Ticket Deflection vs Resolution: What Actually Matters

If you’ve shopped for support tools lately, every pitch leads with a deflection number. “Deflect 60% of your tickets.” It sounds like savings. But ticket deflection in ecommerce only counts a request that never reached a human, and that is not the same as a customer whose problem got solved. A deflected ticket can be a happy customer who got their tracking link, or a frustrated one who gave up and is now writing a one-star review. The number on the dashboard looks identical either way. This post is about the gap between those two outcomes, what the real benchmarks are, and why optimizing for deflection alone quietly damages the thing you actually care about.

Deflection and resolution are not the same metric

Deflection measures volume that didn’t reach you. Resolution measures problems that got fixed. They sound similar and they are constantly confused, but they answer different questions.

Deflection rate: of all incoming requests, how many were handled before a human touched them (self-service article, chatbot, automated reply). Counted at the front door.
Resolution rate: of the requests an automated system handled, how many actually closed the customer’s issue with no follow-up. Counted at the outcome.

The trap is that a deflection can be counted as a win the moment the ticket doesn’t land in your inbox, regardless of whether the customer got what they needed. A click on a help-center article is not a fix. A closed chat window is not a fix. As one benchmark put it bluntly: the gap between self-service attempts and self-service resolutions is wider than most owners assume.

The benchmarks worth knowing

Rough numbers from 2025-2026 CX data, useful as guardrails rather than targets to chase:

Metric	Typical	Good	Stretch
Ecommerce deflection, no AI	15-30%	30%	-
Ecommerce deflection, with AI	40-65%	60%	70%+ (high order-status volume)
Median tier-1 deflection (cross-industry)	~41%	~59% (top quartile)	-
AI resolution rate (ticket closed end-to-end)	~45% median	60-65%	-

Two things stand out. First, median tier-1 deflection sits around 41%, so if a vendor promises you 70% out of the box, ask what’s being counted. Second, the median brand resolves only about 45% of AI-touched tickets all the way through, with the top quartile around 65%. A realistic resolution target is 60-75%. The space between “deflected” and “actually resolved” is where customer trust leaks out.

Why chasing deflection backfires

Push deflection too hard and you start blocking people from reaching you. That is easy to do and it shows up in the satisfaction numbers.

The pattern in the data: CSAT trends negative once deflection climbs past roughly 80%. Beyond that point you are usually deflecting tickets that should have reached a human, by hiding the contact form, forcing a bot loop, or auto-replying to messages the automation didn’t understand. Escalation data makes the cost concrete: satisfaction drops by about 22 points when a customer has to escalate after a failed first attempt (roughly 89% CSAT on a clean resolution vs 67% once they’ve been bounced). Every forced deflection that fails becomes a second, angrier contact, and now you’ve spent more effort than if you’d answered it straight the first time.

So the failure mode is specific: a high deflection rate with a low resolution rate means you’re not saving work, you’re deferring it and adding frustration on top. The customer still needs an answer. They just get it later, after they’ve had a worse experience.

What to measure instead

Deflection is fine as one input. It’s a bad primary goal. A better setup:

Lead with resolution rate. Of the messages your automation answered, how many closed without the customer replying again or escalating? That’s the number that maps to a solved problem.
Pair every deflection number with CSAT. Deflection without satisfaction is just volume you pushed out of sight. If deflection rises and CSAT falls, you’re hiding tickets, not handling them.
Watch the re-contact rate. If “deflected” customers keep coming back within a day or two, those weren’t resolutions. They were delays.
Segment by type. WISMO (“where is my order?”) is the one category where high automation and high satisfaction genuinely coexist, because the answer is a specific, lookup-able fact: order status and a tracking link. Address changes before fulfillment are similar. Refund disputes and damaged-item complaints are not, and forcing automation onto them is where CSAT tanks. If you want the mechanics of doing the easy category well, the complete WISMO automation guide and auto-answering “where is my order?” emails both go deeper.

The healthy version of automation isn’t “answer fewer tickets.” It’s “resolve the answerable ones correctly, and route the rest to a human fast.” That’s also why a draft-and-approve model holds up well for small stores: the system pulls the real order data and writes the reply, you glance at it and send. The ticket gets resolved, not just deflected, and a person stays in the loop on anything sensitive. That’s the approach behind LzyReply, though the principle holds whatever tool you use.

The takeaway

Deflection tells you how much volume you avoided. Resolution tells you whether your customers are okay. For a small Shopify store, the second one is what keeps reviews positive and repeat orders coming, so make it your headline metric and treat deflection as a side effect of doing the answerable tickets well. If a number ever forces you to choose between deflecting a ticket and solving it, solve it. The math works out better every time.