I recently went looking for songs I could actually sing. I checked around 350 of them, tagged about 120 as possibilities, and after singing, transposing, practicing, and listening to myself, ended up with eight on my serious shortlist.

Eight out of 350 is 2.3 percent.

At first that seemed like a fact about karaoke. Then I recognized the funnel.

Over the last few years, I have met roughly 300 people while deliberately trying to expand my friendships. Lots of them were perfectly nice. Many were interesting. Some became friends. But if I use a much harder standard and count the guys who became unusually strong fits for me, I get about six.

Six out of 300 is 2 percent.

Maybe real fit is a 2 percent proposition.

Quality Requires Something to Choose From

We are constantly told to choose quality over quantity. It sounds wise because nobody wants 50 mediocre friends, 50 mediocre employees, or 50 mediocre songs. But the advice skips an important question.

Where does the quality come from?

Suppose you are hiring for a mid-level position. You might reasonably want 200 resumes to make one hire. You certainly don't want to spend three hours interviewing every applicant. You scan quickly, reject most of them, talk briefly to some, interview a handful, spend more time with the finalists, and eventually invest years and hundreds of thousands of dollars in one employee.

The cost of evaluation rises as the evidence of fit rises.

Sales works the same way. Sales trainers have been saying "make the calls" forever. It can sound like motivational boilerplate until you think about the underlying math. If only a small percentage of prospects are ever going to be good fits, five calls tell you almost nothing. You need a field large enough for the fits to appear.

The trick is not learning to enjoy rejection. It is building a process where rejection barely matters.

At Reclaim the Money, we start with an enormous universe of records, identify leads, identify the leads economically worth pursuing, contact those people, and eventually sign a tiny fraction. We could never afford to treat every record as a serious prospect. The business depends on eliminating almost everyone cheaply.

Get very good at losing candidates cheaply.

Several Ordinary Requirements Can Produce a Rare Fit

There doesn't have to be anything mystical about 2 percent.

Real fit is usually multidimensional. Imagine that only 30 percent of candidates pass one meaningful requirement, 30 percent pass another, and 20 percent pass a third. If the requirements are reasonably independent, only 1.8 percent pass all three.

Nothing individually was terribly rare. Their intersection was.

A spouse may need to fit on attraction, temperament, values, lifestyle, intelligence, sex, geography, family plans, money, availability, and reciprocal attraction. A friend needs fewer things to line up, but still enough that the intersection can become small quickly. An employee can be talented and still be wrong for the job, manager, compensation, location, culture, or moment.

The same thing happens with something as trivial as a song. I can like a song that doesn't fit my voice. I can sing a song that doesn't fit my personality. A song can fit both but do nothing for a room. Several things have to be true at once before I want to keep it.

The 2 percent number is not important enough to turn into a law. The useful assumption is that real multidimensional fit may be much rarer than we intuitively expect.

If you expect one in five and encounter twenty misses, you may conclude something is wrong.

If the real number is closer to one in fifty, twenty misses barely constitutes a search.

Technology Makes Quality Cheaper

There is another reason I could find those eight songs.

Checking 350 songs would once have been ridiculous. I would have needed records, radio play, CDs, or a patient employee at a music store. Today a karaoke box, an app, Spotify, Pandora, and other music services let me check songs almost instantly. I can jump to the chorus, try a few lines, change the key, decide it isn't mine, and move on.

Technology didn't necessarily make me better at choosing songs. It made losing songs incredibly cheap.

Dating apps did something similar to potential partners. Job sites did it to employment listings. Real estate sites did it to houses and apartments. Search engines did it to information. Modern databases and automated communication did it to sales prospects.

When the cost of examining and rejecting candidates falls, you can examine more candidates. When you can examine more candidates, you can afford to become more selective.

Cheap rejection can produce higher quality.

Dating Makes the Cost Curve Obvious

Modern dating is almost a caricature of the process.

A person might see hundreds of profiles, pause on dozens, match with several, chat with a few, go on a handful of dates, have second dates with fewer, and eventually build a life with one person.

The striking feature is not just the shrinking numbers. It is the increasing cost.

A swipe costs a second. A message costs minutes. A first date costs hours. A relationship can consume years.

That is exactly how a good selection system should allocate resources. Spend almost nothing on early candidates. Spend progressively more as evidence accumulates. Invest enormously once real fit has survived enough filters.

The mistake would be spending relationship-level resources on swipe-level candidates.

We do versions of that everywhere. We research every plausible job for hours before applying. We practice a mediocre song endlessly because it was one of the first ten we tried. We keep working a sales lead that doesn't want us. We schedule the sixth dinner with someone we weren't particularly excited about after the second.

Sometimes perseverance is admirable.

Sometimes we have simply built an expensive rejection system.

Experts May Throw Away More, Not Less

Creation seems at first like an exception. Surely a sculptor doesn't inspect 300 sculptures and select one. She creates one.

But where did it come from?

She has seen thousands of bodies, sculptures, photographs, poses, materials, and forms. She may sketch alternatives, change proportions, reject ideas, make studies, and remove material until one form remains. The finished work is singular. Its ancestry is not.

Writing works the same way.

A strong 1,000-word piece between an experienced writer and an AI collaborator can easily emerge from 30,000 words of exploration, arguments, false starts, alternate headlines, examples, objections, jokes, and discarded formulations.

That means roughly 97 percent of the words don't survive.

Experience hasn't eliminated quantity. It has made us faster at generating possibilities and better at killing them.

The novice imagines expertise means producing the right answer immediately. Often expertise means recognizing the right answer quickly when it finally appears.

Knowing good stuff when you see it is valuable.

Good luck seeing it if you only look at twelve things.

Gold Miners Already Know This

Gold mining makes the whole idea almost embarrassingly obvious.

A miner doesn't choose a promising-looking rock and spend the afternoon trying to turn it into gold. Mining works by processing enormous quantities of material, discarding almost all of it, and progressively concentrating the tiny fraction that contains what you want.

Pan it. Screen it. Separate it. Concentrate it.

The worthless material isn't evidence that the process failed.

Processing worthless material cheaply is the process.

Nobody tells a gold miner to stop obsessing about quantity and concentrate on quality. The bowl of gold exists because a vastly larger quantity of material passed through the system.

Human fit is obviously not gold mining, but the search architecture is surprisingly similar.

The quality was rare. The field was large. The losses were cheap. Investment increased as promising material survived.

There Is a Boundary

This does not apply to everything.

You don't screen 300 biological children and select the one who fits you best. You don't audition parents before being born. Some of life's most important relationships are assigned rather than selected.

When there is no candidate choice, development, adaptation, acceptance, and repair have to do work that selection cannot.

But notice how quickly the quantity principle returns when choice returns. Prospective adoptive parents can be evaluated. Schools can be compared. Treatments can be considered. A child can try several sports before discovering one that fits.

The principle belongs specifically to problems where quality must be selected from heterogeneous possibilities.

Quantity Is the Raw Material of Quality

This takes me back to friendship.

I originally started thinking seriously about adult friendship because I wanted more strong friendships and wasn't finding enough of them. Much friendship advice focuses on what to do after meeting someone: be vulnerable, listen well, initiate, show up, invest, repair.

All useful. But none solves a candidate shortage. If real friendship fit is anywhere near 2 percent, meeting twenty people and failing to find a great friend is not evidence that adult friendship is impossible. You may simply have searched less than half of the expected field required to find one.

Want five unusually good fits? You may need exposure to hundreds of people.

That sounds exhausting until you realize the solution isn't to invest deeply in hundreds of people. It is exactly the opposite. Meet broadly. Filter cheaply. Let weak fits disappear without drama. Spend progressively more time on the people who keep surviving. When something exceptional finally appears, invest.

The same principle can find an employee, a customer, a spouse, a song, or a friend.

Quality over quantity is excellent advice once you have found quality. It is terrible advice for finding it.