Marriage is primarily
a matching problem.
Our position: probably false. We have no evidence, and the counterarguments below are not answered.
Stated as the industry states it implicitly.
That the scarce, hard, valuable thing is finding the right candidate — and that everything after that is logistics.
Because the tool fits the belief.
Honestly: because the entire visible technology industry around marriage is built on it, and inherited belief is hard to see from inside.
Matching is also the part that is legible to software — a search problem over structured attributes, which is exactly what computers have been good at for forty years. An assumption that happens to match the available tool is rarely examined.
What the old system actually supplied.
- Family
- Community
- Information
- Intermediary
- Verification
- Decision
The part worth studying is not that it was slow. It is what it supplied. Information moved through overlapping social networks with reputational consequences for misrepresentation. Trust was transitive and enforceable because the intermediary had standing to lose. Verification was social rather than documentary. Families were parties to the decision, not spectators. And the months of coordination that followed were carried by people with an interest in completion.
Discovery got solved. The rest left the system.
Digital platforms took the discovery layer and made it enormously better: reach expanded from a dozen villages to a country. That is a real achievement and should not be minimised.
What they did not take: verification, which is still social or absent; family coordination, which moved to messaging apps; the negotiation; the months of execution; and everything after the wedding. The industry’s own metrics measure the part it solved.
None collected.
This section stays empty until it is not. What would fill it: where participants report the process actually failing; whether failures cluster at discovery or after it; the ratio of time and money spent on discovery versus everything else; what share of introductions from each channel complete; and what “worked” means to participants a year later rather than at the point of match.
If discovery turns out to carry most of the cost and most of the failure in the process, then the assumption holds and our architecture is wrong. Pre-registered prediction: under 30% of aggregate cost and failure attaches to discovery.
The strongest cases against us.
Including the one about ourselves, which belongs here rather than in a critic’s reply.
- A
The market has spoken
Large businesses were built on matching. If verification and coordination were the binding constraints, someone would have monetised them. Weak but not empty — it may instead show which parts are monetisable, not which parts matter.
- B
Discovery genuinely was the constraint
Geographic isolation collapsed the candidate pool. Solving that may have been the single largest available improvement, and the remaining problems only look primary because the first one was fixed.
- C
Our own selection bias
We build ecosystems, so we are predisposed to conclude the answer is an ecosystem. This is stated here rather than left for someone else to find.
- D
The remaining problems may not be system problems
They may be irreducibly human, in which case a coordination layer adds overhead without changing outcomes.
1850, today, and the boundary.
Most people build the matrimonial website
Given full modern technology and no modern institutions, three routes are available: digitise the register, build profiles and filters, or map the whole journey. Most people pick the second. That is the finding — not that it is stupid, but that it is what reasoning from the existing process produces.
What present technology changes is not matching
Verifiable identity makes documentary verification possible without a central registry. Payments make staged, conditional obligations enforceable. Coordination tooling makes a months-long multi-party process trackable. None of those are matching — and if the assumption were true, all of them would be marginal.
The cleanest falsification available
At the boundary, matching becomes nearly free. If matching is the primary problem and it becomes free, the system should look resolved. If the difficulty instead migrates to trust, coordination and execution, the assumption was wrong. This is partly observable already, in markets where discovery is effectively solved.
Four, and one of them is a serious rival.
- A — Better matching
- Status quo plus better models. Cheap, incremental, real gains. Leaves every post-discovery failure untouched. A genuinely serious rival, and not dismissed.
- B — Verification first
- Attacks the least-served layer. But verification without discovery has no entry point.
- C — Coordination first
- Attacks the longest, costliest phase. Heavy to build, and may be overhead people refuse.
- D — Full ecosystem
- Matches the actual journey. Largest, slowest, easiest to get wrong. This is what we currently believe, and that belief is the thing under test.
If it had never been a matching problem.
Most likely an institution for verification and enforceable commitment between families — with introductions as a by-product of being trusted, rather than the product. Whether that is better is unknown. That it is a coherent alternative history is the point.
Designed. Not run.
Method, public: follow a cohort through the full journey, from first introduction to twelve months after the wedding, recording where time, money and failure actually concentrate — and specifically whether stated dissatisfaction traces to the match or to what happened after it.
Location, partners, recruitment and operational design: not public. The method is the part that can be checked; the operating geography is not ours to publish.
Result: not run.
2026-09-12
“Marriage is primarily a matching problem.” Status: under investigation. Our position: probably false, but we have no evidence and the counterarguments are unanswered.