HealthTech · EdTech · 2024 — 2025
International medical graduates apply to US residency programmes with almost no reliable signal about which ones will actually consider them — so most unmatched candidates spend their fees on programmes that were never going to open the file. Match A Resident replaces that guesswork with programme requirements verified by calling all 4,575 ACGME programmes every year.
Executive Summary
A residency application is expensive, and an applicant only gets one cycle a year. The decision that determines the outcome is not how good the personal statement is — it is which programmes end up on the list.
Most of that list-building happens on rumour: forum posts, last year's spreadsheet, what a friend heard. Programme requirements are not reliably published, and they change. An applicant can spend a full cycle applying to programmes that filter them out on visa status or graduation year before a human reads a word.
Match A Resident's answer is unglamorous and hard to copy: call all 4,575 ACGME programmes every year, ask what they actually require, and build the compatibility list from that. The surrounding services — personal statement review, MyERAS review, interview prep, advising — sit on top of that data spine.
Section 02 — The Problem
What a programme says publicly and what it filters on are different things. Visa sponsorship, years since graduation, attempt history — the criteria that actually decide are often not on the website at all.
Applicants pay per programme. Applying to one that will never review the file is not a small inefficiency — it is spend removed from the applications that could have worked.
The match runs annually. A list built on bad information does not fail gradually — it fails once, completely, and the applicant waits a year to try again.
Section 03 — Research & Discovery
Plenty of products could score an applicant against a programme list. What is genuinely hard is having a programme list that is true this year — and the only way to get it is to ask every programme, every cycle, and record the answer.
That reframes the product problem. The engineering question is not how clever the matching is; it is how to keep a verified dataset of 4,575 programmes across 18 specialties current, and how to turn it into a compatibility decision an anxious applicant can act on in thirty seconds.
By the time an applicant is filling in ERAS the expensive decision is already made. The research that changes the product is about the weeks before that — where the candidate list comes from, who they trust, and what makes them add a programme they have no realistic chance at.
A dataset refreshed by thousands of phone calls a year is not a background task — it is a workflow with throughput, coverage and quality problems of its own. Treating the internal tooling for it as a first-class product surface is what keeps the moat from decaying quietly.
Section 04 — Current State Analysis
Section 05 — Competitive Research
The category is crowded with advisors, courses and forums, and all of them are cheaper to start than this. What none of them can produce quickly is a programme dataset verified by direct contact every year — the cost of building it is the reason it is defensible.
That shapes where the product should compete. Matching advice quality with everyone else is a losing race; the argument is that the list itself is built on something checkable, and everything else is in service of it.
Most applicants are not choosing between paid services. They are choosing between paying for a list and assembling one from forums, spreadsheets and last year's advice — which is free, feels sufficient, and is exactly the behaviour that produces a wasted cycle.
"We call all 4,575 programmes every year" is doing the entire persuasive job on the homepage, and that is deliberate. A moat the buyer only appreciates after paying is a moat that never converts anyone.
Section 06 — Key Insights
The entry point asks for a profile and returns matches in about thirty seconds. Anyone deciding where to spend thousands of dollars will not first invest an hour in a form.
Applicants do not need programmes sorted by prestige. They need to know which ones will not discard them on visa status or graduation year — a different, more useful output.
ERAS opens on a fixed date and the match runs once. Demand is not smooth, and a feature that lands after submission opens has effectively missed a year.
Section 07 — Design Strategy
A user under this much stress needs an honest filter, not reassurance that everything is possible.
Every extra question before the first result costs conversions from people who are already overwhelmed.
There is one submission window a year; anything that lands after it waits twelve months to matter.
The dataset is the product. Everything else can be rebuilt; a missed call cycle cannot.
Section 08 — The System
The applicant clears the programme's stated requirements for this cycle. Worth the application fee.
Some criteria met, others uncertain or unverified. Needs the applicant's judgement rather than the system's.
The programme's own stated criteria exclude this applicant — most often visa sponsorship or graduation year. The most valuable answer the product gives.
The third state is where the money is saved. A product that only surfaces good matches is pleasant; one that reliably tells an applicant where not to apply is worth paying for.
Section 09 — The Features
The applicant supplies scores, graduation timing, visa needs and specialty interest; the generator returns compatible programmes with scoring, drawn from the dataset refreshed by calling every programme annually.
Personal statement drafting and review, MyERAS review across all 19 components, 70+ interview practice questions with feedback, programme-specific answers, letter templates, one-to-one advising with resident physicians, and SOAP preparation for applicants who go unmatched.
Section 10 — User Flow
A single branching question that changes almost every downstream requirement, asked first because it costs nothing.
USMLE scores, graduation timing, visa needs, specialty interest — the criteria programmes screen on, not biography.
Matched against the verified dataset, so the output reflects what programmes said this year rather than last year's rumour.
Where the product converts into a decision. The list has to leave the platform in a form the applicant can actually apply from, or the value stops one step short of the outcome.
Section 11 — End-User Experience
This product is used at one of the most stressful moments of a medical career, often by someone applying from another country, in a second language, against a system whose rules are not written down anywhere they can reach.
That changes what good means. Plain language beats polish, an honest no beats a hopeful maybe, and every unnecessary question is a reason to abandon. The thirty-second promise on the entry point is not a growth tactic — it is an acknowledgement of the state the user is in.
The place to watch is the profile step. Every additional field improves the match and costs conversions from people who are already overwhelmed, and that trade-off is the whole product design problem in one screen.
Section 12 — Impact & Outcomes
The figures above describe the product, not a result you produced — the service has been running since 2004 and has guided 240,000+ IMGs over that period, most of it before any involvement of yours.
Outcome figures for this period are not published, so none are claimed. The measures that would settle it are list-purchase conversion, profile completion rate, and — the one that actually matters to an applicant — match rate among users against the unmatched baseline.
Section 13 — Reflection
The thirty-second promise is the product's core bet, and it is testable at a much finer grain than it was treated. Knowing which specific field loses people would turn an argument about form length into a decision with an answer.
Programme requirements change during the cycle, not only between cycles. An annual call campaign is a huge advantage over scraping and still leaves a window where the data is confidently out of date — and the product does not currently show the applicant how fresh any given answer is.