://gladly.com) for more information.*