Look at the distribution of incoming requests in e-commerce customer service and the same picture always appears: the bulk of them cluster around five questions, asked over and over. Most of the team's time goes to questions whose answers already sit in the system.
This post lays out a plan to close those five end to end. The critical phrase is "end to end": giving information is not enough, the transaction has to be completed too.
The five questions
1. Where is my parcel? The most frequent. The answer lives in the carrier's system, and the agent has to be able to look there.
2. When will it arrive? Estimated delivery date. Customers are usually looking for a date, not the word "in transit."
3. How do I return this? A process question, but usually with a transaction request behind it: opening a return.
4. Is this in stock, and when is it coming? The question where lost sales are most direct. If the answer is no, a back-in-stock notification should be registered.
5. Can I change my order? Address, size, color or cancellation. Time-sensitive; done before dispatch it costs nothing.
What these five share is that the answer exists in a system and each can end in a transaction. That makes them the cluster best suited to automation.
Information is not enough — finish the transaction
The most common half-built setup looks like this: the agent reads out the shipping status, the customer says thanks, the call ends. Is the customer satisfied? Partly. Was the problem solved? No — if the parcel is late, what the customer needs is not information but a resolution.
An end-to-end setup looks like this:
- If the parcel is late: open a delay record, notify the carrier, tell the customer when they will hear back.
- On a return request: open the return, generate the shipping code, send it to the customer.
- If out of stock: create a back-in-stock notification and confirm it to the customer.
- On an address change: update the address if not yet dispatched, or open a redirect request if it has.
These steps require integration. Without it the agent is just an information screen and the team's load does not fall — the questions simply go to the agent while the transactions still land on the team.
The difference between voice and chat
Chat dominates in e-commerce, but a call coincides with the moment patience runs out. Customers call when chat gave them no answer or when the matter became urgent.
So priorities on the voice channel should differ:
- The caller's record and latest order must be known in the first second of the call.
- The agent should not open with "may I have your order number"; if it already knows the number, confirmation is enough.
- Answers on voice must be short, and long process explanations moved to the written channel.
- If there is a complaint tone, the handoff threshold should fire earlier.
WhatsApp calling fits this scenario particularly well: the customer is calling from a number that already has message history, so the agent knows who they are and what you have been discussing.
Peak season plan
Load in e-commerce is not flat. Campaign days, season changes and pre-holiday periods multiply demand. Growing the team for those is impractical because the peak lasts a few days.
Preparing for peaks:
- Update the knowledge base before the campaign. Campaign terms, return exceptions and delivery times must be current in the agent; this is where errors most often appear.
- Add the expected question types in advance. Without campaign-specific questions defined, the agent falls into a knowledge base gap.
- Review handoff thresholds. When the team is already full during a peak, unnecessary handoffs create queues.
- Test capacity. Does concurrent call capacity cover your expected peak?
Measurement
The numbers that show the health of a voice and chat agent deployment in e-commerce:
- Self-resolved request rate. Requests closed without ever reaching the team.
- First-contact resolution rate. Requests with no second contact on the same topic.
- Transaction completion rate. Requests where an action was taken, not just information given. This is the real indicator of an end-to-end setup.
- Effect on return rate. If address verification and order change flows work, this number should fall.
- Peak-day answer rate. Are requests being missed on campaign days?
Where to start
Do not try to build all of it at once. Start with the single highest-volume question — in most businesses that is shipment tracking. Make that flow work end to end, measure it, then add the second.
The advantage of this approach is that it produces a measurable result within the first two weeks and earns the team's trust in the agent.
You can review the product side of e-commerce scenarios on the e-commerce assistant page.