parley
DocsPlatform
Resources
JournalIdeas on running AI supportCustomer storiesHow teams use ParleyGuidesStep-by-step playbooksChangelogWhat shipped, by versionHelp centerAnswers and support
CustomersPricing
LoginGet started
DocsPlatformCustomersPricingJournalCustomer storiesGuidesChangelogHelp center
Get startedLogin

Take charge of your AI customer experience

Get startedTalk to sales

Platform

Platform overviewChannelsPlaybooksIntegrationsAnalyticsPricing

Resources

DocumentationGuidesCustomer storiesJournalChangelogHelp center

Company

AboutCareersContact salesContact supportTerms of servicePrivacy

Security

Trust centerSOC 2ISO 27001GDPRHIPAAData processing
parley

© Parley. All rights reserved.

Journal · Measurement

Measure resolution, not deflection: a CX scorecard your finance team will trust

Deflection counts the customers you kept away from people. Resolution counts the problems you solved. Only one of them shows up on the P&L.

LMLeon MarshCX Analytics Lead · 7 min read

Deflection rate is the most quoted number in AI support and the least trusted in the finance office. It counts conversations that ended without a person. It does not count whether the customer's problem went away, or whether they came back an hour later on the phone.

A scorecard that finance will trust starts from resolution and follows the money.

Define resolution before you launch

A conversation is resolved when the customer's goal is met and they do not contact you again about the same thing within a set window. Pick the window per journey: a day for order tracking, a week for refunds. Agree it with finance before the first conversation, not after the first good month.

The five numbers

  • Resolution rate: resolved conversations divided by all conversations the agent started
  • Repeat contact rate: customers who came back about the same issue inside the window
  • Cost per resolved conversation: platform cost plus human time on handoffs, divided by resolutions
  • Handoff quality: satisfaction for handed-over conversations, measured on its own
  • Revenue kept: refunds avoided by offering an exchange, cancellations saved, orders completed in the conversation

Each one has a clear owner, and each one can get worse when the others improve. That tension is the point.

Count the conversations you did not finish

An agent that resolves 70% of the conversations it chooses to start can look better than one that resolves 60% of everything. Report resolution against all incoming volume on a journey, not just the slice the agent took.

Report resolution against all incoming volume, not just the slice the agent took.

Make it auditable

Finance will ask how a conversation was counted. Every number on the scorecard should click through to the conversations behind it, with the trace that shows what the agent did. In Parley, the analytics view links each metric to its conversations so a sceptical reviewer can sample them in a minute.

Review monthly, change weekly

Share the scorecard monthly with finance and leadership. Inside the CX team, look at it weekly and tie every change to the agent to the number it is meant to move. If nobody can say which number a change was for, it probably should not ship.

LMLeon MarshCX Analytics Lead, ParleyAll posts

Keep reading from the journal

All posts
Knowledge6 min read

Writing support policies an AI agent can follow word for word

Read
Voice5 min read

Writing for the ear: what changes when your agent picks up the phone

Read
Engineering6 min read

Rehearse before you ship: using simulations as a release gate

Read

Try it on your own conversations

Start building an agent in minutes, or talk to us about your support volume.

Get startedTalk to sales