One question drives half your peak volume. What is it costing you to answer it?
Where is my order is the largest ticket category most retailers have. It means the contact centre is sized around a question the operation could answer for itself.
Most order status contact comes from orders that did not arrive as promised.
Order placed
A promise date the operation has not confirmed it can hit.
A promise date checked against real stock and real carrier capacity.
Dispatch
Tracking activates late or not at all. The customer sees nothing.
Tracking live at dispatch, with the exception flagged before the customer notices.
The wait
No update. The customer refreshes a page and then gives up on it.
Proactive updates at the points where people normally start worrying.
Contact
An agent opens three carrier portals to answer one question.
Answered from live order and carrier data without a person touching it.
Return
A form, a wait, and a refund the customer has to chase.
Started and completed in the same conversation, on any channel.
High order status volume is an operating signal rather than a contact centre workload problem. It reflects how well fulfilment and inventory accuracy and carrier performance line up behind the promise you made at checkout.
Five systems know where the order is. The customer asks the one that does not.
An agent answering a single order status question typically opens three of these and spends up to five minutes doing it.
Ecommerce platform
What was ordered, but not where it is.
Warehouse system
Whether it was picked, but not whether it shipped.
Carrier portals
Where it is, in three different formats.
Helpdesk
That they asked, and how often.
Phone system
The most expensive channel per resolution.
Most retailers arrive with the same question. Can AI absorb peak without making the experience worse?
It is a fair question, because deflection on its own can make the experience worse. Median AI deflection in ecommerce reached about 41 per cent in 2026, and some of that is customers being turned away rather than helped. A deflected customer often picks up the phone, which is the most expensive channel.
Fortay Connect is an independent CX and communications consultancy. We work across the market leading communications and contact platforms, which is what lets us stay outcome focused and technology neutral. Retailers bring us in to find why the contact is happening at all, work out which of it can be resolved rather than deflected, and build that on whatever fits the estate.
Most of the value is in reducing why customers contact you at all.
The Customer Conversation Review
Three weeks. We measure what your contact actually costs by channel and by driver, then show what can be removed before peak. You keep the findings.
What the review coversWhat this looks like in practice
Order status and returns handled end to end against live order data, so the contact is resolved rather than moved to a more expensive channel.
Four areas where the work shows up in cost per order
Ordered by how quickly the value arrives. The first one usually pays for the rest.
Order status answered, not deflected
Connected to live order and carrier data so the answer is real rather than a link to a tracking page.
- Identity verified from phone number or postcode or order reference
- Live status pulled from the carrier rather than from the last known update
- Exceptions flagged proactively before the customer asks
- Handled on voice as well as chat, because voice is where the cost sits
Returns completed in one conversation
The January wave is predictable, which makes it the easiest to plan for.
- Purchase verified and reason captured in the same interaction
- Return or exchange processed without a form and a wait
- Refund status answerable without a second contact
- Return reasons fed back as product and listing data
Peak absorbed without peak hiring
Capacity that scales on the day rather than being recruited and trained in September.
- Volume surges handled without a queue collapse
- Seasonal recruitment reduced to the roles that need judgement
- Agents left with the contacts that actually need a person
- Surge behaviour tested before peak rather than during it
The contact you stop causing
The cheapest contact is the one that never happens, and the data tells you which ones those are.
- Contact rate measured per shipped order rather than in raw volume
- Drivers traced back to fulfilment node and carrier and listing
- Promise date accuracy reported as an operational metric
- Root cause reduction separated from simple deflection in reporting
Fashion and apparel
Return rates are the business model. Speed of refund decides whether they order again.
Grocery and convenience
Substitutions and slot changes, where the contact arrives inside a two hour window.
Home and furniture
Long lead times and two person delivery, so the wait is measured in weeks not days.
Electronics
Warranty and setup questions that turn into technical support after the sale.
Health and beauty
Subscription changes and replenishment, where the contact is about the next order.
Peak is a deadline
A retailer deciding in November is unlikely to have anything in place for that trading period, because anything needing order data wired in and tested takes longer than the runway left at that point.
Sources: published ecommerce support benchmarks for 2026 and retail contact cost studies.
What the AI never does
- Promise a delivery date it has not verified against live carrier data
- Approve a refund or goodwill payment outside your published policy
- Handle a product safety or injury report. That goes to a person immediately
- Claim stock availability it has not checked against the live position
- Hold a customer in a loop. There is an exit to a person at every step
What we put in place
- A live connection to order and carrier data, so the AI can act rather than only talk
- Resolution measured rather than deflection, with repeat contact tracked
- Load testing against your forecast peak before the trading window opens
- Contact rate reported per shipped order, so improvement is provable
- A fallback plan for carrier failure during peak
From discovery to embedded change
We work across the market leading platforms, so the design follows the outcome you need rather than a single product set.
Measure the contact
Volume and cost by channel and by driver, expressed per shipped order rather than in totals.
Separate cause from symptom
What can be removed at source, what can be resolved automatically, and what needs a person.
Design the architecture
Connected to live order data, with an exit to a person at every point.
Test before peak
Load tested against your forecast, then measured through the trading window and after it.
Three ways in, depending on how far along you are
The conversation scorecard
Ten questions on where your contact comes from, what it costs by channel, and what happens when volume triples. You get a score, a benchmark against retailers of similar size, and your two weakest areas.
- What share of your contact is order status
- Whether you measure contact rate per shipped order
- What happens to a call at eight in the evening in December
- How a return gets started today
The Customer Conversation Review
We measure what contact actually costs you by channel and by driver, then separate what can be removed at source from what should be automated. Week one is discovery. Week two is analysis. Week three is the readout.
- Cost per contact and cost per resolution across every channel
- Your order status volume traced back to its operational causes
- What can realistically be in place before the next peak
- A 90 day plan you can execute with us or with anyone else
The fee is credited in full against the first implementation phase if you proceed within 90 days. The findings are yours outright and portable to any provider.
Book a scoping callProcess specific AI for regulated firms
How to scope and frame discovery before any technology decision, including the tractability questions to work through first.
- Which processes are worth automating and which are not
- What to establish before a shortlist exists
- The governance questions that stall these projects
Frequently asked questions
What share of retail contact is order status?
Roughly 25 to 40 per cent in a normal trading week, climbing past 50 per cent during peak. It is usually the single largest ticket category a retailer has, which means the contact centre ends up sized around a question the operation could answer for itself.
Is deflection the same as resolution?
No, and confusing the two is the most expensive mistake in this category. Deflection removes a contact from a queue. Resolution completes the customer's task. A deflected customer usually rings you, and phone is the channel that costs the most per resolution once repeat contact is counted.
Will this work with our ecommerce and helpdesk stack?
In most cases yes. The layer needs live access to order data, carrier tracking and your returns policy, which the major platforms and helpdesks expose through APIs. The harder question is usually the warehouse system and the carrier integrations rather than the storefront, and establishing that is part of the review.
Can we get something live before peak?
It depends when you start. Order status resolution connected to live data is achievable on a short runway if the data is accessible. Anything involving warehouse or carrier integration work needs longer. A retailer deciding in November is realistically planning for the following year, and we will say so rather than promise a rushed build.
Do we still need seasonal agents?
Usually yes, but fewer and for different work. The aim is to leave your people with the contacts that need judgement, such as complaints, high value orders and anything where a customer has already been let down once, and to stop recruiting for a question a system can answer.
The cheapest contact is the one that never happens. Find out how many of yours qualify.
Three minutes for a read on where you stand, or a scoping call if you want us to measure it properly.