Don’t Quit Dating Apps—Quit the Swipe Factory: Redesign Dating Search Around Low CPA and More Real-World Trials
Conclusion: dating is not only a popularity contest; it is also a search-design problem
If dating apps are exhausting you, the answer is neither “swipe through another thousand profiles” nor “all apps are trash, delete everything.”
A more useful answer looks suspiciously like process engineering:
Keep apps as low-frequency search tools, move the main evaluation into real conversations, and avoid expensive one-shot events when subsidized or low-cost in-person options let you run more trials for the same budget.
The point is not to rank people like inventory. It is the opposite: stop working as the quality-control clerk of an infinite human e-commerce catalog and observe the parts of compatibility that only exist once two people actually talk.
This article borrows the advertising term CPA (cost per acquisition) only as a metaphor. Here it means something like “total search spending per mutually desired second date with someone who also clears your non-negotiable values.” Nobody is being “acquired.” We are measuring the cost of the search process.
1. Why the swipe factory gets exhausting
Profile. First photo. Second photo. “Hmm.” Next. Next. Next. Do that hundreds of times and dating starts to resemble incoming-goods inspection.
Research supports part of this experience. In Pronk and Denissen’s work on the rejection mind-set, participants became progressively more rejecting while evaluating online-dating candidates. Across three studies, the probability of accepting a candidate fell by an average of 27% from the first to the last option.
That does not mean “using an app reduces your chance of romantic success by 27%.” It means repeated sequential evaluation itself was associated with a growing tendency to reject.
More choice is therefore not automatically more useful choice. An infinite shelf can make you increasingly good at finding reasons not to pick anything. The swipe factory may become harsher the longer the shift runs.
2. When attention concentrates, pre-meeting traits get overpriced
A 2025 conjoint study analyzed 5,340 swiping decisions made by 445 online daters. A one-standard-deviation increase in physical attractiveness increased selection success by roughly 20%, while the same increase in intelligence improved it by about 2%.
Bruch and Newman also found a pronounced desirability hierarchy in large online dating markets across four U.S. cities.
The sensible conclusion is not “looks are everything.” It is that traits that are easy to observe before meeting can become disproportionately powerful in a market that makes decisions before meeting.
Voice, timing, humor, warmth, listening, the ability to pick up conversational threads, and the ability to make a stranger comfortable are difficult to price from a static profile.
If your evaluation tends to rise after five minutes of conversation, competing only in a profile-card market is like taking an exam after removing your strongest subject from the syllabus.
Exact concentration figures for a specific app should not be invented unless the platform publishes them.
3. Match count is a weak KPI
Ten matches can turn into three people worth messaging, one sustained conversation, and zero viable partners once a hard-value mismatch appears.
Track the funnel instead:
| Stage | Metric |
|---|---|
| Exposure | Profiles viewed |
| Contact | Unique people actually spoken with |
| Interest | People you would like to see again |
| Reciprocity | People who also want to see you again |
| Fit | People who clear important values and logistics |
| Action | Actual second dates |
| Outcome | Ongoing relationships |
Likes are not currency. Matches are not revenue.
A more useful metric is “second dates per ¥10,000” or “cost per serious candidate.”
4. Move hard filters upstream
If topics such as children, marriage intention, smoking, geography, or family boundaries are true non-negotiables, confirm them early enough to avoid expensive downstream failure.
Viewing 100 people, matching with 10, talking to three, dating one three times, and then discovering a fundamental incompatibility is not merely sad; it is costly search architecture.
But do not diagnose strangers from photos or profiles. Labels such as “unstable” or “low tier” are both ethically poor and statistically unreliable. Evaluate stated values, observable behavior, and actual conversations.
5. If you are stronger in person, make in-person interaction the main event
Joel, Eastwick, and Finkel collected more than 100 pre-date traits and preferences from speed-dating participants and then used machine learning to predict romantic attraction.
Models could predict some general tendencies—who tends to desire others and who tends to be broadly desirable. But they could not predict the relationship-specific variance: the unique “these two people click” effect. Participants had to meet for four-minute dates for that information to appear.
That is a strong reason to preserve real-world trials.
| Channel | Pre-filtering | Uses in-person skill | Typical cost pattern | Main weakness |
|---|---|---|---|---|
| Dating app | Strong | Low–medium | Subscription | Swipe fatigue, attention concentration |
| Expensive singles event | Medium | High | High per event | Large loss when the draw is poor |
| Subsidized/local event | Medium–strong | High | Can be very low | Lotteries, irregular schedule |
| Friend introduction | Sometimes strong | High | Low | Small pool |
| Bar/cold approach | Weak | High | Variable | You must screen relationship status and intent yourself |
You do not need to abandon apps. Think app = search engine, in-person = finals.
6. Sometimes talking to ten people costs little more than a premium sort feature
As of September 2026, with’s male VIP add-on was ¥2,900 per month by credit card and ¥3,500 via in-app purchase, on top of the paid messaging plan.
Premium discovery tools can save time. But if a few thousand yen can instead buy structured conversations with several people at a low-cost event, someone whose strength appears in person should compare those marginal returns.
Apps still provide useful filtering. with’s rules prohibit users who already have a partner and users seeking something other than romance or marriage. Its terms also explicitly state that the company cannot guarantee that every member is single. Filtering reduces work; it does not eliminate verification.
7. Split a ¥10,000 budget into small experiments
Instead of paying ¥9,000 for one giant draw, an illustrative budget might be:
- two ¥1,000 municipal events = ¥2,000
- two ¥3,500 small 1-to-1 events = ¥7,000
- reserve = ¥1,000
This is a budgeting principle, not a promise that four suitable events will always exist at those prices.
A practical heuristic:
| Grade | Rough rule |
|---|---|
| SS | Free–¥1,500, single/eligible participants, multiple real conversations |
| S | Up to ¥3,500, 1-to-1 format, likely 6+ conversations |
| A | Up to ¥4,500, all-participant rotation or a strong interest filter |
| Usually skip | Over ¥6,000 with unclear attendance or conversation structure |
Adjust for travel, age range, food value, and your own constraints.
Fine dining is not the enemy. Have the expensive French lunch after you know who you are having it with. Do not automatically spend fine-dining money on an unknown event draw.
8. Public matchmaking is infrastructure, not a tax exploit
In fiscal 2026, Aichi Prefecture offered subsidies of up to ¥200,000 per registered organization for eligible matchmaking events listed on its official Aicon Navi portal. Eligible expenses can include venue rental, advertising, printing, and certain food costs.
That does not prove any specific ticket price was reduced by a specific amount. It does show that a publicly supported market exists whose purpose is to lower barriers to meeting.
Examples as of September 10, 2026:
- Moricoro Park Mega Matchmaking 2026: free, capacity 400; 2,640 applications, a 6.6× application ratio.
- Chita Peninsula En-musubi, Oct. 25, 2026: ages 25–39, 20 men and 20 women, ¥1,000.
- Kariya events for people in their 20s and 30s: examples with 10 men and 10 women, 1-to-1 conversation, male price around ¥3,500.
- Mie’s “Miemusubi”: registration and system use are free for eligible marriage-seeking singles; optional supporter attendance at a face-to-face meeting costs ¥1,000.
- Nagoya City: announced three matchmaking events planned for roughly November 2026 through January 2027; details were still pending at the time of writing.
Cheap programs can be oversubscribed—the Moricoro event proves that. The hidden cost becomes lottery risk.
If you meet the eligibility requirements, using a program designed to reduce participation barriers is simply using the infrastructure as intended.
9. Long-tail does not mean attending the same venue every week
Repeatedly using one organizer, one neighborhood, and one age band can produce repeat participants.
Rotate the market:
prefecture → municipality → NPO → chamber/community group → subsidized private organizer → hobby event → small 1-to-1 → large event.
Rotate geography and timing as well.
The metric is not event count. It is new unique people with whom you had a real conversation.
The long tail is the collection of small, cheap, overlooked entry points across different pools.
10. Automating swipes can merely move the pain downstream
When browsing becomes tedious, automation is tempting. But automating only the sending step may create this pipeline:
automatic likes → more matches → profile review after matching → select the few you actually want → more message work.
The filtering burden moved; it did not disappear.
Automation that violates platform rules also creates account risk. A safer optimization is to use native filters, short sessions, and explicit stopping rules. View 10–15 profiles and stop. If you already have two promising conversations, pause new search until they resolve.
Every factory needs an off switch.
11. Research communicators also have incentives
If an influencer or expert who explains dating research has supervised, advertised, sold, or otherwise been connected to a dating product, that does not automatically make the explanation false.
It does mean you should consider which evidence they have incentives to emphasize.
with was historically marketed as supervised by Mentalist DaiGo. That does not establish that DaiGo owned the service, nor does it prove he intentionally suppressed the rejection-mind-set paper. No evidence for that stronger accusation is established here.
The useful rule is information hygiene:
- Open the cited paper.
- Find reviews and contrary findings on the same topic.
- Check commercial relationships.
- Verify the scope of the statistic.
If someone says “research shows 27%,” ask: 27% of what?
12. Use one scorecard
| Metric | Record |
|---|---|
| Total cost | Ticket + travel + necessary costs |
| Unique conversations | People you actually talked with for meaningful time |
| Want to meet again | Your yes count |
| Mutual yes | Both sides want another meeting |
| Hard-filter pass | Important values/logistics align |
| Second dates | Actually happened |
| CPA2 | Total cost ÷ second dates |
If second dates are zero, record zero rather than manufacturing an infinite-looking precision metric.
After three to five trials per channel, your own data begins to beat generic advice.
13. Limits and ethics
Low CPA does not mean “find cheap people.” It means reduce waste at the entrance to the search process.
A dating partner is not inventory. The other person is also choosing you. The real success metric is mutual willingness to continue.
More trials also require more courtesy, not less. Eligibility rules for public events—age, residence/work/study, single status, marriage intent—must be respected.
14. Summary: stop the swipe factory and talk to actual people
The problem is not that apps are inherently evil. The problem is that rapid profile evaluation and real-life romantic compatibility measure different things.
So:
- use apps as low-frequency, hard-filtered discovery tools;
- track the funnel to second dates, not raw likes;
- if you are stronger in person, move the main evaluation offline;
- hunt for free and low-cost public/local events;
- rotate organizers, geography, and formats;
- split ¥10,000 across trials rather than one expensive draw;
- return to primary research when claims sound too neat.
Likes are not currency. Eat the expensive French meal after you know your date. And yes, you are allowed to shut down the swipe factory.
References and official sources
- Pronk & Denissen (2020), https://doi.org/10.1177/1948550619866189
- Witmer, Rosenbusch & Meral (2025), https://doi.org/10.1016/j.chbr.2024.100579
- Bruch & Newman (2018), https://doi.org/10.1126/sciadv.aap9815
- Joel, Eastwick & Finkel (2017), https://doi.org/10.1177/0956797617714580
- Aichi matchmaking-event subsidy: https://www.pref.aichi.jp/soshiki/kosodate/konkatuhojyo.html
- Moricoro Park 2026 results: https://www.pref.aichi.jp/soshiki/kosodate/konkatsu2026.html
- Aicon Navi, Chita event: https://aiconnavi.jp/event/detail/12761
- Aicon Navi, Kariya event: https://aiconnavi.jp/event/detail/12667
- Mie “Miemusubi”: https://deai-mie.jp/aimatching-miemusubi/
- Nagoya “NAGOYA Hachimusubi”: https://www.city.nagoya.jp/houdou/3003942/3003946/3005098.html
- with VIP pricing: https://support.with.is/hc/ja/articles/5384014443289
- with guidelines/terms: https://with.is/safety_center/community_guidelines / https://with.is/terms_of_use
