Deflection vs. Resolution: Why the Words You Choose Matter
Deflection avoids a person; resolution answers the customer. Buyers should know which they're buying.
Two vendors can quote the same impressive number and mean opposite things by it. One counts the customer who got their answer. The other counts the customer who gave up. Both call it a success, and both put it on the same slide. If you're buying support software, the gap between those two definitions is the most expensive thing you'll skim past.
This is the difference between deflection and resolution. They sound like cousins. They're not. One measures whether a person was avoided; the other measures whether a customer was actually helped. Knowing which one a vendor is selling you tells you almost everything.
Deflection vs resolution: avoided vs. answered
Deflection is a count of contacts that didn't reach a human. The customer hit a help article, a chatbot, or an automated reply, and never opened a ticket. That's the whole bar: no human touched it. Notice what's missing — nobody checked whether the customer's problem went away.
Resolution is a count of customers whose problem was actually solved, end to end. The order status was found. The refund went through. The address got changed before the package shipped. The customer left with the thing they came for, not just without a ticket.
Here's why the distinction is load-bearing: a customer who reads three help articles, finds none of them useful, mutters something unkind, and abandons the chat counts as deflected. From the dashboard's point of view, that's a win. From the customer's point of view, it's the moment they started shopping for a competitor.
A high deflection rate can mean your customers got help, or it can mean they gave up. The number alone can't tell you which.
How a deflection stat hides a resolution problem
Deflection is seductive because it's easy to inflate and hard to falsify. You can drive it up with tactics that have nothing to do with helping anyone:
- Bury the contact button so fewer people find their way to a person.
- Add friction — a long form, a forced bot loop — so people abandon before they reach the queue.
- Auto-close anything that goes quiet, and call the silence a deflection.
- Answer the easy questions confidently and the hard ones vaguely, so the hard ones wander off.
Every one of those moves makes the deflection chart go up and customer experience go down at the same time. That's the trap. The metric and the goal point in opposite directions, and the metric is the one on the report.
| Deflection | Resolution | |
|---|---|---|
| What it counts | Contacts that didn't reach a person | Problems actually solved |
| Customer state | Unknown — helped or gave up | Confirmed helped |
| Easy to inflate? | Yes (hide the button, add friction) | No — the problem either resolved or didn't |
| Failure mode | Looks great while customers churn | Honest, sometimes uncomfortable |
| Right question | "Did we avoid a human?" | "Did the customer get what they needed?" |
The tell is whether a vendor can connect the metric to an outcome. Ask: of the contacts you deflected, how many came back within a week? How many escalated to a worse channel — a one-star review, a chargeback, a cancellation? A vendor selling real resolution can answer. A vendor selling deflection will change the subject to volume.
What to ask before you trust the number
You don't have to be cynical, just specific. Before you accept any "we handled X% automatically" claim, get the definition behind it.
- What exactly counts as a success here? No-human-touched, or problem-confirmed-solved? Make them say it out loud.
- How do you measure repeats? A real resolution rate has to net out the customers who came back unhappy. If reopens aren't subtracted, the number is fiction.
- What happens to the hard ones? A system that quietly drops or stalls the cases it can't handle is manufacturing deflection. A system that hands them to a person — with full context — is doing the honest thing.
- Can you show the receipts? For a sample of "resolved" contacts, can you see what the AI actually did and verify the customer was helped?
That last point is where the two worlds separate cleanly. Deflection asks you to trust an aggregate. Resolution lets you audit a case. If every AI action leaves a record you can open and read, you can stop arguing about definitions and just look.
Measure the thing you actually want
The fix isn't to ban the word "deflection." It's to stop treating it as the goal. Deflection is a means; resolution is the end. Track both, but report against resolution — and net it for reopens, so a customer who came back angry doesn't get counted as a win twice. If your contact volume drops because customers genuinely don't need you, wonderful. If it drops because they've given up on you, you want to know that before it shows up in churn.
This is why BearScope is built around resolution, not avoidance. Your people and AI agents work conversations to an actual answer on one governed system, every conversation is scored so a quiet abandonment can't masquerade as a win, and every AI action leaves a receipt you can audit to confirm the customer was really helped. See the approach in the product overview, check how plans are structured on the pricing page, or book a walkthrough.
See it on your own conversations.
Bring your busiest day. We'll score every conversation in it.
Book a walkthrough →