Supportman (@supportmanHQ) header
Supportman (@supportmanHQ) profile picture

Intercom CSAT, AI scoring + weekly reports, in Slack. What separates a 1-star reply from a 5-star one. #custserv #cx

Joined February 2020
Photo by Supportman (@supportmanHQ)Photo by Supportman (@supportmanHQ)Photo by Supportman (@supportmanHQ)Photo by Supportman (@supportmanHQ)Photo by Supportman (@supportmanHQ)Photo by Supportman (@supportmanHQ)Photo by Supportman (@supportmanHQ)Photo by Supportman (@supportmanHQ)Photo by Supportman (@supportmanHQ)Photo by Supportman (@supportmanHQ)
Supportman puts your Intercom CSAT in Slack, in real time. Bad ratings get their own channel. Friday reports for the team, DMs for each agent. And AI QA that scores every closed conversation — including the ones customers never rated.
1
2
160
Four different clocks get called "resolution time." Most teams report the easiest one to pull, which is time to close. That's how fast the queue moves. It says nothing about whether the problem went away.
1
10
From the field note, made-up numbers: Team A closes in 2h median, Team B in 5h. Within 7 days, 36 of Team A's 200 customers are back on the same issue; 12 of Team B's. Team A handles 24 more conversations to solve the same 200 problems.
1
1
Priscilla Brooke (Buzzsprout) on where her team's QA rubric came from: "So we came up with educational, accurate, clear, and human." Four scales, written by the six people who get graded on them. supportman.io/live-chat/tone…
1
1
24
Stacy Justino (PetDesk), same shape, different words: "So our four criteria are accuracy, completeness, customer excellence, and empathy/tone." Empathy made the list because their customers are vets having a hard day. supportman.io/live-chat/ten-…
1
3
Neither rubric came from a vendor template. Both have four criteria, one of them specific to the company, and the team helped write it. Brooke: "And as a team we developed that rubric, because again it's all about buy-in."
5
A Fin resolution "came back" when the same customer writes in again about the same problem. Intercom reports the resolution. It only reports the return if they reply in the same thread. Most don't.
1
13
Worked example from the field note, made-up numbers: 400 Fin resolutions, 46 came back inside 7 days. 9 in the same thread, 37 as new conversations. 80% of the returns invisible to a reopen report, none of them refunded.
1
4
Chloé Koers-Bourrat (Smartly) was reviewing 100 to 120 chats a week on a QA platform. Now: "I'm not reviewing more than 10, maybe 20 chats per week. The drop is huge, and we're aware of this." supportman.io/live-chat/prev…
1
5
Amanda Drws, CX analyst, from the other side of the same decision: "A lot of it is really just stress soothing. People are anxious. It's an anxiety tax." Her rule: QA at 10% or less of tickets. supportman.io/live-chat/stop…
1
6
Both cut review volume on purpose and say the program got better for it. Fewer reviews, read properly, with feedback the agent can act on. Two Supportman Live Chat episodes, links above.
2
Intercom has two features called "topics." They don't talk to each other, and mixing them up wastes setup time.
1
21
Fin's Topics Explorer: AI-generated, zero setup, shows what Fin thinks customers are asking about right now. A separate system entirely.
1
3
Fin doesn't need the same QA rubric as a person. It needs its own, because it fails in different ways.
1
17
Run it weekly on a sample, not once a quarter on everything. That's the difference between catching a regression and reading about it in a churn interview.
1
9