A support-wins channel sounds simple: a customer leaves a positive rating, the rating appears in Slack, everyone celebrates. Done well, it makes invisible support work visible and turns customer praise into something the team learns from. Done badly, it becomes a scoreboard where the same three names appear every week and everyone else quietly wonders whether their work counts.
We have already made the case for celebrating support wins — the retention and engagement evidence lives in that article. This one is the build guide: how to set the channel up, what to automate, and the design decisions that keep a recognition feed from mutating into a leaderboard.
Supportman pipes positive Intercom ratings into a Slack channel automatically — a support-wins feed with zero manual posting.
Decide what the channel is for
Write a one-sentence purpose before creating the channel:
This channel shares customer feedback, celebrates specific support behaviours, and helps the team learn what customers value.
That sentence rules out several tempting uses:
- Ranking agents by rating count
- Pressuring teammates to compete for praise
- Posting private customer information unnecessarily
- Treating a positive CSAT as proof of flawless work
- Using silence in the channel as evidence of poor performance
The last one deserves emphasis. CSAT response rates are often in the single digits and vary sharply by channel and queue, so an agent's absence from the feed usually says more about where they work than how well they work.
Name the channel something obvious — #support-wins does the job — and make membership broad enough that the people who contribute behind the scenes (the macro author, the escalation engineer, the docs writer) can see it and appear in it.
Automate the moment
Recognition is most useful while the work is still fresh, and manual posting fades the first week the queue gets heavy. Automate the feed so appreciation never depends on a manager remembering to check a report.
Keep the post itself small: the rating, the customer's remark, the assigned agent, a link to the conversation, and the time. In practice a post looks like this:
★★★★★ from a conversation with Harbourline Freight
“Priya stayed with me through the whole SSO mess and explained every step in plain English. Best support I have had this year.”
Agent: Priya · View conversation · 10:42
Limit customer identifiers to what the team needs, restrict channel access, and follow your data-handling rules.
Three decisions to make before you switch it on:
- Which ratings get posted? Start with every positive rating, remark or no remark. Filtering to remarks-only makes each post richer, but on a team where one rating in ten includes a comment it can starve the channel — and it starves the quieter queues first.
- Where do middling and negative ratings go? Not here. Route them to an operational channel designed for response, with an owner and a follow-up habit. Mixing incident response into the celebration feed muddies the channel's purpose within a month.
- What if the channel is nearly silent? If your CSAT response rate is 8%, a small team may see only a few posts a week. That is a survey-coverage problem — look at when and how you ask for ratings — and never a reason to read a quiet feed as a quiet team.
Supportman sends selected Intercom ratings into Slack as soon as customers submit them. You choose which rating types reach the celebration channel and route dissatisfied feedback somewhere built for response.
Add meaning to the rating
A five-star post is an outcome. It becomes recognition — and teaching material — when someone names the behaviour behind it.
Weak response:
Great job!
Stronger response:
You acknowledged the migration deadline, gave the customer a safe workaround, and kept ownership through the engineering handoff. That is exactly what “make it simple” should look like.
The stronger version does three jobs at once: it tells the agent precisely what to repeat, it shows the rest of the team what the standard looks like in a real conversation, and it attaches a company value to observable behaviour rather than a poster on the wall. Aim the same specificity at ordinary work — a careful check, a context-rich handoff, a promised follow-up completed — and do not wait for dramatic saves. Gallup and Workhuman's research links recognition of small everyday accomplishments, aligned with company values, to far stronger recognition experiences and clearer expectations at work; the companion article walks through those numbers.
You do not need commentary on every post. One specific sentence on the posts that contain a lesson is enough.
Credit the whole contribution
The rating names one agent. The win often has more authors: whoever wrote the macro, found the workaround, handled the earlier escalation, or covered the queue.
“Tag contributors when they are known” fails in practice because knowing is nobody's job. Make it mechanical instead:
- The assigned agent threads the assist. When a rating lands, the named agent replies in the thread crediting who helped: “Assist to Marcus — his billing-migration macro did half the work here, and Dana found this workaround during last month's escalation.” The person best placed to know who helped is the one who received the praise.
- The manager asks one question in the weekly recap: “who else touched this one?” Asked every week, it trains the team to surface invisible contributions on their own.
- Pick an assist reaction. One agreed emoji — a handshake, a relay baton, whatever the team chooses — marks team wins, so anyone scrolling the channel later can find them without opening every thread.
There is evidence this peer layer matters. Networks of Gratitude, an analysis of 422,000 workplace appreciation messages, found that appreciation frequently flowed between employees at the same level and across different parts of the organization — peers notice work that managers never see — and that employees expected managers to act on recognition, not merely witness it. Both halves apply here: let peers do the tagging, and treat repeated assists as evidence in coaching, staffing, and performance conversations.
Design against the leaderboard
Public recognition is not automatically positive. A study in Management Accounting Research found that a public-only peer-recognition feed encouraged simplified “me versus them” comparisons and decreased feelings of appreciation — people started tracking who appeared in the feed, and who did not. Read the Management Accounting Research study.
You do not need to publish a ranking for the channel to become one. Three failure modes show up on real teams, and each has a countermeasure.
The queue-mix problem
Imagine a 12-person team. One agent works live chat: 40 short conversations a day, each ending with a rating prompt. A colleague works escalations over email: five hard cases a day, most closed long after the customer stopped replying, rarely rated at all. Within a month the chat agent has appeared in the feed five times as often — not because the work is five times better, but because the queue generates five times the rating prompts.
Countermeasure: treat feed frequency as a property of the queue, never of the person. Do not publish counts. When you pick examples for the weekly recap, choose by behaviour and deliberately pull from the low-rating queues — the escalation case with no rating attached can still be the win of the week.
The time-zone problem
An APAC agent's five-star rating posts at 3pm Sydney time and scrolls away while the US team sleeps: zero reactions. The same rating on a US morning collects a dozen. Nobody intends anything, but after a quarter the APAC team reads the channel as proof they are invisible.
Countermeasure: make the weekly recap draw from every shift by rule, not by whatever sits near the top of the channel. Ask a lead in each region to react and add one context reply to overnight posts when they come online — a reaction eighteen hours later still counts.
The accidental ranking
A well-meaning exec drops into the channel and replies “another one for Sarah!” — and has just started keeping score in public. Counting is ranking, even when it is affectionate.
Countermeasure: put the norm in the channel description (“we celebrate behaviours, not tallies”) where every new member sees it, executives included. When a tally comment appears, redirect in the thread — “this one is a great example of holding ownership through a handoff” — so the exec keeps face and the norm holds.
Underneath the specific fixes sit three standing rules: never publish raw rankings, pair public praise with private and specific feedback, and ask people how they prefer to be recognized — some of your best agents would rather not be posted about at all.
Create a weekly ritual
Real-time posts create the moment; a five-minute weekly recap keeps it from disappearing beneath the next queue. On Friday, a lead picks one customer quote, one quiet or behind-the-scenes contribution, and one behaviour worth repeating — then says thank you. That is the whole ritual. It is also where two of the anti-leaderboard countermeasures live: the recap is the place you deliberately draw from every shift and queue, and where you ask “who else touched this one?”
Audit the health of the channel
Once a month, spend fifteen minutes on the last 30 days of the channel. Export or scroll the posts and tally three things: which agents were named, how quickly posts drew a first reaction, and how many posts received a specific context reply. Then check the tallies against a few starting-point thresholds — rough, but far better than intuition:
- Coverage: if fewer than roughly 70% of agents appeared this month, look at queue and shift mix before assuming anything about performance — and fix the mix or the recap selection, not the people.
- Latency: if the median time from post to first reaction is over a day, the channel has become a log rather than recognition. A rotating “first responder” who reacts and adds one line during their shift fixes this cheaply.
- Specificity: if fewer than one post in five drew a comment naming a behaviour, the feed is drifting toward applause-only. Seed it yourself — one specific reply a week from a lead resets the tone.
- Concentration: if two or three names account for most mentions, re-read the queue-mix section before the team does.
These are health checks, not targets. The moment “70% of agents recognized” becomes a quota, people manufacture recognition to hit it and the channel stops meaning anything.
Make customer praise part of the workflow
Expect week one to be quiet: a handful of automated posts, few reactions, and a slightly awkward silence until someone models the behaviour. Plan for that — write the first two context replies yourself, thread the first assist credit, and run the first Friday recap even if there are only four posts to choose from. The habits in this article are small, but none of them start themselves.
Supportman handles the mechanical half: it moves positive Intercom ratings into the Slack channel where the team already works, with the customer's remark and conversation context attached, the moment the rating is submitted.