Why We Built Self-Service Therapist Matching

Before self-service matching existed, starting therapy on YouMindo meant scrolling a directory of therapist profiles and picking one, usually based on whoever had a friendly-looking photo and an opening in their schedule. It worked, in the sense that people did pick someone. It didn't work in the sense that a meaningful share of first-time clients never made it to a second session.

When we dug into why, the answer wasn't what we expected going in. It wasn't cost, and it wasn't scheduling friction, both of which we'd assumed were the bigger barriers. It was that people didn't feel equipped to choose a therapist at all, so they either picked almost at random off a list of bios, asked a friend to pick for them, or closed the tab and never came back to finish signing up. That last group was the one we built the quiz for.

What the old directory actually asked of people

Browsing a directory sounds simple, but it quietly asks a new client to do something most people have never done: evaluate a therapist's fit for them based on a few paragraphs of professional bio. Specialty tags helped a little. They said nothing about whether someone wanted a therapist who challenges them directly or one who leads gently, whether language mattered, or whether age or gender of the therapist would affect how safe someone felt being honest.

Building the intake quiz

The matching quiz asks about six things: primary concerns, language, gender preference, age range, prior therapy experience, and preferred therapeutic modality. Each answer is optional except concerns and language — we didn't want to force a preference on anything where 'no preference' is a legitimate and common answer. The quiz takes under three minutes, which was a hard constraint from the start; every version that ran longer saw a meaningful drop in completion partway through.

How the scoring actually works

Each therapist profile is scored against a client's answers across the same six dimensions, weighted so that stated concerns and modality experience count for more than softer preferences like age range. We resisted the temptation to keep adding dimensions as we thought of them — every additional input is a chance to overfit to preferences that sound meaningful but don't actually predict a good working relationship, and we didn't have the outcome data to justify adding more than we already had.

The debate about weighting gender and language

Internally, we argued about how much weight to give gender and language preference relative to clinical specialty. Some of us worried that leading with those factors over specialty risked steering people away from the therapist best equipped to help them clinically. Our clinical advisors settled it: if someone doesn't feel safe enough with a therapist to be honest, clinical specialty match doesn't matter. Comfort is a precondition for the clinical work, not a secondary preference.

Why we show three matches, not one

We debated showing a single top match, styled as a confident recommendation. We chose three instead, ranked but not dramatically differentiated in presentation, specifically so the quiz reads as a narrowing tool rather than an authority making a decision for you. People still browse the full directory afterward if they want to — the quiz shortens the list, it doesn't replace the choice.

The dating-app version we didn't build

Early design explorations leaned toward a swipeable, photo-first browsing experience, because it tested well for speed and engagement. We killed it before it shipped. Optimizing for a fast decision is the wrong goal for choosing a therapist — a decision made quickly, based mostly on a photo, is exactly the kind of decision that produces the mismatches we were trying to fix in the first place.

The mistake: burying the override

Our first release put the 'browse all therapists instead' option in small text at the bottom of the results page. Usage data showed almost no one found it, which meant the small number of people the quiz genuinely didn't serve well had no visible way out. We moved it to a persistent link at the top of the match results. It's a small placement change that mattered more than most of the scoring logic underneath it.

What the data shows so far

Clients matched through the quiz have meaningfully lower first-session no-show rates and are more likely to still be with the same therapist eight weeks later, compared to the directory-browsing cohort from before the quiz existed. We're cautious about reading too much into this — people who complete a three-minute quiz may simply be more engaged to begin with — but the direction is consistent enough that we've kept investing in it.

What matching can't fix

No quiz can substitute for the first session itself, where fit actually gets tested. A client can score as a strong match on paper and still find, in the room, that the working relationship isn't right — and switching therapists remains a normal, encouraged option, not a failure of the algorithm. We built the quiz to improve the odds of a good first match, not to guarantee one.

Matching well doesn't make therapy work. It just removes one more reason for someone to give up before it gets the chance to. That's a smaller claim than 'we found your perfect therapist,' and it's the honest one — but for the people who stopped at the directory page before, it's turned out to be enough of a difference to matter.