APSY

Role
Product Design - behavioural analysis, UX redesign, UI
Duration
6 months
Tools
Figma, Microsoft Clarity
Team
PM, Engineering

Context

Overview

Apsy changes how individuals and businesses create digital experiences. You describe an app in conversation, agents build it, and you publish, no drag-and-drop, no database setup, source code included. The product worked. The problem was that almost nobody who signed up ever got to the end of it, and the analytics said exactly where they stopped.

My Role

I led the analysis and the redesign, reading the funnel data, running behavioural analysis in Clarity, designing and fielding the survey to inactive users, facilitating the brainstorming session, and designing the five changes that came out of it.

Project Brief

Analysing engagement from November 2024 onward turned up severe drop-offs at every stage of the app creation process. The numbers were not subtle.

1,476
users logged into the platform
44.9%
attempted to build an app

55.1% never started one

< 5%
of built apps were published in Gamma
0.6%
of users who attempted a build published to the store

These findings came from platform data and behavioural insight, and they pointed at bottlenecks costing conversion at every stage. The brief that followed was a comprehensive UX redesign: improve clarity, guide people through the build, and reduce friction wherever it appeared.

Research

Two sources, deliberately paired. Clarity told us where people stopped and what they did there. The survey told us why, because a heat map cannot say “I did not know how to preview my app”.

Behavioural data

Microsoft Clarity

High drop-off rates
49% of users left before reaching the “app name” stage.
Heat-maps
Repeated clicks on non-interactive elements, people were trying to press things that were not buttons.
Session replays
Users revisiting the same steps over and over, which reads as uncertainty about whether they had finished.

Two patterns underneath all of it: people spent excessive time on the onboarding screen, and drop-offs consistently landed at the points where they had to make a decision without enough context or an example to work from.

Survey of inactive users

100+ responsesSigned up, never finished

I sent a targeted survey to people who had signed up but never completed a build, and asked what stopped them.

49%
did not understand the next steps after naming their app
35%
felt onboarding lacked sufficient explanation of features
28%
wanted more visual or video-based guides
“The platform looks promising but feels overwhelming without clear instructions.”
“I didn’t know how to preview my app while building it.”

Insights and needs

Seven problems came out of the two research streams. Each one had a fairly obvious counterpart, which is what made the brainstorming session productive rather than speculative.

What we found

What it needed

Trust issues

Users hesitated to trust Apsy because the landing page looked unpolished.

Professional landing page

A redesign that inspires trust and credibility.

Guidance needed

Many struggled to navigate the platform and asked for tutorials or onboarding.

Tutorials and onboarding

A step-by-step tutorial to address confusion and improve engagement at the earliest stages.

Support access

Difficulty finding clear, accessible support frustrated users.

Clear support

Easy-to-find support tools or a chatbot.

Customisation challenges

Users could not easily match the options to their app type or vision.

Simplified customisation

Tools that match app types to what the user is actually trying to build.

Preview absence

With no way to preview, users were unsure whether they were making progress.

Preview feature

Real-time app previews to build confidence mid-build.

Unclear roadmap

People were confused about what happens at each stage.

Transparent roadmap

Visual milestones for the app-building journey.

Email overload

Too many emails diminished engagement rather than driving it.

Streamlined communication

Fewer, more targeted emails so people feel guided rather than pestered.

Brainstorming

With the insights in hand, the team ran a brainstorming session to work through the most pressing of them. The aim was to reimagine the experience around the pain points we could evidence, rather than around what we assumed people wanted, and to pick the smallest set of changes that would move the funnel.

Five workstreams came out of it.

What we built

Five workstreams

Each one started from a specific thing people told us, and ended with something we could point at in the product. I kept them separate deliberately, bundling them into one big redesign would have made it impossible to tell which change moved which number.

01

Landing page redesign

Research & insights

The landing page was doing two jobs badly: explaining what Apsy is, and getting people to sign up. Session recordings showed visitors scrolling past the value proposition entirely and bouncing before they hit any CTA. In the survey, the most common answer to "what stopped you signing up?" was some version of "I couldn't tell what it actually does."

Key changes implemented

Rewrote the hero to lead with the outcome rather than the technology. Cut the page from nine scroll-sections to five. Pulled the primary CTA above the fold and repeated it once at the end. Replaced the abstract illustration with a short product visual. Removed a secondary CTA that was competing with sign-up.

Outcome

25% increase in sign-ups · 15% reduction in bounce rate · 20% more clicks on the primary CTA

02

Onboarding tutorials

Research & insights

This was where we were losing the most people. Users who did sign up landed in an empty builder with no obvious first move. Support tickets clustered around the same three questions in the first session, and the drop-off curve fell off a cliff between account creation and first model created.

Key changes implemented

An interactive tutorial that walks through building the first model rather than describing it. Contextual tooltips on the controls people asked about most. A progress indicator so the end is visible from the start. Short video walkthroughs for the steps that are hard to explain in a tooltip. Quick-start templates so nobody begins from a blank canvas.

Outcome

Onboarding drop-off reduced by 35% · Onboarding-related support tickets down 40%

03

Support accessibility

Research & insights

Help existed, but it was buried. People who got stuck mid-task had to leave what they were doing, find the help section, and describe a problem they were no longer looking at. Several survey responses described giving up rather than going looking.

Key changes implemented

A floating support button that sits on every screen and opens a help-request form in place, pre-filled with where the user is in the product. Getting help no longer costs you your place in the flow.

Outcome

Help requests now arrive with context attached, which shortened the back-and-forth on each ticket

04

Customisation simplification

Research & insights

Customisation was all-or-nothing. You either accepted the defaults or configured everything by hand, and the second option asked for decisions most people didn't have the vocabulary to make yet. Novice users in particular stalled here.

Key changes implemented

Pre-designed model types that cover the common cases as one-click starting points, with a free-text input for anyone who wants to describe something the presets don't cover. Structure for people who want it, an escape hatch for people who don't.

Outcome

Fewer abandoned sessions at the configuration step; the presets became the most common entry point

05

Email communication

Research & insights

Emails were frequent, generic and long. Open rates were falling and a handful of survey responses mentioned unsubscribing, not because they'd lost interest in Apsy, but because the volume had become noise.

Key changes implemented

Cut the send frequency. Restructured the templates around a single message with one action. Personalised subject lines against what the user had actually done in the product rather than a broadcast schedule.

Outcome

Improved open rates and a lower unsubscribe rate across the sequence

Apsy builder screens in dark mode: the app editor, the Create Account form, the dashboard of app previews, and the Gamma assistant offering a quick tour

The result

What actually moved

The two workstreams with clean before-and-after measurement were the landing page and onboarding. They are also the two that sat directly on the biggest leaks in the funnel, which is not a coincidence, they were sequenced first for exactly that reason.

+25%
increase in sign-ups
−15%
bounce rate on the landing page
+20%
clicks on the primary CTA
−35%
drop-off during onboarding
−40%
onboarding support tickets
+5%
attempted publish to store

The other three workstreams, support, customisation and email, shipped without instrumentation tight enough to claim a number, so I am not claiming one. What I can say is that help requests started arriving with context attached, the preset model types became the most common entry point into a build, and the email sequence stopped generating unsubscribes at the rate it had been.

Ongoing

What comes next

None of this is finished. The five changes fixed the leaks we could evidence; the plan from here is to keep the evidence coming rather than wait for the next drop in the numbers.

  1. 01

    Continuous feedback

    Keep collecting user feedback on a regular cadence so improvements stay driven by what people report, not by what we assume.

  2. 02

    AI chatbot support

    Introduce a chatbot for instant help, so the floating support button has an immediate answer behind it rather than a queue.

  3. 03

    A wider model library

    Expand the range of pre-designed model types so the presets cover more of what people actually turn up wanting to build.

  4. 04

    Behaviour-driven email

    Personalise the email sequence against in-product behaviour rather than a schedule, the right nudge at the point someone stalled.

  5. 05

    Regular usability testing

    Test on a standing schedule to catch friction as it appears, instead of discovering it in a funnel report six months later.

Reflection

Takeaways

A working product can still fail at the door

Apsy did the hard thing well. Agents really did build the app. Almost none of that mattered to a user who left forty seconds in because they could not tell what to do next. The gap between what a product can do and what a first-time user can get it to do is where most of the value was sitting.

Analytics tell you where, surveys tell you why

Clarity showed a wall at the app-naming step. It could not tell me that people did not know a preview existed. Pairing the two was the difference between redesigning the screen where people left and redesigning the thing that made them leave.

Splitting the work made the results legible

Five separate workstreams took longer to sequence than one redesign would have. It also meant that when sign-ups moved 25%, we knew which change did it. A single big launch would have given us one number and no attribution.

Structure beats blank canvases

Every change that worked did the same thing: replaced an open-ended decision with a starting point. Templates instead of an empty builder, presets instead of full configuration, a progress indicator instead of an unmarked path. People are not short of ideas, they are short of a first move.