Farhan - Portfolio

Context

KYC needed to be resumable without making people start over

The earlier KYC flow followed one fixed sequence. Users completed KTP first, then Selfie, before moving into identity details and submission.

Autosave made that structure worth revisiting. Saving progress had limited value if users still had to follow the same order or repeat work after leaving the flow.

I designed two flow directions, set up the usability study, analyzed the behavioral and survey evidence, then translated the decision into product rules for implementation.

Autosave became more useful once saved progress changed how the flow itself behaved.

My role

My scope ran from the experiment into implementation behavior

I worked across the flow concepts, study setup, analysis, and the interaction rules needed once the selected direction moved toward production.

Flow designdesigned Flow A and Flow Bfixed sequence compared with a task-based structure

Researchset up the usability studyMaze tasks followed by a post-test survey

Analysiscombined behavioral and survey evidenceincluding SUS, misclicks, confidence, and open feedback

Deliverydefined Autosave behaviorsaved progress, retake, resume, and submission states

User journey

The journey changes once progress no longer depends on one session

This map follows the user from starting the upgrade through submission, including the moment where KYC is left and resumed later.

Journey layer

01Start upgradeUnderstand what is needed

02ReviewSee tasks and progress

03First taskComplete KTP or Selfie

04Come backResume saved progress

05Remaining taskFinish what is still incomplete

06SubmissionSend KYC for verification

User goalWhat the user is trying to achieve

Know what to prepare before starting

Understand what is complete and what still needs work

Finish one task with confidence

Continue without losing earlier work

Focus only on the unfinished task

Submit once everything is ready

User actionWhat happens in the interface

Read the upgrade benefits and requirements

Choose KTP or Selfie from Review

Capture the photo and confirm the result

Open KYC again and see the earlier task still complete

Complete the task that is still active

Check the final data and continue to verification

Experience riskWhere unnecessary repetition or uncertainty can appear

Requirements only become clear after work has started

A forced order prevents users from adapting to their current situation

It is unclear when progress is actually saved

The user expects to start again

Completed work gets repeated

The visible progress does not match what is submitted

Design responseHow the product answers that need

Show both tasks before KYC begins

Use Review as the progress hub

Save each task after the result is confirmed

Restore the state of each task when the user comes back

Keep completed work intact while the remaining task stays actionable

Bring the tasks together before the next verification step

The journey focuses on KTP and Selfie because those two tasks were the core of the Autosave experiment.

Flow exploration

We tested a fixed sequence against a flexible task hub

Both concepts saved progress. The difference was how much the surrounding KYC structure changed once that progress could persist.

Flow A

Keep the existing sequence

Users followed the predetermined order and the guide remained a required step before moving to the next task.

Autosave sits inside the old interaction pattern.

Flow B

Turn Review into a task hub

Users see the status of KTP and Selfie in one place, choose where to start, and return to the hub after completing one task.

Autosave changes the interaction structure as well as persistence.Try Flow B

Research question

The extra flexibility needed testing because it also introduced more choice

Flow B gave users more control and made progress easier to inspect. It also added a decision that did not exist in the earlier sequence.

I wanted to check whether that tradeoff made KYC harder to understand. Both directions were therefore tested as end-to-end experiences rather than compared only as wireframes.

Method

Randomized usability study with 30 participants

The study ran unmoderated in Maze and ended with a post-test survey.

30participantstotal study sample

15 and 15group splitFlow A and Flow B

Unmoderatedtest formattask-based usability study

Maze and surveydata sourcesbehavioral and self-reported

Research evidence

Inspect the study setup and results

The evidence is grouped here so the study can be inspected without turning the main case into a sequence of static charts.

One randomized study across two flows

Participants completed one of the Maze flows and then answered the post-test survey.

Randomized usability study setup for Flow A and Flow B — N 30 with 15 participants assigned to each flow.

Reading the results

The study did not produce a statistical winner

SUS and total misclicks showed no significant difference between Flow A and Flow B. Several perception measures followed the same pattern.

I then looked at the descriptive data and open feedback without treating them as inferential proof. Flow B showed a lower misclick trend, steadier confidence, and feedback that was more supportive of the progress hub.

Flow B also matched the product model better because each task was already meant to persist independently. There was no meaningful usability penalty that justified keeping the older sequence.

Flow B gave us enough confidence to move forward while staying within what the study actually supported.

Claim boundary

The product decision stayed separate from the statistical claim

The study informed the direction, but it did not justify saying that Flow B was statistically proven to be better.

Not supported

Flow B won the A B test

There was no significant difference in SUS or total misclicks, so the study did not establish a statistical winner.

The inferential result remained non-significant.

Supported for the decision

Flow B was a reasonable production baseline

Usability did not meaningfully deteriorate, while the behavioral trend, feedback, and Autosave model all supported the more flexible structure.

Decision confidence came from the full body of evidence.

Interactive prototype

Try Flow B starting with KTP or Selfie

This walkthrough shows the behavior that became the basis for Autosave. Choose the first task, return to Review with saved progress, then complete what remains.

TRY THE FLOW01 / 07

Set expectationsUpgrade entryThe upgrade page explains the two verification tasks before the user enters KYC.

System flow

Autosave keeps progress intact when someone leaves KYC

This flow focuses on behavior the prototype cannot show. Progress is stored per task and restored when KYC is opened again.

UserUser actions

Frontstage UIVisible product states

Autosave stateProgress that persists

VerificationHandoff after submission

entry

Enter KYC

Start fresh or continue existing progress.

Review reads progress

KTP and Selfie keep independent status.

action

Complete one task

The user starts with either KTP or Selfie.

saved state

Persist task state

A completed task is saved on its own.

later session

Leave and return

The user can continue in a later session.

state

Restore progress

Completed work still appears complete.

Show the remaining task

Review only asks for unfinished work.

action

Finish the last task

Earlier progress does not need repeating.

saved state

Persist complete state

KTP and Selfie are now both complete.

Enable Continue

The CTA becomes available when requirements are complete.

handoff

Send to verification

Submission hands the data to verification.

Flow connections

enter to review
review to first-task
first-task to save-first
save-first to remaining
save-first to leave-return, continue later
leave-return to restore
restore to remaining
remaining to finish
finish to save-complete
save-complete to continue
continue to verification

The interactive prototype above shows the KTP and Selfie ordering. This flow only explains the persistence and resume behavior underneath it.

Design specification

Four rules carried the interaction into implementation

These rules keep Flow B predictable when someone retakes a photo, leaves KYC, or returns in a later session.

Per taskprogress persists independentlyKTP and Selfie keep their own state

Reviewacts as the source of truthcomplete and incomplete states are visible in one hub

Retakeonly changes the relevant taskother completed work does not need to repeat

Submitclears stale progresssaved state should not return after the process is complete

Outcome

The study result and the production outcome are shown separately

The study directly supported the Flow B decision. Conversion moved later while several KYC improvements were live at the same time.

Flow Bproduction baselineselected after the usability study

n.s.SUS and total misclicksno significant difference

11%peak KYC upliftobserved during the broader KYC improvement period

5%more stable upliftversus the earlier period

Not attributed to Autosave alone.

Homepage copy and other KYC changes were also live.

Broader KYC outcome

Conversion moved higher during the wider KYC improvement period

This chart provides post-implementation context. The same period also included other KYC changes.

DERIVED · Internal KYC conversion tracking · KYC improvement period after implementation

TAKEAWAY

Flow B formed part of the experience used during this period, while the conversion uplift cannot be isolated as an Autosave-only effect.

View data

Label	Uplift versus earlier period
Earlier period	0%
Improvement-period baseline	5%
Improvement-period peak	11%

The 5% and 11% values describe the baseline and peak uplift during the broader KYC improvement period. Homepage copy and other changes were live at the same time, so this chart is not used as a causal estimate for Autosave.

Reflection

A non-significant study can still support a strong product decision

The study did not produce a statistical winner. That made the decision boundary clearer. Flow B could move forward because usability did not meaningfully deteriorate and its behavior matched what Autosave needed to support.

A next iteration should measure resume rate, retakes by task, and completion after users return. Those measures would make the value of persistence easier to separate from other KYC changes happening around it.

Making KYC Resumable Without Starting Over