Should You Refer Anyone to a Consumer Connection App?

Sooner or later someone on your caseload asks what app they should download, or a commissioner asks why your referral list is all church halls when there are products that do this now.

You need an answer that is more useful than “those are not evidence-based” — which is true, and which does not help the person in front of you who has no plans this weekend.

Here is a framework for sorting the category and deciding when a referral is defensible.

Step 1: Sort the tools by what they actually do

The single most common mistake is treating these as one category. They are four, and only two of them will produce contact with a new person.

Group A — Seats you with strangers. Buffet, Timeleft, 222, Pie. An algorithm builds a small group and books a venue. The user turns up. TechCrunch’s coverage of Buffet’s launch describes the mechanism typical of the group: a short questionnaire, then a match with people and a place. Group A is the only category that requires nothing of the user socially before arrival.

Group B — Recurring interest groups. Meetup, and every run club and league in your area. The same people meet repeatedly around an activity. This is the category with the strongest theoretical case, because repeated contact is the mechanism the literature actually supports.

Group C — Event logistics. Eventbrite, Luma, Partiful. These sell tickets, host event pages, and manage invitations. They do not introduce anyone to anyone. Partiful in particular assumes the user already has a guest list, which is exactly what an isolated client does not have. Referring someone here as a loneliness intervention is a category error, and it is the most frequent one.

Group D — Dating. Hinge and its equivalents. If a client’s stated need is friendship, this is the wrong tool, however familiar it is to them.

If you take nothing else from this piece: never refer a socially isolated client to Group C or D as a connection intervention. It looks like action and it is not.

Step 2: Be straight about the evidence

There is no trial evidence for any Group A product. No randomised trial, no controlled cohort, no validated loneliness instrument administered before and after. What exists is company-reported usage — Timeleft reports more than three million guests seated, Meetup reports around 60 million members — and journalism describing how the products work.

You are already working in a field where the evidence is thinner than the enthusiasm. Reviews of social prescribing for loneliness have consistently found small samples, weak comparison conditions and high risk of bias. The trial literature that does exist, such as the 2025 randomised trial of befriending in aged care, sits at some distance from a paid restaurant booking between adults in their twenties.

So the honest position is: this is an untested option, not a discredited one. That is a defensible thing to offer a client, provided you say which it is.

Step 3: The four cases where a referral is defensible

1. The client’s barrier is organising, not showing up. This is the specific problem Group A solves. If someone will attend things but cannot face arranging them — choosing, inviting, proposing a date, risking a lukewarm reply — an algorithm that hands them a confirmed booking removes exactly that step. If the barrier is anxiety about being in the room at all, Group A does not help and may make things worse.

2. There is nothing in your local directory that fits. Third-place availability has fallen across every category tracked, and the losses have concentrated in the areas that already had the least. If your directory has three options and all of them are daytime, or faith-based, or aimed at over-70s, a commercial product may be the only thing available to a working-age client. Say that is why you are suggesting it.

3. The client can absorb the cost. Most Group A products charge, and with Timeleft the restaurant bill sits on top of the booking fee. Recommending a paid product to someone whose isolation is partly financial is not a neutral act. Check first.

4. It runs where they live. City coverage is the binding constraint on the entire category. Buffet operates in Los Angeles with expansion planned; 222 restricts sign-up to ages 18 to 27. Check availability and eligibility before raising it, not after — a client who gets excited and then finds the app does not serve their city has been handed another small failure.

Step 4: Refer into Group B wherever you can

If a client meets the criteria above and both a Group A and a Group B option exist, send them to Group B.

Group A gives a new table each time; Group B gives the same table repeatedly. Recurrence is the mechanism, and only Group B has it built in. The cost is that Group B asks much more on the first night — there is no algorithm doing the social work, and walking into an established group as the unknown person is a real barrier for exactly the clients you are most worried about.

A reasonable sequence, where both exist: use Group A to rebuild the habit of going out, then move to Group B for the thing that lasts.

What this means in practice

Add to your referral protocol:

• A category label on every digital option in your directory, using the A/B/C/D split above. This prevents the Partiful error at the point of use rather than relying on each worker to know.
• A coverage and eligibility check before the option is raised with a client. Record city and age limits alongside the entry.
• A cost field, including secondary costs like the restaurant bill.
• A standard sentence for how these are offered: that it is an untested option, that the service is not endorsing the company, and that no one is expected to keep paying for something that is not working.
• Follow-up at the point that matters. Research on how people perceive social prescribing suggests the referral itself is often experienced as being passed along. Ask at week four, not week one — whether they went a second time. The second visit is the outcome worth recording.

And record the referral as what it is. If commissioners later ask which parts of your pathway have an evidence base, you want your own notes to distinguish the tested components from the plausible ones. Methodological commentary on this field keeps returning to the same point: the shortage is not evidence that loneliness matters, it is estimates precise enough to allocate against. Do not let your own service quietly manufacture the impression that this gap has been closed.