The week-three break.
What we found
Retention holds for three weeks, then falls 28% in week four, and 802 of the 844 teams lost that week never got a second collaborator in by day 18.
Solo teams carry the break
Exhibits B–D · product events
Split the same 3,412 teams by whether a second person had joined and done something by day 18, and the curve comes apart.
Teams with a second collaborator lose 2% of their active teams between week three and week four. Solo teams lose 77%: of 1,042 still active in week three, 240 come back in week four.
Solo teams are 38% of trials, but they account for 95% of the break. Fewer than one in five return in week four, and by week eight one in ten is left.
| Cohort at day 18 | Teams | Week 3 | Week 4 | Wk 3 to 4 | Week 8 |
|---|---|---|---|---|---|
| Second collaborator in | 2,11062% | 1,94192% | 1,89990% | −2% | 1,77284% |
| Solo | 1,30238% | 1,04280% | 24018% | −77% | 13010% |
| All trial teams | 3,412 | 2,98387% | 2,13963% | −28% | 1,90256% |
Invite intent shows up in session two, then disappears
Exhibits E–G · 2,190 teams that began session two alone
Teams that start alone reach for the invite dialog early. In session two, usually on day 2, 60% of them open it: they know who should be in the workspace.
71% close it again without sending anything, and most of those asked said there was nothing to share yet.
Opening the dialog did not protect a team. Of the 933 that closed it, 562 were still solo on day 18, and 17% of those were active in week four – no better than solo teams that never opened it (21%).
What trial teams told us
Interviews and support tickets
Interview“By the time I had a board worth showing people, the trial was half over and everyone had moved on to something else.”
Ticket #48213“I tried to add our designer, but there is nothing in here yet. What am I inviting her to?”
Interview“It clicked when I assigned Maya a review. She answered inside Wayline and I stopped chasing her on email.”
Interview“I kept meaning to invite the others once it was set up properly. It never felt set up.”
Ticket #49007“Can we extend the trial? We never got round to using it with the team.”
11 interviews, 19–31 Jul 2026. 214 trial support tickets tagged invite or sharing.
Why it happens
Exhibit H · medians for teams that began session two alone
Day 2Signal
Invitation intent appears in session two, then disappears.
The intent is real and early: 60% of teams working alone open the invite dialog in their second session. Most close it again, because the workspace holds nothing a colleague could act on yet. 58% of those asked said so in as many words.
Exhibits E–G
The first real reason to invite arrives ten days later.
Day 12Trigger
The first shared workflow arrives too late.
The first item assigned to someone other than its creator shows up at a median of day 12. Teams that invite later almost always do it within a day of that moment. For the rest, day 18 passes without a second person, and teams that do not invite a second collaborator by day 18 rarely return in week four.
Exhibits B, D
So the handoff has to move to where the intent already is.
Day 0Next move
Bring a real handoff into the first-run path.
End setup by handing one real item to a named teammate: a review, a checklist, a board with an owner. The invite then arrives with work attached, in the session where people already want to send it.
What we recommend
Exhibit I · decision needed by 19 Aug 2026
Bring a real handoff into the first-run path, and test collaborative setup before adding more acquisition.
At today’s rates, every 1,000 new trial teams include about 380 still working alone on day 18, and about 310 of those are gone by week four. More acquisition buys more of that loss.
We propose one experiment on the first-run path, and a hold on the planned Q4 increase in paid acquisition until it reads out. The split above is observational, so the experiment measures week-four retention directly instead of assuming the gap is causal.
- Hypothesis
- Ending setup with one item handed to a named teammate raises the share of trial teams with a second collaborator by day 18 from 62% to at least 68%.
- Primary metric
- Second collaborator active by day 18. Baseline 62%.
- Read-through
- Week-four retention. Baseline 63%.
- Guardrails
- First-session setup completion (71%) and trial-to-paid conversion fall by no more than 1 point. Invite emails reported as spam stay under 0.3%.
- Design
- New trial teams split 50/50 at workspace creation.
- Duration
- Six weeks of enrolment from 24 Aug, about 2,100 teams, each followed to day 28. Readout in the week of 2 Nov 2026.
- Owner
- Growth product manager, activation. With the first-run design lead and one analyst.
Open the decision note
Decision note for the Head of Product, needed by 19 Aug 2026
- We ask for
- Approval to ship the handoff experiment in the 24 Aug release.
- A hold on the Q4 paid-acquisition increase until the readout in the week of 2 Nov.
- Cost
- One squad for about three weeks: two engineers and one designer. No new tooling.
- If it works (68% or more)
- Roll the handoff out to all new trials, then revisit acquisition with day-18 collaboration as the gating metric.
- If it does not (under 65%)
- The handoff is not the missing piece. Next candidate: team templates chosen at signup. Between 65% and 68%, extend enrolment by three weeks before deciding.
- What it could be worth
- At about 1,500 trial starts a month, six more points of collaboration is roughly 90 more teams with a second person by day 18. If the retention gap is fully causal, that is up to 64 more teams active in week four each month; if it is partly causal, fewer.
- Either way
- The acquisition hold ends at readout. The break is a first-run problem, and more trials at today’s collaboration rate repeat it.
Method and data notes
Data pulled 10 Aug 2026
- Sources
- Product event stream: workspace created, session start, invite dialog opened, invite sent, invite accepted, item assigned.
- 412 answers to the close prompt on the invite dialog. 11 interviews, 19–31 Jul 2026: six teams whose trial lapsed and five that converted. 214 trial-period support tickets tagged invite or sharing.
- Window
- Team trials started 6 Apr–14 Jun 2026, each followed for 56 days, so every team has a full eight weeks of data. Trials last 30 days; weeks 5–8 include teams that moved to the free plan.
- Exclusions
- 146 internal and test workspaces. 58 workspaces provisioned with several members at creation through single sign-on, since they never needed an invite.
- Definitions
- Active in a week: at least one member created, edited, commented on or completed an item. Weeks count from workspace creation; week 1 is days 0–6. Second collaborator: a second member who accepted an invite and took one of those actions by day 18. Session: activity separated by 30 minutes of inactivity. Shared workflow: an item assigned to someone other than its creator.
- Week-to-week change is the net change in active teams, so −28% is 2,139 active teams in week 4 against 2,983 in week 3.
- Why day 18
- We tested cut-offs from day 7 to day 28. Day 18 separates week-four retention most sharply, a 72-point gap; the gap is 66 points at day 14 and 69 at day 21.
- Caveats
- This is an observational split. Teams that invite early may differ in size or intent. The gap holds within self-reported team sizes of 2–5 people (70 points) and 6–20 people (73 points), but that does not make it causal, which is why we recommend an experiment rather than a rollout.
- Contact
- Growth research. Queries and the event definitions live in the investigation workspace.