Unicall Logo
← Back to Blog
E-commerce 9 min read

Automate Where Is My Order Calls: A Practical Guide

Every vendor quotes a WISMO percentage. Almost none of them can tell you where it came from. Measure your own, split the calls by what actually went wrong, and automate the ones live order data can answer.

Steve G.
Steve G. AI Growth · Published November 16, 2026
A large fulfilment warehouse with rows of yellow bins and stacked cardboard parcels on blue shelving — where every “where is my order” call begins

If you plan to automate where is my order calls, the first number you need is not in any vendor deck. It is the number of order-status calls your store receives per 100 orders shipped, split by what actually went wrong with the parcel.

That sounds pedantic. It is the difference between a business case that holds up in month three and one that quietly falls apart. A caller whose parcel is simply in transit needs a ten-second answer. A caller whose parcel shows “delivered” and is not on the doorstep needs a claim, a decision and a follow-up. Both get counted as “WISMO”.

This guide is for heads of customer care and e-commerce operations: the data, a two-week audit, the call flow, a cost model and vendor questions.

The WISMO numbers you have read are mostly unsourced

Search for WISMO statistics and you will find three bands repeated everywhere: 10–25% of contacts, 25–40% of inbound volume, and 50–60% at peak. In August 2026, Ship24, a tracking-software vendor with every commercial reason to like a big number, tried to trace those figures back to primary research. It could not find a published study with a sample, a survey instrument or a method behind any of them (Ship24, August 12, 2026).

The review is right about why this matters. “Share of contacts” moves with what counts as a contact, which channels are included and the season. And a share of contacts is not a rate per order: sizing a team needs the first, judging your logistics needs the second.

If a WISMO figure has no denominator, it is not a benchmark. It is a sales line.

What the research does support is the cause. Delivery problems are common, and they make shoppers nervous:

  • Problems are the norm, not the exception. In Narvar's 2025 State of Post-Purchase survey of 3,461 US online shoppers (August 2025), 74% said they had a late delivery in the past year and 86% at least one delivery issue. Two-thirds said they feel anxious after clicking “buy” (Narvar, November 2025).
  • “Delivered” is not always delivered. In Narvar's 2026 holiday survey of 1,348 US consumers, 37% said they experienced a late package last year and 23% were told a package had been delivered when they never received it (Narvar, August 2026).
  • What shoppers want when it goes wrong is simple: acknowledgement (46%), a clear explanation (46%) and real-time updates (45%), according to the same 2025 survey.
  • Volume keeps growing. US retail e-commerce reached $340.2 billion in the second quarter of 2026, seasonally adjusted, 17.1% of total retail sales and up 12.2% on a year earlier (US Census Bureau).

Both Narvar studies are vendor research. Note what happens when one vendor asks a similar question twice: 74% in one survey, 37% in the other, with a different sample, wording and time frame. Your own call data beats any survey.

Run a two-week WISMO audit first

Before you talk to a vendor, tag every order-status call for two normal weeks, in the wrap-up code or by sampling 200 recordings. Capture five things per call:

  1. The order's real state at the time of the call (see the five types below).
  2. Whether the customer already had tracking and whether it was accurate.
  3. The days since the promised delivery date, if any.
  4. What the agent had to do: read a status, contact a carrier, issue a refund or reship, change an address.
  5. Handle time and time of day, including calls that rang out after hours.

Then compute two numbers: order-status calls per 100 orders shipped, and the share of those calls by type. Repeat in your peak month if you can.

The five kinds of “where is my order” call

Not every order-status call is the same job. Here is how we would sort them, and what an AI phone agent should do with each.

Call typeWhat the caller needsData the agent needsAI role
In transit, on timeReassurance and a dateOrder record, latest carrier scan, promised dateResolve
Not shipped yetA ship date, maybe a change or cancellationFulfilment status, stock or pre-order date, cancellation rulesResolve within written rules
Late or stuck in transitAn explanation and a next stepScan history, carrier exception codes, your late-order policyResolve or assist: explain, open a carrier inquiry, offer the remedy your policy allows
Marked delivered, not receivedTo be believed, and a resolutionDelivery scan details, claim rules, order historyAssist: collect the facts, open the claim, hand refund or reship decisions to rules or a person
Change before deliveryNew address, delivery instructions, holdFulfilment stage, what the carrier allowsAssist or human, depending on stage and fraud risk

The first two types are pure lookups: easy to automate, and just as easy to prevent with a better tracking email. The last three are where the phone line earns its keep. The customer is already worried, and a vague answer is how a late parcel becomes a chargeback.

How to automate where is my order calls: the call flow

Here is the flow we would build, in order.

1. Recognise the caller and find the order

Match the calling number to a customer record and their open orders before the first question. Asking for a 12-digit order number should be the fallback, not the opening line.

2. Verify before you reveal

A generic status (“your order has shipped”) carries little risk. A delivery address, a phone number or a change to the order does. Require a second matching detail, such as the email or postcode, before reading anything personal or changing anything. Put that rule in the system, not in the agent's instructions; as we argued in Confidently Wrong, behaviour that matters belongs in code.

3. Answer with a date, not a tracking status

“Label created” and “in transit” are carrier vocabulary. The caller wants to know when. Say the expected date, say what it is based on, and say plainly if it has slipped.

4. Handle the exception on the same call

For late, stuck or missing parcels, follow your written policy step by step: acknowledge, explain what the scan history shows, take the action the policy allows (carrier inquiry, claim, reship request) and say when the customer will hear back. That covers all three things shoppers told Narvar they want.

5. Leave a trail and a promise

Every exception call should create or update one ticket, not three, with the transcript and scan history attached. If the customer calls again tomorrow, the agent should already know why. Offer a follow-up by text or email when the status changes, and keep that promise.

6. Tag the reason for logistics

A spike in “delivered, not received” calls for one carrier and one region is a logistics problem a phone agent cannot fix. It can only report it, fast.

An illustrative cost model

Use your audit numbers. The figures below are assumptions for illustration, not benchmarks: a store shipping 100,000 orders a month, an average handle time of 4 minutes per order-status call, and a fully loaded cost of $35 per agent hour. Replace each with your own.

Calls per 100 ordersCalls per monthAgent hoursMonthly handling costPeak month at 2× volume
22,000133$4,667$9,333
44,000267$9,333$18,667
66,000400$14,000$28,000

Formula: orders ÷ 100 × calls per 100 orders × handle minutes ÷ 60 × cost per hour. It leaves out calls that ring out after hours, repeat contacts and orders lost when a worried customer gives up; add them if your audit shows them.

Be honest with the result. At 2 calls per 100 orders on a small store, order-status calls alone may not justify a voice agent; a better tracking page might. The case usually works when order status is bundled with returns and store-policy calls, which is how we scope it in our guide to AI phone agents for e-commerce customer service. On the price side, our managed service starts at $1,000 a month, billed on ticket volume rather than agent seats, with implementation quoted after the audit.

For reference, the results Unicall publishes as averages across its live deployments are about 60% of calls resolved autonomously and cost per ticket down 52% within 90 days, with sample sizes and periods varying by client. They are not a forecast for your store.

Fix the cause, then automate what is left

The cheapest order-status call is the one that never happens. Before or alongside a phone agent:

  • Show a real delivery date at checkout and hold to it. In Narvar's 2026 survey, reliable delivery dates influence 49% of consumers' purchase decisions.
  • Send the update before the customer has to ask, especially when a date slips. A late parcel announced by you generates fewer calls than one discovered by the customer.
  • Put the same answer on every channel. The tracking page, the email, the chat and the phone agent should read from the same order and carrier data, or customers will call to resolve the contradiction.

The calls that remain are the hard ones, and the ones worth answering well at 2 a.m. in December.

The scorecard to track from week one

  • Order-status calls per 100 orders, by call type, weekly.
  • Resolved without a human, separately for lookups and for exceptions.
  • Repeat calls within 7 days about the same order. This is the honest test of whether the answer was good.
  • Wrong answers found in transcript review, such as a wrong date or a promise outside policy. Read a sample every week.
  • Answer rate and time to answer, including after hours and peak days.
  • Claims and chargebacks on orders that generated a call, compared with the period before.

Questions to ask any vendor

Where does the agent get the order and carrier status, and how fresh is it?
Ask for the exact data sources (store platform, order management system, tracking platform) and a live demo against a real late order, not a happy path.
What does it require before it reads an address or changes an order?
You want a verification rule enforced by the system and visible in the logs.
What happens on a “delivered, not received” call?
Ask to hear a recording. Look for acknowledgement, the right questions, one ticket with context, and a clear next step.
Can it issue refunds or reships, and who sets the limits?
If yes, the limits should be written rules in your systems. If no, ask how the hand-off reaches your team and how fast.
Will you report calls per 100 orders and repeat calls, not just calls handled?
Calls handled is an activity metric. Fewer repeat calls and fewer claims are the outcome.

For how voice AI providers differ on delivery and accountability, our comparison hub applies the same ten buyer criteria to each vendor, from public information only.

FAQ

What share of customer service calls are “where is my order” calls?

There is no reliable published figure. The ranges that circulate, from 10% to 60% of contacts, could not be traced to a study with a sample and method in an August 2026 review by Ship24. Tag two weeks of your own calls by reason and calculate order-status calls per 100 orders shipped; that is the number to plan with.

Can an AI phone agent answer order status calls without verifying the caller?

It should not give out delivery addresses, contact details or account information before an identity check. A simple rule is to match the order number or phone number to one more detail, such as the email or postcode, before reading anything beyond a generic status.

What should the AI do when a package shows delivered but never arrived?

Acknowledge the problem, read the carrier's delivery scan, ask the standard checks your policy requires, then open a claim or ticket with everything attached and tell the customer when they will hear back. Refunds and reshipments should follow written rules enforced by your systems, or go to a person.

Do proactive tracking notifications make an AI phone agent unnecessary?

No. Good notifications remove the easy calls, which is worth doing first. The calls that remain are the exceptions: late, stuck or missing parcels. Those are the calls where a fast, informed answer on the phone protects the next order.

How long does it take to automate where is my order calls?

With a done-for-you provider, a first scope covering order status typically goes live in 4 to 6 weeks without replacing the phone system or helpdesk. Most of that time goes on the order-data connection, identity rules and testing the exception paths, not on the conversation itself.


Unicall.ai builds and operates done-for-you AI voice agents for customer care, led by e-commerce. The agent answers order-status calls with identity verification and live order data from Shopify, your OMS or your CRM, opens tickets with deduplication, and hands over to your team with full context when it should. See how it works with Shopify stores.

Find out what your order-status calls really cost

Book a free call-flow audit: we sort your order-status, delivery and returns calls by type and show which ones an AI agent should resolve, assist or leave to your team. Or call the live agent and test it yourself.

Get your free call-flow audit

Or call the live AI agent: (888) 890-8020

See how Unicall compares Get a free call-flow audit