Cost Per Ticket: Ecommerce Benchmarks and How to Lower It (2026)

How much does a support ticket cost?

A support ticket in ecommerce costs roughly $5 to $20 to resolve when a human handles it, and under $1 when self-service or an AI agent resolves it first. The cost per ticket that actually lands in your P&L is a blend of the two, and it moves almost entirely on one variable: how many contacts reach a person versus how many get answered before they do.

Retail and ecommerce sit at the lower end of the cross-industry range. Benchmark data for 2026 puts retail ecommerce around $2.70 to $5.60 per ticket on a blended basis, well below SaaS ($18-$35) or B2B ($30-$60), because so much ecommerce volume is simple and repetitive: where is my order, how do I return this, did my refund go through. That mix is both the reason ecommerce CPT is naturally lower and the reason it is so easy to lower further.

Cost per ticket matters because it is the one support metric that translates directly into money. Resolution rate and CSAT tell you how well you are doing. Cost per ticket tells you what that performance costs, and it is the number a finance team will actually use to decide whether you can hire, whether to buy software, and whether support is a cost center or a margin you can defend during peak season.

How to calculate cost per ticket

The formula is simple: total monthly support spend divided by total tickets resolved that month. The mistake most teams make is using too narrow a numerator. Agent wages are the largest line, but they are rarely more than 70-80% of the real cost. Leave out tooling, management time, and overhead and you will understate CPT by a third and overstate the savings from any change you make.

Pull a clean numerator first. Add up everything support actually consumes in a month, then divide by every ticket that closed, including the ones a help center article or an AI agent handled end to end. Those resolved-without-a-human contacts belong in the denominator even though they barely touch the numerator, which is exactly why adding them pulls the blended number down.

Pick a consistent period and stick to it. A single month is fine for a stable store, but if your volume swings with promotions or seasonality, average three months so one quiet week or one BFCM spike does not distort the figure. The same rule applies to labor: prorate any agent who splits time between support and another function rather than counting their whole salary, or you will inflate the numerator with hours that never touched a ticket.

• Agent labor: salaries, benefits, and payroll taxes for support staff, prorated by the share of time spent on support
• Platform costs: help desk, live chat, AI subscription, telephony, and any per-seat add-ons
• Management overhead: team lead and manager time spent on coaching, QA, and escalations
• Training and onboarding: amortized over average agent tenure
• Infrastructure and integrations: middleware, custom development, and data tooling that support depends on

Cost per ticket benchmarks (2026)

The global blended average across channels and regions runs about $8 to $12 per ticket, with North America at the high end (roughly $15 to $20) and Asia-Pacific lower (around $5 to $10), reflecting wage differences more than any difference in efficiency. Ecommerce specifically benchmarks below the cross-industry average because of its simple ticket mix.

Use the ranges below as a sanity check, not a target. Your real CPT depends on agent wages in your region, your channel mix, your ticket complexity, and how much volume you deflect before it reaches a person. A three-person team in a high-wage market will run a higher CPT than the same team in a lower-wage one, even at identical quality.

Where you sit in these ranges says more than the absolute number. Landing at the top of your bracket usually points to one of two things: low volume spreading fixed costs thinly, or high handle time from complex tickets and weak tooling. Both are fixable, and knowing which one is yours tells you whether the answer is more automation, better content, or a different team shape entirely.

Cost per ticket by channel

Channel choice is one of the largest cost drivers you control. Phone-heavy teams consistently pay two to three times the per-ticket cost of chat-first teams handling the same issue types, because voice is strictly one-to-one and ties up an agent for the full length of the call. Chat lets one agent run several conversations at once, and self-service has almost no marginal cost at all.

The table below reflects 2026 benchmark ranges per contact, including the re-contact cost that voice tends to carry because of its lower first-contact resolution. The pattern is consistent across studies: every step from phone toward self-service takes cost out, which is why routing simple questions to chat and a help center before they reach a phone queue is one of the highest-leverage moves in support.

Concurrency is what makes chat and messaging cheap. A phone agent handles one customer at a time; a chat agent can run three or four conversations at once, and an AI agent handles effectively unlimited concurrent conversations with no extra labor at all. That is why social and messaging channels (WhatsApp, Instagram, Facebook Messenger) land near chat economics despite feeling like a heavier lift, and why peak-season volume that would require a phone hiring spree can often be absorbed on chat with the headcount you already have.

What drives cost per ticket up

Cost per ticket is mostly labor, so anything that adds agent minutes or removes leverage pushes the number up. For most ecommerce stores, two or three of the drivers below dominate. Finding which ones apply to you is the difference between a generic cost-cutting project and a targeted one that actually moves the number.

How AI lowers cost per ticket

AI lowers cost per ticket through one mechanism: deflection. Every contact an AI agent resolves on its own removes the labor cost of a human handling it, while adding only a few cents in AI cost. Because labor is the bulk of CPT, even modest deflection moves the blended number, and the effect compounds as deflection climbs because each resolved ticket you remove is a full human cost saved.

The model below uses illustrative figures, a $15 human CPT and $0.50 per AI contact, to show the shape. Each point of deflection is worth roughly $0.14 of blended CPT reduction in this example. Note what the structure rewards: the highest-volume, simplest ticket types (WISMO, return status, refund timing) are exactly the ones an agent resolves most reliably, so deflection tends to land on your cheapest-to-automate tickets first.

Deflection also ramps rather than arriving all at once. A freshly connected agent might resolve 30-40% in its first weeks; as you feed it better help docs, connect order data, and let it learn from real conversations, that rate climbs. The CPT curve is steepest at the start, so even a conservative early deflection number produces savings while the agent improves toward the higher end of the range. The compounding works in your favor: the simplest tickets go first and cheapest, leaving humans for the complex, high-AHT cases where their judgment is actually worth $15.

Why cheap tickets can still be expensive

A low cost per ticket is not automatically a good thing. CPT measures what you spend to close a contact, not what that contact cost you in lost sales or lost loyalty. Two stores can both run a $6 CPT while one keeps customers and the other bleeds them, because the number says nothing about resolution quality, response time, or whether the ticket should have existed at all.

The cheapest ticket is the one a customer never has to open. A clear shipping notification that prevents a WISMO question, a returns policy a customer can self-serve, a product page that answers the pre-sale question before it becomes a ticket: none of these show up as a low CPT, because they show up as no ticket at all. That is why CPT should always be read next to ticket volume and resolution quality, never on its own.

There is a real failure mode here. A team chasing a CPT target can hit it by deflecting aggressively and answering fast but badly, and the cost simply moves somewhere the support dashboard does not see: a refund that should not have happened, a customer who does not come back, a one-star review. None of those land in cost per ticket, but all of them are more expensive than the ticket would have been. Optimize CPT as a budget number, not as a quality goal, and let resolution rate and CSAT keep it honest.

• A slow or wrong answer is cheap on CPT but expensive in lost repeat purchases and refunds
• Deflecting a ticket by frustrating the customer into giving up is a false saving
• Preventing tickets (proactive notifications, better content) beats resolving them cheaply
• Read CPT alongside resolution rate, CSAT, and first response time, never alone

Build your own cost model

Before you invest in AI support or extra headcount, build a simple cost model. It takes about 20 minutes and gives you a defensible number to bring to a finance conversation, instead of a vendor's marketing figure. The point is not precision to the cent; it is a realistic, transparent estimate you can stand behind.

1. 1Calculate your current CPT: total monthly support spend (labor + tooling + overhead) divided by monthly tickets resolved.
2. 2Estimate a realistic deflection rate. Use the benchmark table above as a starting point; 40-55% is reasonable for most ecommerce stores, higher if you have heavy WISMO and returns volume.
3. 3Project the new blended CPT: (deflection rate x AI cost per contact) + ((1 - deflection rate) x human CPT).
4. 4Multiply the CPT reduction by monthly ticket volume to get monthly gross savings.
5. 5Subtract the AI platform cost to get net monthly savings, then divide the platform cost by net savings for a payback period.
6. 6Add the benefits CPT does not capture: 24/7 coverage, instant first response, lower repeat-contact rate, and any revenue the agent influences through recommendations.

Mistakes that distort the number

Most bad CPT figures come from a handful of repeatable errors. Each one either flatters the number or panics a stakeholder, and all of them are avoidable once you know to look for them. The thread running through every one is scope: which costs you count, which contacts you count, and which benchmark you compare against. Get those three right and the number becomes trustworthy enough to build a budget on.

• Counting only wages in the numerator, which understates true CPT by roughly a third
• Leaving AI- and self-service-resolved contacts out of the denominator, which makes AI look like a pure cost instead of a deflection
• Mixing channels into one average and hiding an expensive phone queue inside a healthy-looking blend
• Comparing your CPT to a cross-industry benchmark instead of an ecommerce one and concluding you are doing great when you are average
• Treating a per-resolution vendor price as your CPT; that pricing model punishes you for resolving more, which is the opposite of what you want

How Bookbag changes the math

Bookbag is an AI customer support agent built for ecommerce, and it attacks cost per ticket on the input side: it resolves the high-volume, simple tickets that make up most of an ecommerce queue, so fewer contacts ever reach a person. It tracks orders, processes returns and refunds within the rules you set, answers pre-sale product questions, and recommends products, then hands off to a human with full context when a ticket genuinely needs one. Merchants typically see up to around 70% of tickets resolved autonomously, which is the deflection that pulls blended CPT down.

The pricing is built so deflection works for you, not against you. Bookbag uses flat monthly plans with a message-credit allowance (one credit is one AI reply) and a spend cap you set, not per-resolution fees. There is no success penalty: when your agent resolves more tickets, your cost stays predictable and your blended CPT keeps falling. That is the opposite of per-resolution models, where every win adds to the bill.

Bookbag is not the cheapest help desk on the market, and if your volume is tiny the math may not yet favor any automation. But for a store with real, repetitive volume across channels, moving the simple tickets to an agent that actually resolves them is the most direct lever on cost per ticket there is, and it adds 24/7 coverage and instant first response that headcount alone cannot buy.