E-commerce customer requests: what we learned from 10,000 tickets
Published on November 4, 2026

When you read one support request, you see a customer with a problem. When you read ten thousand, you see patterns. The same subject lines, the same triggers, the same points in the order process where customers become unsure and reach for the keyboard. The first ten thousand tickets that went through Humane came from Shopify stores of various sizes and industries in German-speaking markets. We categorised them, answered them or put them forward as drafts, and saw which replies close the ticket and which trigger the next follow-up.
This article is not a collection of statistics. It is the list of things we would not have expected in this form, and which you will probably recognise in your own inbox. For industry figures we link to external sources; our own observations we describe as what they are: observations.
1. The most common request is not a question for the shop
That was to be expected, but not the extent of it. According to Gorgias, delivery questions are on average 18 percent of incoming requests in e-commerce, making them the largest single category. In the stores we see, the delivery question is almost always number one, and it has two faces.
The first face is harmless: the customer has a tracking number, but the status means nothing to them. They want one plain sentence and a date. The second face is the conflict case: status “delivered”, parcel not there. Both land in the inbox under the same subject line, and both need completely different replies. A support team that treats both the same annoys one group with bureaucracy and the other with canned text.
What this means for you: Sort delivery questions by status, not by subject. Anything with a normal carrier status is routine. Anything with “delivered”, “undeliverable” or no scan for days is a case. In detail: reducing WISMO requests.
The first day after dispatch produces the most unnecessary tickets
A large share of delivery questions arrives before the carrier has even scanned the parcel. The shipping email has gone out, the customer clicks the link, sees “shipment announced” and nothing else. From their point of view the order has been “on its way” since yesterday and is not moving.
These tickets have one property that sets them apart from all others: they can be prevented with a single sentence in the shipping email. “The status usually updates in the evening or the next morning.” Stores that have this sentence get visibly fewer first-day requests. That is not automation, that is communication.
What this means for you: Before you automate anything, open your Shopify “Shipping confirmation” template and add that sentence.
2. The second email is a different category from the first
A customer asking for the first time is curious or slightly unsure. A customer writing a second time about the same order is impatient, and by the third time they are angry, regardless of how politely they phrase it. You can see it in the tickets by length: the first email is one sentence, the second is three, the third has exclamation marks.
The most common reason for the second email is not that the first reply was wrong. It is that it contained no date. “Your parcel is on its way” does not reassure. “Expected Thursday” does, even if Thursday is later than hoped. This matches what customers say in general: according to Zendesk CX Trends 2026, 74 percent of consumers are frustrated when they have to repeat information, and a customer writing for the second time has to do exactly that.
What this means for you: Every second message about the same order belongs outside automation and with a human who takes over the case. And every first reply needs a date.
3. Most order changes arrive within an hour of purchase
Wrong address, wrong size, ordered twice, forgot the voucher. Almost all of these requests arrive shortly after the order confirmation, often within minutes. The pattern is so clear you could set your watch by it: the customer reads the confirmation email, spots the mistake, writes immediately.
The problem is response time. If you only read the email the next morning, the order is in the warehouse, the label is printed, the address can no longer be changed. A request that would have cost two minutes turns into a return with a replacement shipment. Customers expect speed here: according to the HubSpot State of Customer Service survey 2025, 67 percent of consumers expect their issue to be resolved within three hours.
What this means for you: Order changes are the category where speed directly saves money. A reply within an hour, even just “we have changed the address”, prevents a return. How to answer these requests in practice: Changing a Shopify order after purchase.
4. The “other” category is never as big as you think
Many merchants believe their requests are too individual for categories. When you actually go through the tickets, a handful of categories cover almost everything:
- Delivery question and shipment status
- Registering a return, return status, refund status
- Order change and cancellation
- Product question before purchase
- Complaint after purchase
- Discount, voucher, invoice
What is left after that is a small remainder that is genuinely individual. That remainder matters, it contains the interesting cases, but it is small.
The feeling that everything is individual arises because the individual cases take more time and therefore stick in memory. The routine cases are forgotten, even though they fill the day.
What this means for you: Count for one week. The list of the 12 most common customer requests in e-commerce covers most of the inbox in most stores, and for each of these categories there is a standard reply you can write well once.
5. The replies that close tickets are shorter than expected
We compared which replies trigger a follow-up and which do not. The good replies are short, name one concrete piece of information (date, amount, next step) and say what the customer should do if it does not work out. The bad ones are long, apologise twice and end with “Should you have any further questions, please do not hesitate to contact us”.
The same applies to the replies the agent writes. During the training week merchants almost always shorten the drafts, and when they add the reason, the agent learns from it. “Not like that” becomes a rule that applies to all following replies. How that works: Correct once, right every time after.
What this means for you: Cut every sentence in your templates that contains no information. Replace “shortly” with a date.
6. Angry customers are rare, but they cost the most
The share of genuinely angry messages is small. But each one costs many times the usual effort, and each one is a potential public review. According to Zendesk CX Trends 2026, 85 percent of CX leaders say a single unresolved issue is enough to lose a customer.
What we see in the tickets: anger rarely starts with the first problem. It starts with the second unanswered email, with the canned text in response to a specific question, with the sentence “that is the carrier’s responsibility”. And de-escalation almost always succeeds with the same pattern: name the problem, take responsibility, give a concrete next step with a date.
What this means for you: Angry messages need to be recognised and handled differently from routine. At Humane the agent detects the tone and puts such tickets in front of you according to your rules, instead of answering them automatically. What automation should fundamentally not take over here is described in The 90 percent limit.
7. Customers want to know who is answering, but they do not insist on a human
A Bitkom survey from 2025 shows that 62 percent of online shoppers would prefer a quickly reachable human contact when there is a problem. In the tickets we see what is behind that: customers do not want a human for its own sake. They want someone who knows their order, answers the question and does not dodge. A reply that names the real shipment status and a date is not perceived as a “bot”. A reply that asks for the order number that is already in the subject line is.
What this means for you: The yardstick is not “human or machine” but “does the reply know my order”. Anything that fails that condition should not be sent automatically.
In short
Ten thousand tickets later, the most important lessons are fairly unspectacular. Most requests are predictable, some of them can be avoided with one sentence in the shipping email, and the replies that work are short and contain a date. The second email about the same order is always a case for a human. And customers do not notice whether a reply was generated automatically; they notice whether it knows their order. The complete framework for support in your store is in Customer service for online shops: the complete guide.
If you want to see which patterns are hiding in your own inbox: try 100 customer requests for free, no credit card; in the first week every request is only categorised and put forward as a draft.