Skip to main content

Flutter Mobile App Case Study · Mobile product engineering

SettleTab: Making Shared Bills Easier to Understand.

From a receipt to an editable, shareable breakdown of what everyone owes.

Splitting a group bill sounds simple until people order different items, share some purchases, and need to account for tax and tip. SettleTab brings those decisions into a mobile workflow. Users can capture a receipt or enter a bill manually, review its items, assign purchases, calculate individual totals, and share the finished breakdown.

Role
Shashank Sangule · Founder & Lead Developer, DivWeavers
Platform
Android
Status
Live on Google Play
View SettleTab on Google Play
Product
SettleTab
Focus
Itemized bill splitting, receipt review, group collaboration and sharing
Stack
Flutter · Dart · Riverpod · Supabase · PostgreSQL · Edge Functions · AI/Vision APIs · Google ML Kit · RevenueCat · OneSignal

Case study at a glance

Product problem
Shared bills need an itemized representation people can check, including individual purchases, shared items, tax, and tip.
Role and scope
Shashank Sangule, Founder & Lead Developer at DivWeavers: Flutter app, receipt processing, split calculation, Supabase backend, sharing, testing, and Android release infrastructure.
Implemented solution
Scan or enter a bill, correct the proposed items, assign participants, calculate deterministic totals, and share a read-only result. The screens and narrative demonstrate the mobile-to-backend workflow.
Evidence limits
SettleTab can open external payment destinations; SettleTab does not transfer money. A public Google Play listing exists for the Android application; installation eligibility, runtime service health, adoption, and revenue are not independently verified here.

Product engineering / 01

Your order.
Your share.

Scan or enter a bill.
Review, assign, and share.

Live on Google Play

01 / Context

Making Shared Bills Easier to Understand

Splitting a group bill sounds simple until people order different items, share some purchases, and need to account for tax and tip. SettleTab brings those decisions into a mobile workflow.

Users can capture a receipt or enter a bill manually, review its items, assign purchases, calculate individual totals, and share the finished breakdown. Creating and managing splits requires an account, with email/password or Google Sign-In.

The problem

The Challenge: One Bill, Different People, Different Shares.

A total amount doesn't explain how much each person should pay.

Some items belong to one person. Others are shared. Receipt recognition isn't always perfect, and the calculations need to reflect the final reviewed bill.

The product needed a clear way to move from uncertain receipt input to a breakdown people could inspect.

Product screenshots

Correct the items before splitting.
Editable SettleTab receipt for The Green Table with five items, tax, tip, and a ₹1,886 total.
Assign purchases, including shared items.
SettleTab assignment screen showing individual items, a shared pizza, and unassigned Garlic Bread for Alex, Maya, and Jordan.

Product intent

Reduce entry work without hiding the source data.

The product goal is a clear path from receipt input or manual entry to a reviewed bill, participant totals, and a breakdown people can share.

  • Reduce repetitive receipt entry
  • Keep extracted items editable
  • Use reviewed items as the input to split logic
  • Keep final calculations deterministic
  • Make results accessible to people without the app

02 / Engineering

My Engineering Contribution

I worked across the Flutter application, receipt-processing flow, split calculations, Supabase backend, group collaboration, sharing, subscriptions, notifications, testing, and Android release infrastructure.

The project required mobile and backend work to behave as one connected product.

Designing for uncertainty

The system combines uncertain input with exact expectations.

Receipt recognition can be incomplete or wrong, while the people settling a bill expect totals they can inspect and reproduce. That makes correction a primary path rather than an exceptional state.

  • Physical receipt input is not consistently structured
  • Extracted values require review
  • Financial rules must remain deterministic
  • The workflow must stay understandable when automation is imperfect

User and system workflow

The Solution: Capture, Review, Assign & Share.

SettleTab follows a straightforward flow.

  1. 01. Capture or Enter the Bill

    Start with a receipt image or manually enter the bill details.

    Account required
  2. 02. Review the Items

    AI and OCR assistance can propose receipt information, but users can correct items, prices, tax, and tip before continuing.

    Camera · gallery · manual
  3. 03. Assign Purchases

    Assign individual items to participants and divide shared items.

    User correction
  4. 04. Calculate Each Share

    Application code calculates item shares, tax, tip, and rounding from the reviewed bill data.

    Individual · shared
  5. 05. Share the Result

    Send a breakdown through the device's sharing options or share a finalized, read-only browser result when the split is saved online. People receiving the browser link can view the result on a read-only browser page without installing the app or creating an account.

    Application logic
Start with the bill you have.
SettleTab home screen offering receipt capture, gallery upload, and manual entry.
See what each person owes.
SettleTab summary for a ₹1,886 bill showing Alex ₹667, Maya ₹690, Jordan ₹529, and Share Split.

Architecture

System boundary

Described functional boundary

The checked-in architecture connects Flutter to Supabase and a read-only browser result; deployed extraction providers remain configuration-dependent.

SettleTab architecture diagram illustrating mobile receipt capture, human review screen, deterministic calculation engine in Dart, and Supabase cloud sync.
  1. Flutter input and review

    Signed-in users capture or enter a bill and correct its item list.

  2. Supabase data and extraction

    Auth, PostgreSQL, private receipt storage, realtime collaboration, and server functions support the workflow.

  3. Dart calculation

    Reviewed items and assignments determine tax, tip, and reconciled totals.

  4. Shareable result

    Saved finalized splits can be read on a public, read-only browser page.

Engineering challenges

Where product judgment mattered

Unstructured receipt input

Challenge

A receipt image does not arrive as reliable item, price, and total fields.

Why it mattered

The rest of the split depends on a representation people can inspect.

Approach

Use AI/OCR assistance to propose structured receipt data, then put that structure into an editable review step.

Trade-off

Review adds a step, but it prevents faster extraction from silently becoming incorrect product state.

Result

The described workflow produces editable itemization instead of treating recognition output as final.

Automation without surrendering control

Challenge

Manual entry is slow, but probabilistic extraction can be wrong.

Why it mattered

A bill-splitting workflow needs both lower input effort and user trust.

Approach

Automate the first pass and make correction a normal part of the workflow.

Trade-off

The interface must support inspection rather than presenting automation as instant certainty.

Result

Users remain in control of the data used for the split.

Deterministic financial logic

Challenge

Extraction can be probabilistic; a final bill cannot be opaque.

Why it mattered

Participants need totals that can be traced back to reviewed items and rules.

Approach

Keep bill-splitting calculations in deterministic product logic after review.

Trade-off

This preserves a boundary between two processing stages instead of asking one AI response to do everything.

Result

The calculation remains reproducible and explainable.

Important decisions

The trade-offs behind the workflow

AI Assists with Data Entry. It Doesn't Decide the Final Bill.

Decision

Receipt images are inconsistent, and extraction can produce errors. Instead of treating extracted information as final, SettleTab makes it editable. This keeps users involved before the calculations begin.

Why

The two jobs have different reliability requirements.

Alternative considered

Ask one model response to extract the receipt and finalize the financial result.

Trade-off

The product needs an explicit handoff between extracted and reviewed data.

Why acceptable

That handoff keeps the consequence visible and the final calculation inspectable.

Financial Rules Stay in Application Code.

Decision

The amounts owed are calculated using defined product logic rather than asking a language model to produce final totals. Individual and shared items, proportional tax and tip, and rounding remain explainable.

Why

Financial calculations require absolute reproducibility and explainability.

Alternative considered

Rely on probabilistic model outputs to divide itemized costs.

Trade-off

Requires maintaining explicit Dart business logic and edge-case handling for rounding reconciliation.

Why acceptable

Every participant can inspect and verify exactly how their total was calculated.

Sharing Extends Beyond the Mobile App.

Decision

Not everyone in a group needs to install another application to read the result. A finalized split saved online can be viewed through a read-only browser page. SettleTab can open external payment destinations, but it does not transfer funds itself.

Why

Friction kills bill settlement if every group member is forced to download an app and create an account.

Alternative considered

Require all participants to install SettleTab and create accounts to view their breakdown.

Trade-off

Requires a public read-only web rendering surface and secure Supabase sharing endpoint.

Why acceptable

Recipients can view their bill breakdown instantly in any browser without signing in.

Inside the product

Assist the input step; keep the product rules explicit.

Flutter handles capture, manual entry, review, assignment, and summary. Supabase provides authentication, PostgreSQL data, private receipt storage, realtime collaboration, and server functions. Receipt processing proposes bill data; Dart application logic divides shared items equally, allocates tax and tip proportionally, and reconciles rounding.

Failure and correction

Building for Imperfect Inputs

The receipt-processing flow includes alternative paths when automated extraction is unavailable or unsuccessful.

Users can correct proposed items or enter the bill manually.

This was an important design choice: automation should make the task easier without making the entire workflow dependent on perfect recognition.

03 / Beyond the app

A result people
can pass along.

A finalized split saved online can be opened in a browser without installing SettleTab or creating an account. The shared view is read-only.

SettleTab can open external payment destinations; it does not transfer money.

Read-only SettleTab web page for The Green Table showing a ₹1,886 bill with Alex selected and owing ₹667.
Send a result people can read without the app.

04 / What shipped

What Shipped

SettleTab is live on Google Play for Android.

Its implemented workflows include receipt capture and manual bill entry, editable itemization, shared and individual purchases, calculated totals, group collaboration, saved balances, and shareable breakdowns.

The engineering spans Flutter, Supabase, PostgreSQL, authentication, receipt storage, server functions, subscriptions, and push notifications.

Keep track of what the group owes.
Weekend Dinner group balances showing Alex owed ₹1,219, Maya owing ₹690, and Jordan owing ₹529.

Implemented behavior

A concrete mobile workflow replaced two poor extremes.

The live Android workflow turns receipt images or manual bill items into editable, itemized splits and shareable totals. Join codes support collaborative item assignment, while saved groups carry balances across bills.

Product engineering

What This Project Demonstrates

SettleTab shows how I approach products that combine a user-facing experience with backend rules, third-party services, and AI-assisted input.

The main lesson is that automation and reliability don't need to be competing goals.

AI can help users begin the task, while reviewable information and deterministic calculations keep the result understandable.

What carries forward

The correction path is part of the product.

Probabilistic extraction is most useful when the user can inspect what it produced. Financial behavior remains easier to trust when the final rules are ordinary, reproducible product logic.

SettleTab assignment and summary screens showing how a shared bill becomes individual totals.

Next / Engineering case study

Tare

Tare is a mobile habit reflection app that saves original check-ins, proposes structured fields for user review, and uses recorded history for progress views and insight paths.

Explore Tare

Building a product with similar technical complexity?

Share the product problem, the current system, and where delivery is stuck.