A merchant that sells everything cannot have one honest category. For Amazon, Paytm or a large supermarket, you have three workable options: categorise each order by what was bought, use a single Shopping - mixed bucket for the whole merchant, or split one payment across several categories. Pick based on how much the detail matters to you, and let AI help with the easy rows only.
Why is a mixed merchant such a problem?
Your bank statement shows who you paid, not what you bought. A payment to a large online store could be a bag of rice, a phone charger or a birthday gift. A payment through a wallet or payments app could be a mobile recharge, an electricity bill or a payment to a friend. Software sees only the merchant name and the amount, so it is guessing.
What are my options?
| Option | Effort | What you get | Weakness |
|---|---|---|---|
| Categorise each order | Highest | Accurate categories | Needs order history or receipts |
| One mixed bucket | Lowest | Fast and consistent | Hides what the money was for |
| Split a payment | Medium | Exact for big mixed orders | Needs the order breakdown |
Categorise by order
Open your order history, match each payment by date and amount, and file it under what it was. It takes time, but for a merchant that takes a large share of your spending, it is worth it. Say you spent ₹9,600 across one store in a month: ₹4,100 on groceries, ₹3,200 on electronics and ₹2,300 on gifts. Those add up to the full ₹9,600, and your Groceries and Gifts categories become honest.
A mixed bucket
If the total is small or you do not care about detail, make a category called Shopping - mixed and send the merchant there. You lose nothing in your overall total, and you can review the bucket once a quarter to see whether it has grown big enough to deserve a closer look.
Splitting
Where a single large order includes several kinds of things, record it as separate entries that add up to the payment. Check that your tracker supports splits, or enter the parts by hand, adjusting so the total still matches the bank amount.
Where does AI help and where does it fail?
AI does fine when a merchant is clearly one thing, such as a supermarket chain you shop at only for groceries, or a recharge with a typical amount. It struggles when the same name covers many uses. An amount band helps a little: a ₹199 payment and a ₹19,999 payment to the same store are probably different things. But it is still a guess, so mixed merchants are the ones to review yourself.
SpendixAI checks your own rules first, then your history and similar merchants, and only then asks an AI model, using the merchant name and an amount band. For a mixed merchant, a good plan is to leave the category as a rule you control or send it to review, and to add merchant aliases if the same shop appears under more than one name.
What is a sensible default?
- Supermarkets that mainly sell food: Groceries, unless a particular visit was clearly something else.
- Large online stores: Shopping - mixed, and categorise individually only the big or recurring orders.
- Wallet and payments apps: look at the amount and the narration; recharges and bill payments get their own category, and payments to people go to a transfers category.
- Anything above an amount you set, such as ₹5,000: always check by hand.
A small monthly routine
Once a month, open the mixed bucket and sort by amount. Re-file anything large that has an obvious home, and leave the small remainder where it is. Ten minutes is usually enough, and the habit keeps one noisy merchant from distorting a whole category.
Useful ideas start with a clearer picture of your own spending.
Explore your reports

