Intercom Fin resolution rate
What Intercom counts as a resolution, what that number leaves out, and how to check the same conversations against your own definition.
Intercom reports three numbers for Fin. Involvement rate is how often Fin tried. Resolution rate is how often it resolved, out of the ones it tried. Automation rate is the two multiplied.
Resolution rate is the one quoted in meetings, and it is the one with a definition worth reading before you act on it.
What Intercom counts as resolved
Two things, and only one of them involves the customer saying so.
| Confirmed | The customer answers something affirmative after Fin's last answer. |
|---|---|
| Assumed | The customer stops replying and does not come back asking for more help. |
Fin's own help centre puts the silence needed for an assumed resolution at 24 hours. Intercom's copy of the same article states the rule without a number, so take the figure as Fin's.
Intercom does guard the definition at one end: a reply to a greeting is not an answer, so it cannot be a resolution.
What the number cannot tell you apart
An assumed resolution is silence, and silence has more than one cause. Intercom's documentation is direct about this.
- A customer who read the answer, was happy, and left.
- A customer who gave up and went to a competitor.
- A customer who went quiet after a teammate joined, having never asked for a person.
- A conversation a teammate closed by hand, where Fin never escalated.
All four land in the same bucket. The one documented exception is frustration Fin detects itself, which escalates and is not billed.
The definitions moved
Intercom changed what counts as Fin being involved, which lifted resolution rate without any agent changing. Comparable history starts at 1 October 2025.
Existing custom reports kept the old definitions. One workspace can print two different numbers for one period, from two screens, both correct.
Measuring the same conversations yourself
Evidova reads the Intercom conversations you already have and scores each one against rules you approved, rather than against disengagement.
Connect Intercom in Setup.
We poll for what changed and read each conversation in full, so nothing about Fin changes.
Import the history you already have.
The same rules run over past conversations, so you get a baseline on day one instead of in a month.
Compare a week of ours against the same week of Intercom's.
The gap is the set of conversations that ended in silence for a reason nobody wanted.
What the two numbers are each good for
| Intercom resolution rate | How much work Fin took off the queue, and what it costs. |
|---|---|
| An Evidova score | Whether the answer was right, against a rule you wrote and can change. |
| Where they disagree | Conversations that closed without anyone being helped. |
Intercom does measure answer quality, through the CX Score reason named for it. It needs the Pro add-on, it covers chat and email but not phone, and it rates the experience from one to five rather than checking an answer against a source.
Every score we produce opens onto the turn it came from, so a number you doubt is a conversation you can read.
Read next
- Quickstart: connect a platform · Ten minutes, one path, from a webhook to your first score.
- Webhook reference · The URL, the signature, and what we answer a delivery with.
- All twelve platforms · Every voice and chat stack Evidova reads, side by side.