Consumer product company / Boston
Over two years the team shipped three consumer products, grew past 10,000 student signups with no paid media, and used funnel and cohort analysis to reach a clear conclusion about the economics of the category. This page documents the thesis, the products, the evidence and the decision.
The problem
U.S. adults reported loneliness before COVID-19
Higher heart disease risk linked to poor social connection
Higher stroke risk linked to poor social connection
Higher dementia risk linked to chronic loneliness and isolation in older adults
“Loneliness is far more than just a bad feeling. It harms both individual and societal health. Our relationships are a source of healing and well being hiding in plain sight.”
Loneliness was not a fringe concern. In 2023 the U.S. Surgeon General called it a public health priority on the scale of smoking, and the figures above come from that research. Reading it was part of what convinced us the problem was worth years of work.
So Weave began with a simple question. Students live surrounded by thousands of peers and still report record isolation. Could technology pull people out of their feeds and into the same room? We spent two years finding out.
Source: U.S. Surgeon General, Our Epidemic of Loneliness and Isolation (2023). Risk estimates reflect the populations and measures cited in the advisory.
Community in practice
The team hosted meetups and events through Weave, ran focus groups at the Harvard i-lab, and studied how shared public space creates connection.
Product evolution
How our products evolved, from a neighborhood community app to a campus calendar and then campus dating.
A neighborhood social graph. Map based local posts, interest groups, events and conversations in one app.
Every Harvard event in one place. Aggregated fragmented school, club and department calendars into a single feed.
A campus scoped connection app built for one clear job, getting students to meet in person.
Go to market
Every campus launch was run the same way so results could be compared. Print a coded flyer, saturate the highest traffic locations, talk to students directly, then read the funnel.
One campus at a time until the network was usable, rather than thin coverage everywhere.
Separate scan destinations per location made it possible to compare dining halls, dorms and libraries.
Team members and student ambassadors onboarded users in person during high traffic moments.
Measurement
Signup totals hide the thing that decides whether a consumer product works. The team instrumented six stages per campus and weekly retention curves per cohort, then compared them side by side.
| Campus | Flyer scans | Signups | Profile complete | Activated | Referred a friend | Day 7 return |
|---|---|---|---|---|---|---|
| Harvard | 4,200 | 2,140 | 1,290 | 780 | 240 | 215 |
| MIT | 3,100 | 1,520 | 910 | 560 | 175 | 150 |
| Boston University | 3,600 | 1,810 | 1,080 | 640 | 190 | 175 |
| Northeastern | 2,900 | 1,420 | 855 | 500 | 145 | 135 |
| Boston College | 2,100 | 1,040 | 620 | 360 | 105 | 95 |
| Yale | 1,800 | 890 | 535 | 310 | 90 | 80 |
| All campuses | 17,700 | 8,820 50% | 5,290 60% | 3,150 60% | 945 30% | 850 27% |
Campus level figures are representative of internal tracking across the 2025 launches and are shown to illustrate the method rather than to report audited totals.
Each row is a weekly signup cohort and each column is the share of that cohort still active in the following weeks. Later cohorts held better early retention, which is what justified continued campus expansion.
| Cohort | Users | Week 0 | Week 1 | Week 2 | Week 3 | Week 4 | Week 5 |
|---|---|---|---|---|---|---|---|
| Sep 7 to Sep 13 | 72 | 100.0% | 59.7% | 52.8% | 41.7% | 40.3% | 31.9% |
| Sep 14 to Sep 20 | 92 | 100.0% | 68.5% | 55.4% | 47.8% | 25.0% | 15.2% |
| Sep 21 to Sep 27 | 109 | 100.0% | 60.6% | 52.3% | 42.2% | 18.3% | 13.8% |
| Sep 28 to Oct 4 | 60 | 100.0% | 65.0% | 55.0% | 35.0% | 21.7% | — |
| Oct 5 to Oct 11 | 35 | 100.0% | 54.3% | 45.7% | 28.6% | — | — |
| Oct 12 to Oct 18 | 29 | 100.0% | 79.3% | 62.1% | — | — | — |
| Oct 19 to Oct 25 | 24 | 100.0% | 70.8% | — | — | — | — |
| Oct 26 to Nov 1 | 20 | 100.0% | — | — | — | — | — |
Values below are transcribed from the original App Store Connect screenshots. They represent different windows, not a single combined funnel.
For context, Similarweb estimated that fewer than one in nine new users of three leading U.S. dating apps were still active after 30 days in 2023. Our weekly cohorts were different: roughly a quarter to a third of new users were still active at day seven, and later cohorts held stronger early retention still. Against the public benchmarks available to us, Campus Dating consistently outperformed established dating apps on retention.
Source: Similarweb, Bumble and Hinge: the Dating Apps Americans Are Most Likely to Keep (2023). Third-party Android estimates shown as context; different measurement windows than our weekly cohorts.
Unit economics
Once acquisition was proven, the question became whether a subscription could carry the business. The team modelled conversion, price, churn and the platform fee across three scenarios.
Contribution never covers campus launch effort and support load.
Breaks even on infrastructure but not on the team required to keep launching campuses.
Profitable per campus, but requires retention behaviour that a matching product structurally works against.
When users find a match, they may have less reason to keep paying. That potential churn, alongside graduation cycles and platform fees, made sustainable growth harder to project. The scenarios showed that scaling to more campuses would have required additional investment before the economics were clear.
With runway limited, the team chose to close rather than spend the remaining capital scaling a model the analysis did not support. The decision followed from the numbers.
Operating principles
Weave spent too long developing the first app before testing the underlying need with people. An offline gathering or a small, manual experiment could have shown what kind of connection people actually wanted. Shorter cycles with real participants would have helped us identify the minimum useful solution sooner.
Products that serve everyone retain no one. Weave Community tried to hold events, discovery, messaging and civic action at once. Campus Dating held one job, and the traction followed almost immediately.
Scans, installs, profile completion, activation, referral and day seven return were tracked separately by campus. Growth problems and retention problems look identical in a signup count and completely different in a funnel.
Acquisition strength can disguise a structurally unprofitable model. Modelling conversion, price, churn and platform fees before scaling spend is what turned a hard outcome into a deliberate decision.
Team
Product, engineering, design, community and brand, built to run consumer experiments end to end.
