Andy Ke

My own productStarted October 2026, pre-pilot

An assistant coach that remembers every student

Coaches who run small group sessions plan tonight's courts in their heads and keep each student's progress in memory. I started CourtPlan to carry both. It is pre-pilot, with no users and no revenue yet. This page covers the problem, the field, the design, and how I'll test demand before writing production code.

Stage

Pre-pilot

A clickable prototype exists; production code waits for a two-week hand-run test.

Design target

Under 2 min

A design target for coach effort per session, not a measured result.

Price hypothesis

$15–30

A price hypothesis of $15–30 per coach a month, tested in Phase 0.

How I built it. I did the market research, the competitive scan, the design brief and the clickable prototype solo with Claude Code.

I paused AnswerClip, my earlier idea, because its demand check came back too thin, and moved to a problem two coach friends who run small group sessions face.

Why this, why now, why me

Why this problem

Three jobs no single tool handles together

Coaches with 2 to 12 students on 1 to 4 courts plan tonight's session in their heads at the fence, 5 minutes before start. Each student's progress lives in memory and gets lost after two missed weeks. Progress updates keep students paying, yet rarely get sent. This is my hypothesis from the 2 pilot coaches; Phase 0 interviews check it against their last sessions.

Why now

A voice note can become a record

Transcription and structured LLM output can turn a 60-second voice note into one entry per student. An LLM limited to the coach's own drill library can draft a plan, while code enforces the court and grouping rules. The coach talks; the software files.

Why me

Two pilot coaches and a fast build loop

I'm an engineer turned PM who builds solo with Claude. Two coach friends who run small group sessions are lined up as my pilots, and they already shaped the product: fee tracking came from them.

I scanned 12 tools: they know who booked, not what to run

Features and prices as seen on October 13, 2026. A selection of the 12 tools scanned.

ToolWhat it doesWhat it misses
OutcoachJournal, parent app, voice-to-text, fee processing and bookings for tennis coaches.No session planner found. The biggest threat if it adds one.
Booking tools (CourtReserve, Playbypoint)Court bookings and payments for clubs.Nothing about what to run on court or how each student is doing.
CoachNowVideo coaching with plan templates.No plan built from tonight's attendance and courts.
Sportplan, PracticePlannerDrill libraries and practice plans for teams.No attendance, court split or memory of each student.
WhatsApp, Sheets and VenmoThe free default: messages, a spreadsheet and payment apps.All manual and scattered. Progress is lost when a student skips.
ChatGPT and other chat assistantsDrafts a session plan on request.No roster, attendance, court rules or memory between sessions.
CourtPlanTurns attendance and courts into a plan from the coach's own drills, a voice note into a journal, and tracks fees after class.No bookings, payments, student app or video, by design.

The open space is planning plus memory: attendance and courts turned into groups and drills from the coach's own library, and a journal that remembers. No bookings to migrate, so it sits beside existing tools.

Market size, bottom-up: coaches times price

TAM

~$75M

Total market (TAM): about 215k racquet coaches worldwide (ITF and PPR counts, plus a padel estimate from FIP club counts) at $29 a month, the top of my $15–30 price hypothesis, gives ~$75M a year.

SAM

~$12M

US serviceable market (SAM): about 35k US certified tennis and pickleball pros (USPTA, PTR and PPR) at $29 a month gives ~$12M a year.

SOM

~$0.7M

Year-3 obtainable market (SOM): ~$0.7M a year. This is an assumption, not a forecast.

The biggest unknown is the share of coaches who run small groups. There is no public data, so I'll ask in the Phase 0 interviews. All three use $29 a month, the top of my price hypothesis, so treat them as upper bounds.

The design: one loop, four moments

  1. Before

    Attendance becomes a plan

    The coach taps who showed up and how many courts there are. CourtPlan splits groups by level and picks drills from the coach's library for each court.

  2. On court

    A timer runs the session

    A live timer counts down each block and shows the drill and cue for every court. One tap adds a quick note on a player.

  3. After

    A voice note becomes the journal

    A 60-second voice note becomes one entry per student, which the coach checks before saving. Then fees are recorded, one tap per student: Venmo, Zelle, cash, a pack or not paid yet.

  4. Monthly

    Progress notes the coach approves

    CourtPlan drafts a short note per student from the journal. The coach edits and approves it, then sends it to the student or parent.

Design target: under 2 minutes of coach effort per session, one-handed on a phone. It is a target, not a result.

Code keeps the rules, the LLM drafts, the coach decides

Each step is labeled by who does it: code, the LLM or the coach. Nothing goes out without the coach's approval.

Voice note to journal

coachRecordThe coach records a 60-second voice note after class, with quick notes from the session already attached.
codeTranscribeCode sends the audio to a transcription service and gets text back.
LLMExtractClaude matches names to the roster and returns entries: worked on, observation, progress, next focus, confidence and any unmatched names.
coachReviewThe coach confirms or edits each entry; low-confidence entries and unmatched names are flagged first.

Planner

coachCheck inThe coach taps who showed up and sets the courts.
codeGroupCode splits players by level and focus into courts, handles odd numbers and builds time blocks, so the rules never break.
LLMPick drillsClaude picks drills only from the coach's library to match each group's next focus, and writes one cue per block.
coachAdjustThe coach reorders blocks or swaps a drill, then runs the plan on the timer.

Planned stack: a Next.js phone web app on Vercel, Supabase for data and audio, OpenAI transcription, Claude Sonnet for structured JSON and Resend for email.

Six screens from the prototype

Screens from a clickable prototype with example data. The students, groups and fees are made up.

Who's here tonight?: The coach taps who showed up and sets the courts. No fees appear before or during class.
Who's here tonight? The coach taps who showed up and sets the courts. No fees appear before or during class.
Tonight's plan: Groups by level on each court, with drills from the coach's own library and one cue per block.
Tonight's plan Groups by level on each court, with drills from the coach's own library and one cue per block.
Live session timer: A countdown for each block, the drill and cue on every court, and quick notes per player.
Live session timer A countdown for each block, the drill and cue on every court, and quick notes per player.
Check before saving: One drafted entry per student from the voice note. The coach confirms or edits each one.
Check before saving One drafted entry per student from the voice note. The coach confirms or edits each one.
Who's paid tonight?: After class, one tap per student: Venmo, Zelle, cash, a pack session or not paid yet.
Who's paid tonight? After class, one tap per student: Venmo, Zelle, cash, a pack session or not paid yet.
Progress note: A monthly note drafted from the journal, approved by the coach before it goes to a student or parent.
Progress note A monthly note drafted from the journal, approved by the coach before it goes to a student or parent.

How I'll find out if anyone wants it

Phase 0: I am the product for two weeks

Before any production code, I plan sessions and write journals by hand for the 2 pilot coaches. Friends tend to be polite, so interviews ask about their last session, not my idea, and the test is judged on behavior. It passes if they send attendance and voice notes without reminders and name a price.

  1. Next

    Phase 0

    Concierge test

    Coaches send attendance and voice notes on their own and name a price. If I have to chase them, it fails.

  2. Later

    Phase 1

    Journal

    Coaches log their own sessions, and most drafted entries need no edits.

  3. Later

    Phase 2

    Planner

    Coaches run the generated plan, with light edits, in most sessions.

  4. Later

    Phase 3

    Progress updates

    Each coach sends notes, and a student or parent replies; the coach says the note was worth sending.

  5. Later

    Phase 4

    Decide

    I ask the pilot coaches to pay. Then I decide to charge and grow, or stop. Any failed gate means stop and decide.

North star. Sessions per week where the coach both planned and logged.

RiskHow I handle it
Friends being politeConcierge test first, judged on what the coaches do, not what they say.
Coaches won't log after classA 60-second voice note, no forms.
Wrong names in voice notesRoster nicknames, flagged unmatched names and a review screen before saving.
Nonsense drillsThe LLM only picks from the coach's own drills, and code enforces the court rules.
A crowded categoryWin on the planner, which I found no sign of in Outcoach, built from the coach's own drills and attendance. Stay out of bookings.
A small marketThe same loop fits other small-group teaching beyond racquet sports, such as swim and music lessons.