How receipt scanning works in bill splitting apps (and why it matters)

Typing in every line item from a restaurant bill is not a feature. It's a punishment.

You're at the end of a good dinner. You want to settle the bill and leave. Instead, you're squinting at a crumpled thermal receipt, manually entering "Butter Chicken x2 - ₹560", "Garlic Naan x4 - ₹220", "Raita - ₹80", then handling the GST and service charge separately.

That experience is why most people give up on itemised splits and just say "let's split equally" even when they shouldn't.

Receipt scanning exists to fix exactly that problem.

What receipt scanning actually does

You take a photo of the bill. The app reads it.

Behind that sentence is a combination of Optical Character Recognition (OCR) and some basic parsing logic. OCR converts the image of text into actual text data. The parsing layer then identifies what's an item name, what's a price, what's a quantity, and what's a subtotal or tax line.

The output: a structured list of items and amounts, ready to be assigned to people at the table.

A close-up of a paper receipt on a table

Where it gets complicated

Not all receipts are created equal, and this matters for the technology.

Clean printed receipts from organised restaurants are easy. The text is aligned, fonts are consistent, item names and prices are in predictable columns. OCR handles these well.

Thermal receipts from smaller Indian restaurants are harder. The print quality is lower, items are sometimes abbreviated to cryptic shorthand ("BTRCHKN x2"), and columns aren't always consistent. The OCR still works, but you might need to correct an item or two.

Handwritten receipts, or photos taken in bad lighting, are the hardest. Most apps will still get the total right, but line-item accuracy drops.

The practical advice: good lighting, flat receipt, phone camera directly above. That combination gives the best result with any scanning app.

How the itemised split works after scanning

Once the receipt is read, you see a list of items. The next step is assigning each item to the people who ordered it.

Some apps ask you to do this manually: tap the item, select the person or people who ordered it. Barabar lets you do this with a few taps. The shared items (a bottle of water for the table, the bread basket) get split equally among everyone. Individual items go to the person who ordered them.

The app then calculates each person's share including their portion of any service charge and GST.

The result is an accurate, itemised split that would have taken 10 minutes to calculate by hand, done in about 60 seconds.

Tired of tracking this manually?

Barabar extracts bills, splits expenses, and settles via UPI. Instant.

Launch barabar

Why this matters beyond convenience

There's a deeper reason receipt scanning is worth building properly.

Itemised splits are more fair than equal splits. Most groups default to equal split not because it's fair, but because the alternative is too much work. When scanning removes the work, people actually split accurately.

This changes how groups behave. The non-drinkers aren't quietly subsidising the bar tab anymore. The person who had the thali doesn't pay the same as the person who ordered two starters and a dessert.

Small injustices in equal splits don't feel large in isolation. Across months of dinners and trips, they accumulate. Receipt scanning removes the root cause.

What Barabar's receipt scanning handles

The scanner reads the full bill, identifies line items and prices, handles multi-item rows (Naan x3 = ₹150), and separates tax lines from food items. You review the parsed list, make any corrections, assign items to people, and get a final split.

It handles Swiggy and Zomato order summaries too. Paste the order breakdown text, or screenshot the order confirmation. Same process.

The future of this

Receipt scanning today is largely OCR plus parsing rules. The next generation of this is smarter: recognising abbreviated dish names, understanding restaurant menu formats, automatically suggesting splits based on past behaviour in the same group.

That's coming. But even the current version saves enough time per use that it changes how people track shared meals. For a group that eats out three times a week, that's hundreds of minutes saved per year.

More importantly, it means accurate splits become the default instead of the effort. And accurate splits mean fewer quiet resentments and fewer awkward money conversations later on.

The 60-second scan is doing more than just reading a receipt.