A few weeks ago I listened to a group of friends vent about dating apps: the endless swiping, the false starts, the fatigue. I could relate. I met my husband online 14 years ago. It worked. But the process was stressful and inefficient enough that I would never go back to it.
What struck me is that the matching problem goes well beyond dating. It shows up in college admissions, job searches, sales, and startup fundraising. New graduates send hundreds of applications. Founders pitch hundreds of times before they find the right investor, while venture capital has roughly a 1–2% hit rate.
Everyone is searching. Very few are matching. Why is that, when we have more access and more powerful tools than ever?
Matching isn’t ranking
Most people treat matching like a ranking problem: find the best candidate, the best job, the best partner. But matching is two-sided, and often multi-sided. The “best” doesn’t exist on its own.
A great employee at one company can fail at another. A promising startup may not fit a given fund’s thesis. A wonderful person may be the wrong partner for you, specifically. “Good” only means something once you specify good for whom, in what context, at what moment.
The math is unforgiving. N candidates times M opportunities, each side carrying its own preferences along many dimensions, plus a long tail of factors nobody can directly observe. The search space explodes.
More options don’t make the problem easier. They make it harder, both for the algorithm and for the human.
This is why people say, “There are so many options, but none feel right.” It isn’t a paradox. It’s what you should expect when you’re searching across many dimensions with incomplete information.
The signal is mostly marketing
On paper, this should be a solved problem. We have unprecedented access to candidates, jobs, and investors. We have rich profiles, ranking algorithms, and recommendation systems. And now AI can write the résumé, polish the profile, and draft the pitch deck.
And yet outcomes haven’t kept pace. In many cases they’ve gotten worse. The reason is simple: most of what looks like data is actually marketing.
Everyone has a reason to present the best possible version of themselves. Candidates inflate their experience. Startups round up their traction. Investors perform selectivity. Dating profiles surface an idealized version of a person who, in practice, does not quite exist.
The result is a system flooded with noise. Top-tier profiles all start to look the same. Real quality gets harder, not easier, to spot. We don’t end up with more clarity. We end up with more sameness.
People keep changing while they search
Most matching models assume preferences are stable. Real life violates that assumption almost immediately. People’s tastes evolve as they search. Their behavior shifts as they learn what plays well. And many of them don’t actually know what they want, even when they think they do.
That sets off feedback loops. Over-optimization pushes everyone toward the same look. Inflated signaling erodes trust. Endless exploration delays commitment. Participants aren’t just being matched. They’re reshaping themselves, and the system, in real time. The whole thing never quite settles.
What AI actually fixes, and what it doesn’t
AI is very good at efficiency. It can process more candidates, sharpen rankings, polish presentation, and take a lot of friction out of the search. What it cannot do is fix the inputs. It doesn’t make people more honest about who they are or what they want. It doesn’t deepen compatibility. It doesn’t build trust. And it has no special insight into whether a match will still feel like a match in three years.
Often, AI makes the underlying problem worse. Better-generated profiles are harder to tell apart. Scaled outreach drowns the channel. The optimization target is engagement, not outcome. We’re pointing very powerful tools at a system whose objective function was never aligned with ours to begin with.
What actually moves the needle
If technology alone can’t close the gap, what can? A few things, in my experience.
Constraint, not more options.
Past a certain point, more options make matching worse. The systems that work well introduce deliberate friction: curated pools, warm introductions, structured funnels. Costco succeeds partly by offering fewer choices than a traditional supermarket, not more. Constraints aren’t a limitation. They’re how you raise the signal.
Know what you actually want.
Matching gets easier when people are honest about their own constraints and priorities, willing to state their non-negotiables out loud, and not trying to be palatable to everyone. The fuzzier your preferences, the lower the odds of a good match. Clarity, on your own side, is half the work.
Trust channels outside the system.
The best matches I’ve seen usually skip the open marketplace entirely. They come through referrals, trusted networks, and repeated interactions over time. Those channels cut through the information asymmetry, introduce real accountability, and carry signals that no profile can. It’s why who you know still beats what the system shows you.
The bottom line
Matching is hard because it combines incomplete and strategic information, compatibility across many dimensions, and human preferences that shift while you’re trying to read them.
AI can accelerate the search. It cannot resolve those structural constraints.
The problem was never that we don’t have enough options. It’s that we don’t have enough clarity, honesty, and constraint inside the options we have.
Until those improve, matching will stay inefficient no matter how clever the tools become.
Matching has never really been about maximizing quantity. It’s about finding one good alignment inside a system designed to obscure it. That was true 14 years ago when I met my husband, and it’s still true now.
Other Posts You Might Like
