# Run a coffee chat program people show up to > How to run an opt-in coffee chat program across teams: who it’s for, group size, cadence, the match message and what to measure. Audience: HR and people teams at companies with remote or growing teams Canonical: https://shuffl.ai/playbooks/coffee-chat-program Coffee chat programs pair coworkers who wouldn’t otherwise meet for a short, unstructured conversation. Most fade within a few months because nobody decided who they were for, how often they run or what to say when the match arrives. This playbook covers why cross-team connection is worth the time, how to choose group size and cadence, how to run it opt-in, and what to measure. ## Why it matters People don’t meet across teams by accident as often as they used to. When Microsoft moved to firm-wide remote work, the share of collaboration time spent with cross-group connections fell by about 25% of its earlier level, and groups became more siloed[^microsoft-remote-collaboration]. Close ties at work are rare. In the U.S., just two in 10 employees report having a best friend at work, and Gallup found the link between that and outcomes such as intent to leave grew stronger after the pandemic began[^gallup-best-friend]. Structured small-group chats can help newcomers. In a field experiment with 1,370 remote interns, those randomly given weekly 30-minute video sessions with a senior manager and three or four other interns were 4.7 percentage points likelier to get a full-time offer than interns given a Q&A forum. Against interns given nothing extra, the gap was about 2 points and not statistically significant[^worldbank-water-coolers]. **Why cross-team chats need a nudge** - About 25% drop in the share of collaboration time spent with cross-group connections after Microsoft went fully remote[^microsoft-remote-collaboration] - 2 in 10 U.S. employees report having a best friend at work[^gallup-best-friend] - About 4 the upper limit on people in a spontaneous conversation[^dunbar-conversation] ## The plan Ask a senior leader to join in the first round and say so in the launch message. It tells people the time is allowed. - Decide who it’s for: everyone, new hires, or people across two teams that need to work together. - Make it opt-in. Use a channel people join, and say how to pause. - Keep groups small: two to four people. Pairs are easiest to schedule; threes survive one person being busy. - Pick a cadence: every two weeks or monthly. Run it for a full quarter before changing it. - Match across teams, and avoid matching someone with their own manager or direct reports. - Send a match message that names who books the time and gives one question to start with. ## Running it Most of the work is in the first two rounds. After that, the program runs itself as long as people can pause and come back. - Launch: one message in a company channel saying what it is, how often, how long a chat is and how to join. - Each round: a group message that introduces everyone with a line about each person and suggests who books the time. - Midway through the quarter: a reminder to people who joined but haven’t met anyone yet. - End of quarter: share the numbers (people taking part, chats made, cross-team matches) and ask for one change. ## An example A 300-person company with offices in two cities and a third of staff remote launches a coffee chat channel. 90 people join in the first two weeks. HR sets groups of three, every two weeks, matched across departments. Each round makes 30 groups. Over a quarter that is six rounds and 180 introductions, and each person meets 12 coworkers they might not have met. After the first month, 10 people pause during a product launch and come back the month after. At the end of the quarter HR shares the numbers in the launch channel and asks whether to move to monthly. Most say keep it every two weeks. ## Common mistakes - Enrolling everyone by default instead of asking people to opt in. - Running weekly matches for a large company and watching people mute them. - Matching people with their own teammates, who already talk every day. - Sending a match with no prompt, so nobody knows who should message first. - Never telling people they can pause, so they leave for good instead. ## Launch message for a coffee chat program ```text We’re starting coffee chats across teams. Every two weeks you’ll be put in a small group of three with people from other teams. Meet for 20 minutes, on a call or in person, whenever suits you. The first person named books the time. Not sure what to talk about? Start with: what are you working on that most of the company doesn’t know about? Join #coffee-chats to take part. You can pause any time and come back when you’re ready. Leaving the channel takes you out. ``` ## Do this in Shuffl 1. Connect the Shuffl app in Slack, then open Connections from the HR workspace and choose New program. Start from the Cross-team coffee template, or choose Review and start on the program Mello suggests. 2. Pick the Slack channel whose members take part. Joining the channel is how people opt in. 3. Under Who should meet whom?, choose People from different groups, departments or communities, set the group size, and under Keep coworkers apart avoid groups of managers and their direct reports. 4. Under Repeats, choose every 2 weeks or every month, set Starts and At, and write the introduction. The preview shows the Slack message. 5. Choose Start program, or use Run now to preview the groups before a round goes out. 6. Check Insights for who is taking part, first-time pairs and introductions across teams. People can take a break from the Home tab of the Shuffl app in Slack. Guide: [Introduce coworkers with Connections](https://shuffl.ai/docs/markdown/connections) - [Create an introduction program](https://shuffl.ai/docs/connections#program) - [Choose who gets matched](https://shuffl.ai/docs/connections#matching) - [Set the schedule and review rounds](https://shuffl.ai/docs/connections#schedule) - [What employees see](https://shuffl.ai/docs/connections#employees) ## Common questions ### How many people should be in each coffee chat? Small. Research on spontaneous conversation finds an upper limit of about four people talking together[^dunbar-conversation]. Pairs are the easiest to schedule; groups of three or four keep going when one person can’t make it. ### How often should coffee chats run? Every two weeks or once a month suits most companies. Weekly works for a small, keen group but tends to wear people out. Pick one, run it for a quarter, and let people take a break without leaving. ### Should coffee chats be mandatory? No. Make it opt-in, with an easy way to pause. People who chose to join turn up. People who were enrolled without asking cancel, and the person matched with them has a worse time than if there had been no match. ### How do I know the program is working? Track who takes part, how many first-time pairs you made and how many matches crossed teams. A short question every quarter, such as whether people know someone outside their team they could ask for help, tells you whether it’s working. ## Glossary - [Belonging](https://shuffl.ai/glossary/belonging) - [Employee engagement](https://shuffl.ai/glossary/employee-engagement) - [Employee experience](https://shuffl.ai/glossary/employee-experience) - [Employee resource group](https://shuffl.ai/glossary/employee-resource-group) - [Onboarding](https://shuffl.ai/glossary/onboarding) - [Psychological safety](https://shuffl.ai/glossary/psychological-safety) ## Related playbooks - [Plan a new hire’s first 90 days](https://shuffl.ai/playbooks/markdown/first-90-days) - [Run an engagement check without survey fatigue](https://shuffl.ai/playbooks/markdown/engagement-check-in) - [Build a recognition habit that lasts](https://shuffl.ai/playbooks/markdown/recognition-habit) - [Celebrate milestones without it feeling automated](https://shuffl.ai/playbooks/markdown/celebrate-milestones) ## Sources [^microsoft-remote-collaboration]: Nature Human Behaviour (via Microsoft Research), “The effects of remote work on collaboration among information workers”, September 2021. https://www.microsoft.com/en-us/research/publication/the-effects-of-remote-work-on-collaboration-among-information-workers/ Peer-reviewed study by Longqi Yang and colleagues using data on Microsoft employees’ work patterns before and after the firm-wide work-from-home mandate in March 2020. The 25% is a relative decline, not percentage points. [^gallup-best-friend]: Gallup, “The Increasing Importance of a Best Friend at Work”, August 2022, updated January 2024. https://www.gallup.com/workplace/397058/increasing-importance-best-friend-work.aspx Gallup article by Alok Patel and Stephanie Plowman. 2022 poll of 16,586 U.S. adults who work for an employer, margin of error ±1 point at the 95% confidence level. [^worldbank-water-coolers]: World Bank, Development Impact blog, “Onboarding summer interns in a virtual work environment: an experiment highlights the pros and cons of “virtual water coolers””, June 2021. https://blogs.worldbank.org/en/impactevaluations/onboarding-summer-interns-virtual-work-environment-experiment-highlights-pros-and David McKenzie’s summary of a randomized field experiment by Choudhury, Lane and Bojinov (HBS working paper 21-125) with the North American cohort of 1,370 remote summer interns at one global firm in 2020. 91.5% of interns received an offer overall. [^dunbar-conversation]: Human Nature (via Northumbria University research portal), “Size and structure of freely forming conversational groups”, March 1995. https://researchportal.northumbria.ac.uk/en/publications/size-and-structure-of-freely-forming-conversational-groups/ Peer-reviewed paper by R.I.M. Dunbar, N.D.C. Duncan and D. Nettle, Human Nature 6(1), 67 to 78, using observations of conversations in various settings. Agent documentation index: https://shuffl.ai/llms.txt