Seungju WI

← Experience

Planfit

AI Problem Solver · Internship

Jun–Dec 2025

70+ experiments

Up to +75% CVR

Owned subscription conversion as a full-cycle AI Solver

Planfit

Context

An AI fitness app. Joined by volunteering for “Solver”, a new role created to clear the bottlenecks between planners, designers, and developers with AI.

Ran one-person sprints across planning, design, front end, and QA to improve free-to-paid subscription conversion. Designed and ran 70+ experiments in about 3 months and wrote 100+ PRDs.

Position
AI Problem Solver · Internship
Period
Jun–Dec 2025 (6 months)
Mission
Convert free users to paid subscriptions
Stack
Amplitude · Figma · Cursor · Claude · Veo

Key Projects

Project 01 — AI Video Paywall · New User

A plain discount,
reframed as a special season.

+20%

CVR uplift

Double the +10% target

A Christmas-only paywall for users within 7 days of sign-up. AI-generated seasonal video (Veo, Midjourney) created a reason to pay now.

01Thinking

Without touching the discount rate,
how do you lift conversion?

  1. 1. Observation

    The coupon had become “a given”

    Free users saw the discount coupon all the time. They already knew it would show up whenever, and purchase conversion was slowly declining.

  2. 2. Diagnosis

    The problem wasn’t price but the lack of “specialness”

    An always-on discount feels less like a benefit and more like noise, so making the coupon bigger wasn’t the answer. What was missing was a “now or never” context.

  3. 3. Hypothesis

    Restore “specialness” with seasonal context

    Wrapping the same discount in a seasonal frame should make users see it as a special opportunity again. Designed a Christmas-only paywall as the first test.

Same benefit, different context: the value users feel changes.
An experiment on perception, not price.

02Execution

AS-IS / TO-BE

The previous static gift-box paywall
AS-ISStatic gift-box image
TO-BEAI seasonal video

How it was built

  1. 01

    FIGMA · CONCEPT

    Define the limited-season concept

    Chose a Christmas-only paywall as the first testbed for “now or never”. Designed the copy, visual tone, and landing structure in Figma.

  2. 02

    VEO · MIDJOURNEY

    Produce the seasonal animation with AI

    Generated Christmas animations with Veo and Midjourney. Multiple versions in days with no outsourcing, so the experiment moved fast.

  3. 03

    REACT NATIVE · A/B TEST

    Build the front end & A/B test

    Implemented the front end myself and A/B tested it against the static image paywall on the new-user segment.

03Impact

New-user coupon payment conversion rose 20%, double the 10% target. The “change the context, not the price” hypothesis held, and seasonal paywalls became the internal reference for later campaigns.

Project 02 — AI Prediction · Free User

Leave the screen alone,
find where it can still work.

+75%

Still in production

Running to this day

Existing users’ screens couldn’t take many experiments. Looked for a lever that affects conversion without touching the UI, and found, adopted, and ran Monetai, an external AI solution that predicts purchase probability and controls coupon exposure.

01Thinking

Without touching the screen at all,
how can you move conversion?

  1. 1. Observation

    Existing users’ screens were off-limits for experiments

    Defined existing users as those still free 14 days after sign-up. They were used to their screens and flows, so running many paywall and UI experiments on them wasn’t an option.

  2. 2. Diagnosis

    The answer was in another layer, not the screen

    If the screen can’t change, find a lever that moves conversion without changing it. That lever was the layer controlling when a coupon is shown.

  3. 3. Hypothesis

    Get the exposure-control layer from an external solution

    Purchase-probability-based exposure control was hard to build in-house, so the research extended to external solutions. Found Monetai, which shows coupons only to users with low purchase probability, and led adoption, operations, and ongoing tuning alone.

Find where it can work without touching the existing screen.
If that lever doesn’t exist in-house, get it from outside.

02Execution

How it was built

  1. 01

    USER DEFINITION

    Define “existing users”

    Users still on the free plan 14 days after sign-up. Started from the constraint that their screens couldn’t be changed casually.

  2. 02

    RESEARCH · SOURCING

    Look outside: finding Monetai

    Researched external solutions for a problem that was stuck in-house. Found Monetai, which predicts purchase probability and shows coupons only to low-probability users.

  3. 03

    PARTNERSHIP · 1:3 MEETING

    Lead adoption & work directly with Monetai

    Ran recurring 1:3 meetings with Monetai’s CTO, developer, and designer; agreed on segment criteria, exposure logic, and how to read results; and led the rollout from pilot to production alone.

  4. 04

    LIVE · CONTINUOUS IMPROVEMENT

    Operate & keep improving with data

    Kept monitoring Amplitude after launch and tuned segments and exposure conditions with Monetai, raising weekly payment conversion further. Still running in Planfit production.

03Impact

Raised weekly payment conversion for existing users by 75% without touching their screens. The exposure-control layer secured through the partnership wasn’t a one-off experiment: it is still running in Planfit production.