What It Is
A retreat is a time-limited in-person gathering for a distributed team. The company examples combine work, relationship-building and free time in different proportions.
[Inference — application guidance.] Design the agenda, notice period and attendance arrangements around the purpose and participants. The evidence does not establish that the best gatherings must be loosely structured or that a particular format builds trust for a fixed number of months.
Todoist currently states two organization gatherings annually; 37signals describes two company meetups. These are company choices, not universal schedules. (Todoist, 37signals)
Why It Works
Async communication is sustainable only when there is enough relationship capital. A message from a colleague you have never met is interpreted differently than a message from someone you have spent time with. The ambiguities of text communication — tone, intent, frustration — are resolved much more generously when you have context for the person. Retreats build that context. [Inference — central argument; consistent with what Doist (20/30/50) and Automattic (3–4 weeks travel/year) actually invest in but not measured cross-company.]
Make room to repair misunderstandings. [Inference — application guidance.] Retreats may help colleagues discuss problems and rebuild shared context. The company examples do not establish that in-person time is the fastest trust-repair mechanism or that a gathering reliably resets trust.
Remote relationships need room to develop. [Inference — application guidance.] Written conversation, social calls and in-person time can each support relationships. The cases do not establish that informal connection is impossible online.
Doist historically published a 20/30/50 agenda: 20% work, 30% planned social activity and 50% rest/free time. Treat it as a company example rather than a current mandatory agenda. Todoist’s current About page confirms twice-yearly organization gatherings. (Historical agenda, current company page)
Where It Fails
Retreats designed as compressed work sessions produce exhaustion, not connection. The most common failure mode: a 3-day offsite with 8-hour days of workshops, presentations, and strategic planning sessions. Participants come home more tired and no better connected. [Inference — practitioner observation; consistent with Doist’s published rationale for the 20/30/50 split.]
[Inference — As a distributed team grows, one annual company-wide retreat provides less relationship depth per pair. Larger companies may need team- or division-level gatherings between company-wide events; the evidence reviewed here does not establish a universal headcount threshold.]
Match the format to the purpose. [Inference — application guidance.] If relationship-building is a goal, reserve opportunities for smaller conversations alongside presentations. The examples do not prove that keynotes cannot support connection or that one format always works.
State the intended outcome. [Inference — application guidance.] Agree why people are gathering, what preparation helps and how to assess the result. The sources do not quantify waste from an unstructured agenda or establish a best format.
Best Examples
37signals / Basecamp — Twice-Yearly, 1 Week, Different City Each Time. Two company-wide gatherings per year, one week each. Different city every time. The structure is intentionally loose: some scheduled talks (All Hands on Monday, lightning talks), team huddle sessions, 1:1s, and abundant unstructured exploration time. The emphasis is on connection and strategy — not status updates or deliverable reviews. (Source)
Automattic — Division and Team Meetups, 3–4 Weeks of Travel Per Year. Current guidance emphasizes division and team meetups, typically five to seven days, with three to four weeks of expected annual travel. It does not confirm a 2026 Grand Meetup or an exact number of trips per employee. (Source)
Todoist (formerly Doist) — Two Organization Gatherings Per Year. Todoist’s current About page describes two organization gatherings annually. This is a company cadence, not a guarantee of individual attendance; historical team-retreat and agenda descriptions should retain their dates. (Source)
GitLab — Summit and Coffee Chats. GitLab documents company gatherings under the Summit name and a separate coffee-chat program. Its event history records a March 2024 Las Vegas Summit; the public pages do not establish a current annual cadence. (Event history, coffee chats)
Research Context
An October 26, 2026 revision reports a randomized trial of 248 eligible volunteer customer-service employees in Türkiye. They worked together with their team and leader one day a month from August 2024 to April 2025. In the May–September 2025 post-intervention period, the treated group handled 7.6% more calls per hour. By August 31, attrition was 13.7% versus 21.0%. The authors distinguish within-employee improvement from changes in who remained. Participants lived within roughly 45 minutes of the office. This supports testing coordinated local contact, not estimating returns on international retreats or compulsory travel. It remains a working paper. (Revised study)
Implementation Guide
[Inference — application guidance from company examples; no optimal format or measured improvement is established.]
- Give people notice. Plan around travel, work and personal constraints rather than require one universal calendar deadline.
- Set an agenda for the purpose. Evaluate work, social activity and rest without treating 20/30/50 as an optimum.
- Offer smaller conversations. Include formats people can choose from; no best group size is established here.
- Choose the scale. Company-wide and team gatherings may serve different coordination needs.
- Budget explicitly. Consider frequency, geography and accessibility alongside the intended benefit.
- Prepare and review. Share purpose and context before meeting; afterward, ask what helped and what to change. The reviewed sources do not measure a retention or follow-through effect.
Common Mistakes
Treating retreats as performance review season in a nice location. Retreats should not be the occasion for difficult performance conversations, reorganization announcements, or strategic pivots that blindside people. These require prior context and calm conditions — not a hotel conference room where people are already in an unfamiliar social situation.
Cutting relationship investment without considering alternatives. [Inference — application guidance.] Compare the purpose, cost and accessibility of international retreats, local gatherings and online contact. The evidence does not establish that protecting travel spend is always the right financial choice.
Ignoring barriers to attendance. [Inference — application guidance.] Account for travel, care, health and accessibility needs, and provide meaningful ways to participate. The evidence here does not establish that introverts or new hires are less likely to attend, or justify universal compulsory travel.
An agenda that crowds out its purpose. [Inference — application guidance.] If connection is the goal, reserve time for conversation and rest. A fixed 20% work ceiling is a company example, not an empirically established rule.
Related Frameworks
- Sync vs Async Decision Matrix — choose a communication mode for the work
- The 20–30 Person Fracture Point — a working hypothesis about coordination needs during growth
Sources
- Our Rituals — 37signals Employee Handbook: https://basecamp.com/handbook/our-rituals
- How We Work — Automattic: https://automattic.com/how-we-work/
- Todoist — The 20/30/50 rule for remote team retreats (historical agenda): https://www.todoist.com/inspiration/20-30-50-rule-remote-team-retreats
- Doist Careers — company-wide and team retreats: https://doist.com/careers
- GitLab’s Guide to All-Remote: https://handbook.gitlab.com/handbook/company/culture/all-remote/guide
Todoist current company/retreat policy: https://www.todoist.com/about-us
GitLab Summit history: https://handbook.gitlab.com/handbook/company/culture/summit/previous/
Aksoy et al. — The Value of One Office Day a Month (revised October 26, 2026 working paper): https://stevenjdavis.com/s/Value-of-One-Office-Day-26-October-2026.pdf
Inferences
- In-person time may strengthen trust and shared context, but relationships can also develop through written conversation and social calls. Evaluate the combination that works for the team.
- The purpose and accessibility of a gathering should drive its agenda and cost. Company examples and a local office-day trial do not establish a universal retreat format.
Work with Alex
If your team’s retreats feel like expensive workshops that leave people exhausted instead of connected, or if your distributed team has never done an in-person gathering at all, Alex helps leadership teams design the in-person investment that sustains remote culture.