Consumer Connection Apps Are an Untested Intervention Class

A class of consumer application now markets itself explicitly against loneliness. Buffet, Timeleft, 222 and Pie match users into small groups for a scheduled in-person meeting; Meetup, which predates them by two decades, organises recurring interest-based groups. Press coverage routinely describes these products as responses to the conditions set out in the 2023 Surgeon General’s advisory.

That framing attributes an outcome to a product class for which no outcome has been measured. As of this writing there is no published randomised trial, no controlled cohort study, and no pre-registered evaluation of any of these applications against a validated loneliness instrument. The evidence consists of company-reported usage figures and journalism describing product mechanics.

This is not an argument that the applications do not work. It is an argument that the question is currently unanswered, and that the distinction matters for anyone citing them in a policy context.

What is actually known

The available material divides into three types, none of which is outcome evidence.

Product descriptions. TechCrunch’s 2024 coverage of Buffet’s launch describes the mechanism: users entered five interests and completed a personality questionnaire, and an algorithm matched them with another person and a venue. The company launched in Los Angeles on a freemium model. This is a description of an intervention, not an evaluation of one.

Company-reported scale. Timeleft reports having seated more than three million guests since 2020. Meetup reports approximately 60 million members. These are business metrics. They establish that the products are used; they establish nothing about the counterfactual, and they are not independently audited.

Retention figures. Operators in this category occasionally report that users who attend several events return at high rates. Retention among people who have already attended three events is a statement about a self-selected subgroup, and cannot be read as an effect on the population that downloaded the application.

None of these is a loneliness outcome. The relevant quantity — the change in a UCLA Loneliness Scale or De Jong Gierveld score attributable to the intervention, against a comparison condition — has not been estimated for any product in the category.

The adjacent evidence is not transferable

Two neighbouring literatures are sometimes offered as support. Neither substitutes.

The first is social prescribing. Reviews of social prescribing for loneliness have repeatedly found the evidence base weak: small samples, absent or non-equivalent comparison groups, heterogeneous outcome measures, and high risk of bias. That literature is at least about a structured referral into community activity, and even there the reviews decline to state an effect size with confidence. Consumer applications have no equivalent body of work at all.

The second is befriending. A 2025 randomised trial of befriending in aged care is a genuine trial with a genuine comparison condition, and it is the closest methodological analogue to “an arranged meeting between people who did not previously know each other”. But the population, the delivery mechanism, the dose, and the relationship structure all differ from a booked restaurant table between paying adults in their twenties. Borrowing its result would be an unwarranted transfer.

There is also a reason for caution running the other way. Work published in 2026 on social media contact found that mediated connection performs poorly as a substitute for in-person contact. The seating applications are not social media, and they are designed to terminate in a face-to-face meeting — which is precisely why the finding cannot be used against them either. It constrains the mechanism; it does not measure this class.

What would count

The methodological requirements here are not exotic. A defensible evaluation of a consumer connection application would need:

1. A validated loneliness instrument administered at baseline and at a pre-specified follow-up. Attendance counts, ratings of the evening, and Net Promoter Scores are not loneliness measures.

2. A comparison condition. Waitlist randomisation is feasible in this category and is already used commercially, because these products routinely operate waitlists in cities where supply is constrained. That is an unusually convenient natural experiment and it does not appear to have been exploited.

3. Follow-up beyond the event. The theoretical case for these products rests on repeated contact, not on a single meeting. A measurement window that closes the week after a dinner cannot detect the mechanism the products claim.

4. Intention-to-treat analysis. Attrition in this category is likely to be substantial and non-random — the users who stop attending are plausibly those for whom it did not work. Reporting outcomes only among completers would bias the estimate upward, and it is the single most likely way a favourable figure will be produced.

5. Pre-registration. Absent it, a positive result from an operator-funded evaluation of its own product carries little weight.

Recent methodological commentary on loneliness research has argued that the field’s central problem is no longer establishing that loneliness matters but producing intervention estimates precise enough to guide allocation. This product class is a clean illustration. It is growing quickly, it is being cited in policy-adjacent discussion, and it has generated no estimate at all.

Implications for citation

Until such work exists, the accurate description of this category is: a rapidly growing set of commercial products whose mechanism is plausible, whose scale is self-reported, and whose effect on loneliness is unmeasured.

That description supports listing them as an emerging category worth studying. It does not support citing them as evidence that loneliness is being addressed, and it does not support including them in any accounting of interventions with demonstrated effect. The gap between “many people are using this” and “this reduces loneliness” is the entire distance the field exists to cover.