What is contacts per order (CPO)?

Contacts per order (CPO) is the number of customer service contacts divided by the number of orders placed over the same period. It shows how often a purchase turns into a support conversation, and it is one of the clearest volume-side signals of whether fulfillment, product information, and self-service are working.

Cost per contact measures what a conversation costs. First contact resolution measures whether it got solved. Contacts per order measures something upstream of both: whether the order itself needed a conversation at all. An order that ships correctly, tracks clearly, and arrives as described rarely generates a contact. One that does not will generate one, and often more than one.

This page covers how to calculate contacts per order, what typically drives it up, and how to bring it down without making support harder to reach. It also covers where the metric misleads.

Contacts per order in one sentence

Contacts per order is how many times, on average, a customer gets in touch about a single purchase.

How to calculate contacts per order

Contacts per order = total contacts for the period ÷ total orders processed in the period

A brand handles 1,000 contacts against 250 orders during a holiday sales period.

Contacts per order = 1,000 ÷ 250 = 4

Four contacts for every order is high. The direction to move is down, toward 1 and eventually toward 0, where most orders never need a conversation at all. Reaching zero across an entire catalog is rarely realistic, since some products are naturally higher-consideration or higher-complexity than others, but the trend matters more than the absolute number.

What counts as a contact, and what counts as an order

The formula is simple. The definitions underneath it are where most of the disagreement between teams happens.

Order definitions vary. A single checkout with three items is usually one order. A return that generates a replacement shipment, or a subscription renewal, can get counted as a second order, depending on how the commerce platform tags it. If the order count quietly inflates from platform-side splitting, the ratio improves for reasons that have nothing to do with the customer experience.

Contact definitions vary more. A customer who emails, then follows up in chat, then calls about the same undelivered package can be logged as one contact or three. The count depends on the platform. Ticket-based systems tend to open a new record for each new channel touch. Conversation-based systems tend to recognize the thread as one ongoing issue.

This has a direct effect on the ratio. Two operations with identical orders, identical costs, and identical customer experience can report different contacts per order numbers, purely because one counts touches and the other counts conversations. Before comparing the ratio across teams, time periods, or channels, confirm both definitions are held constant.

How contacts per order compares to related metrics

Contacts per order is often confused with metrics that measure something adjacent, particularly cost per contact and first contact resolution. Each answers a different question.

Read together, they tell a fuller story than any one of them alone. A rising contacts per order ratio paired with a falling cost per contact often means contacts are being resolved faster but not prevented. A falling contacts per order ratio paired with falling first contact resolution can mean fewer customers are reaching out successfully, not that fewer of them have a problem.

What drives a high contacts per order ratio

The volume almost always traces back to a small set of root causes. They cluster into a few families.

Order status and delivery uncertainty. Unclear tracking, delayed shipments, and backordered items are the single most common driver. Customers who do not know where their order is will ask. They ask again if the first answer isn't clear.

Fit, sizing, and product-description mismatches. For retail and apparel especially, a product that arrives different from what the listing implied generates a contact almost every time. The return or exchange contact usually follows it.

Policy and process friction. Returns processes that require a phone call, exchange rules that are not stated clearly at checkout, or refund timelines that are not communicated proactively all convert routine questions into support contacts.

Unresolved repeat issues. A contact that does not fix the underlying problem generates a second one. This is the multiplier effect. A single bad fulfillment experience can show up as two, three, or four contacts against the same order.

The most useful way to work with this list is to categorize contacts by the customer's underlying problem. "I never received my package" is a category that points at a fix. "Processed a refund" describes an agent action and points at nothing. Sorting contacts by customer intent, with a manageable handful of categories, is what turns a ratio into a list of things to fix.

It also helps to look at this over a longer window than a single peak period. Contacts per order during a holiday sale spike will look different from the annual average. The spike doesn't mean quality changed. Month-over-month or year-over-year comparisons are more reliable than any single snapshot.

How to reduce contacts per order

The interventions that work are the ones that remove the reason for the contact.

Fix fulfillment accuracy first. Reliable, trackable shipping and clear proactive updates on delays, backorders, or substitutions prevent the largest single category of contacts before they start. A customer who already knows about a delay rarely calls to ask.

Expand self-service for the predictable, high-volume questions. Order status, return initiation, and address changes are asked constantly and answered the same way almost every time. Automation and self-service can resolve these instantly. Customers often prefer the instant answer over waiting for a person. The condition that matters is that the answer resolves the question. A contact that gets automated without being resolved does not go away. It comes back, usually through a more expensive channel the second time, and pushes the ratio higher.

Organize support around the conversation. A support system that opens a fresh record every time a customer switches from email to chat to phone fragments one order problem into several counted contacts. The person on the third channel has to start over. Recognizing the full thread as one conversation, with the order and customer history already visible, is what lets the third touch close the issue.

Give the team full context before the first response. Time spent asking a customer to repeat their order number, their issue, or what they were already told is contact volume with no service value in it. It also increases the odds the first contact does not resolve the issue, which brings the customer back for a second one. This is where contacts per order and first contact resolution reinforce each other: fixing one tends to improve the other.

Be careful with anything that reduces contacts by reducing access. Removing a phone number, burying a contact link, or routing everything through a slow-to-respond bot can lower the reported ratio. It does not lower the number of customers who have a problem. It shows up elsewhere: in reviews, in chargebacks, in customers who simply do not come back. A ratio that improves because customers gave up contacting you is not a win.

What contacts per order does not tell you

The metric has a structural blind spot worth knowing before it goes into a scorecard.

It does not weight for severity. A damaged, high-value order and a routine "where is my package" message both count as one contact. The business stakes are very different. A ratio can hold steady while the mix underneath it gets worse.

It does not account for order or customer value. A contact tied to a twenty-dollar order and one tied to a five-hundred-dollar order are identical in the formula. Segmenting the ratio by order value or customer tier usually reveals more than the blended number does.

It can be gamed by making support harder to reach. As noted above, a lower ratio achieved by discouraging contact is not the same as a lower ratio achieved by preventing problems. Pairing contacts per order with first contact resolution and a satisfaction metric like CSAT or NPS is the check against this. If the ratio is falling while satisfaction is falling too, something other than genuine improvement is driving it.

Read alongside those other metrics, contacts per order is a strong early signal of where fulfillment and product information are breaking down. Read alone, it rewards the same shortcut cost per contact does: it treats a hidden contact channel and a solved problem as the same result.

Contacts per order and automation

Automation changes this ratio more than almost any other lever available. The change needs the same care as any other reduction strategy.

The direct effect is real. Order status checks, tracking questions, and return initiations are exactly the high-volume, low-complexity contacts that automation handles well. Resolving them without a person removes them from the numerator immediately.

The distinction that matters is between routing a contact away and resolving it. A deflection rate can rise while contacts per order stays flat or worsens, when the contacts being deflected are not being answered. They are simply delayed until the customer tries again through a different, usually more expensive, channel. The honest version of this improvement shows two things moving together: fewer contacts overall, and resolution happening on the first attempt regardless of whether that attempt was automated or human.

The other piece automation cannot fix on its own is the root cause. An order that consistently confuses customers about sizing, or a fulfillment process that regularly runs late, will keep generating contacts. It doesn't matter how efficiently those contacts get handled. Automation lowers the cost and speed of the contact. It does not lower the number of customers who had a reason to reach out. Only fixing the underlying order experience does that.