How to Automate Supplier Price List Updates in India

Why this problem costs money

Every time a supplier sends a new list, someone has to open it, find each product in your system, compare the old price with the new one, and change it. When the file is a PDF table or an Excel sheet with its own column names, that person also has to work out what each row means. The cost is not only hours. A missed increase on a fast-moving item goes into your margin on every order until someone notices.

Put numbers on it. Assume a supplier list has 400 lines and a buyer spends one minute per line to find, check and key the change. That is close to seven hours for one list. Ten suppliers sending two lists a month would mean more than 130 hours, which is most of one full-time month. These are our working assumptions, not a published benchmark. We could not find a public study we would trust for a per-line time, so the one input to replace is your own minutes per line. Time one real list before you plan anything.

Missed increases are the second cost. Purchase price variance is the gap between the price you planned to pay and the price you were charged. Take a simple case: you planned for 100 per unit, and the invoice comes at 108 for 1,000 units. That gap is 8,000, and it only shows up when the invoice arrives unless someone compared the list at the start.

In India, GST adds a second trap. Some suppliers quote prices before GST and others after it. If a supplier switches basis between revisions, the list can look like a jump when nothing has changed in the goods. Record the basis for each supplier and compare like with like.

The manual way and the automated way

The table shows where the hours go. The reading and typing steps are the easiest to remove. The approval step is the one most guides skip, and it is where the buyer's judgement still matters. Automation should shrink that step to a short list, not remove it.

Step by step: how to automate it

1. Build the cross-reference table before anything else. It has one row per product, with your item code, the supplier's code, the supplier name, the unit (piece, box, kilogram), the pack size, the last approved price and the GST basis. Keep it in Google Sheets or Excel. Matching depends on it, and most failed projects skipped it.
2. Pick one intake channel for one supplier. Email is the simplest. Create a dedicated address for price lists and set rules that save each attachment to a folder per supplier. Automation tools such as n8n or Make can watch that inbox and start a run. Many Indian suppliers send lists on WhatsApp instead. The WhatsApp Business Platform, reached through an approved provider, can receive those files, but the flow is harder to build, so add it after email works.
3. Read the file according to its type. For Excel and CSV, a short Python script using pandas or openpyxl can find the header row by name and ignore the title rows. For text-based PDFs, a library such as pdfplumber can pull out the table. For scanned PDFs, use OCR, for example Google Document AI or Tesseract, and expect more rows to need a person. A large language model can help when a supplier's layout is messy. Ask it to return rows in a fixed format, check that every price is a number, and never let it write to your system directly.
4. Match by code first. An exact match on the supplier code is safe to accept. If there is no match, try the supplier's description against your product names, but send that result to review rather than accepting it. Fuzzy matching should never approve a price on its own.
5. Calculate the change for every matched line. Compare the new price with the last approved price and work out the percentage change. Flag any increase above your limit. Many companies use a tolerance around a fixed percentage of the standard price. One common setting is plus or minus 3 percent, as noted in a 2025 guide from Ramp, but set your own figure by product group. A tighter limit suits bulk raw materials, and a looser one may suit stationery. Also flag new items, removed items, and any change in unit or pack size.
6. Send only the exceptions to the buyer. A short email with a link to a shared sheet works well at the start. The buyer ticks approve or reject, and the sheet records the decision. Once that habit is set, move the same exception list to WhatsApp. Replies inside the 24-hour customer service window are free, but a first message outside it has to use an approved template, so plan for that in the design.
7. Import only approved rows. In TallyPrime, the price list import works from an Excel file, but the price levels must already exist in the company data, and the file needs the Price Level and Price List Date fields. A Google Sheet has to be downloaded as an Excel file first, and a CSV has to be saved as .xlsx. Tally also uses mapping templates that record the header row number, so the import does not depend on where a column sits. Check which field your buying cost is read from, because the Tally price list feature covers cost and last price for each stock item. Zoho Books and other accounting tools have their own import routes, and a developer can confirm which one fits your setup.
8. Keep a log. For each run, record the file name, the supplier, the date, the number of rows read, the rows flagged, the rows rejected and the approver. This log is what you will check when a price on an invoice does not match your system.
9. Run the new process alongside the old one for two cycles. Compare the automated result with a manual check on the same list. Only switch off the manual check when the differences are explained. Then add the next supplier, one format at a time.

What breaks, and how to prevent it

Most failures are not dramatic. A price goes in wrong, quietly, and nobody notices until a customer complains or an invoice comes in higher than expected. The risks below are the ones we would plan for first.

• OCR misreads digits on scanned PDFs, for example reading 8 as 3. Prevention: keep a confidence score on every row, send low-confidence rows to review, and compare the sum of the rows with a total printed on the list where one exists.
• A supplier changes the layout, and the columns move. Prevention: map columns by header name, and stop the run with an alert if an expected header is missing. A run that silently reads the wrong column is the worst outcome.
• Supplier codes change, often when a supplier restructures its catalogue. Prevention: unmatched rows go to a queue, and someone reviews the cross-reference table once a month.
• Pack sizes change, for example from a single kilogram to a 25 kilogram bag. The price looks like a fall of 96 percent. Prevention: compare unit and pack size on every row, and never calculate a change across a pack change.
• The GST basis changes between revisions. Prevention: store the basis in the cross-reference table and reject a row where the basis differs from the last approved one.
• Items under negotiation carry a different price from the list. Prevention: mark them as excluded until the negotiation closes, so the list does not overwrite a contract rate.
• The buyer starts approving everything without reading. Prevention: keep the limit tight and track how often rows are rejected. A reject rate of zero is a warning sign.

Automation has hard limits, and they should be stated once and plainly. It does not negotiate, it does not tell you why a supplier raised a price, and it does not decide whether to change supplier. It moves the data, routes the decision to a person, and keeps the record. Scanned, handwritten or photographed lists will always need a human check.

When not to automate

Automation is not worth building in every case. If you receive one list a month, from one or two suppliers, with under about 50 lines, a shared sheet with a percentage formula and a checklist will do most of the job. My rule of thumb is that if the whole update takes less than about two hours a month, the build cost and upkeep will outweigh the saving.

Hold off if your item codes are not consistent across your own system, because the matching step will fail and you will spend the time fixing codes instead. Hold off if prices are agreed by phone and the list is only a rough guide. Hold off if no one in the business owns the cross-reference table, because it goes stale within a quarter without an owner. In each of these cases, fix the basics first.

What it typically takes

The cost that matters most is time, not software. For a single supplier format, a realistic plan is about two weeks of work: a few days to build the cross-reference table and collect three real files, a few days to build the reader and the flagging rule, and a week of running the old and new processes side by side. These figures are my estimate from the steps above, not a benchmark. Each later supplier format usually takes less time than the first, because the matching table and approval routine already exist.

Staff time changes too. A buyer who used to spend most of a day on each list should, once the process runs, spend a short session on the flagged rows and the unmatched queue. An accounting person needs time at the start to set up price levels or import templates in Tally or another system, and then a short check each cycle that the log matches the books.