Find the money quietly leaking out
A deterministic, offline scan that estimates where margin slips away — products sold below cost, thin-margin lines, over-discounted invoices and high-return customers — with a rupee figure on each leak.
4 products losing money on every sale
9 invoices discounted over 25%
3 customers with heavy returns
Live product UI — sample data.
Below-cost & thin margins
Flags every product sold below its purchase cost with the exact rupee loss, plus thin-margin lines to reprice — so pricing mistakes surface long before the year-end.
Discounts & returns
Highlights invoices discounted beyond a healthy threshold and customers whose returns eat into their revenue, with the amount at stake on each.
One estimated number
Rolls every leak into a single estimated ₹ figure, ranked by severity and traced to the product, invoice or customer — so you fix the biggest first.
What this actually means in practice
The scan sweeps five registers. It reads every non-cancelled sales invoice line — the taxable value actually invoiced, not the price sitting on the product master — and every purchase invoice line, from which it derives a weighted-average unit cost per item. It then reads invoice-level discount against the pre-discount total, credit notes by party, the expense register and the vendor payment register. Line-level detail is what makes it work: a discount that pushed one line under cost is caught even while the master price still looks healthy.
Every finding is a stated rule applied to those rows, never a judgement. A line has sold below cost when its taxable value falls under quantity times weighted-average cost, and the loss reported is exactly that difference. Under ten per cent gross margin is thin. More than twenty per cent off an invoice is excessive discounting. A party returning over a quarter of what they bought is flagged. Identical amount, same party, same day, entered twice is a probable duplicate. Same books, same findings, every single run.
The output is a ranked worklist with a rupee figure on each group and the exact product, invoice, customer or payment behind every line, so the first hour of work goes to the biggest hole rather than the loudest complaint. Typical outcomes are repricing a line, tightening who may discount, reviewing a returns-heavy account, or cancelling a payment made twice. The word possible on duplicates is deliberate — two genuine identical same-day bills do exist, so the engine flags rather than reverses. Nothing is written back.
| Leak | Rule applied | Rupees reported |
|---|---|---|
| Sold below cost | Line taxable value under qty times weighted-average cost | The exact shortfall per product |
| Thin margin | Gross margin above zero but under 10% | Flagged for repricing, no loss figure |
| Excessive discount | Invoice discount over 20% of the pre-discount total | The full discount given |
| High-return customer | Credit notes over 25% of that party's billings | Value returned |
| Duplicate invoices | Same customer, same total, same day | Value of the extra copies |
| Duplicate expenses | Same vendor or category, same amount, same day | Value of the extra entries |
| Duplicate vendor payments | Same vendor, same amount, same day | Value of the extra payments |
What’s included
- Below-cost sales with exact ₹ loss
- Thin-margin product detection
- Excessive-discount invoices
- High-return customer flags
- Total estimated ₹ leak
- Severity-ranked, traced to source
- Runs offline · deterministic
What it does not do
- · Every finding is a candidate, not a verdict. Below-cost clearance can be deliberate and two identical same-day bills can both be genuine — the scan surfaces them for a person to confirm.
- · Costing depends on purchases being entered. Where an item has no purchase history the product master's purchase price stands in, and the margin verdict is only as sound as that figure.
- · The thresholds are fixed rules — ten per cent margin, twenty per cent discount, a quarter returned — not learned from your trade. A business whose normal margin sits under ten per cent will see thin-margin lines that are simply its business.
- · It reports and ranks; it never reverses a payment, cancels an invoice or changes a price. The scan ships in every plan from ₹249; only the conversational layer is Enterprise.
Frequently asked
How is the leak amount calculated?
Deterministically from your own books: below-cost loss is quantity × (cost − sale price), the discount leak is the amount discounted beyond the threshold, and returns are summed per customer. Every figure traces to the exact product, invoice or customer.
Does it use AI or the internet?
No. It is a deterministic, offline computation — no LLM and no third-party service — so a re-run gives the same result.
Which plan includes it?
Revenue-leak detection is deterministic and part of the built-in insights, not the Enterprise conversational AI add-on. See pricing for details.
Does the scan need the internet or an AI subscription?
Neither. It is arithmetic over your own posted registers, so it runs on the ₹249 plan with no language model involved and returns the same groups and rupee totals on every run. An Enterprise copilot can summarise the findings in prose, but the figures it quotes are these ones, entirely unchanged.
Why does a product show a loss when its price list looks fine?
Because the scan costs the line that was actually invoiced, not the master price. A scheme, a line discount or an override at the billing screen can push a single sale under weighted-average cost while the price list stays healthy — precisely the case a period-end margin report averages away and this one keeps visible.
How is ai · deterministic handled differently here?
It rests on below-cost & thin margins, discounts & returns and one estimated number. The distinction that matters is that these are structural rather than cosmetic — the behaviour is built into how records are posted, not layered on as a report you have to remember to run.
Does AI · deterministic post to the real books?
Yes, and that is the whole reason it lives in this product rather than beside it. What below-cost & thin margins records lands in the same posted double-entry ledger the Trial Balance, P&L and GST returns are built from — so there is no second set of ai · deterministic numbers to reconcile against the first.
How does AI · deterministic affect my GST?
It feeds it directly rather than sitting alongside it. CGST/SGST and IGST are split from HSN or SAC and place of supply as each document is raised, so whatever ai · deterministic produces is already correct when GSTR-1, 3B and 2B reconciliation are drawn off the same books. The return matches the accounts because it was never a separate exercise.
Can I try AI · deterministic before committing?
Yes — 14 days, no card. Test ai · deterministic against your own masters and your own transactions rather than sample data: the questions worth answering here are about your business's edge cases, and a demo dataset is built not to have any.
Can I get my ai · deterministic data out again?
Yes. On the Desktop edition everything AI · deterministic records sits in a local database on a disk you choose, with scheduled encrypted backups you control — stopping payment leaves you holding a readable file. On the Cloud edition your data is isolated to your business by row-level security and exportable from the reports it feeds.
Find the money slipping away
Every sale is scanned for below-cost pricing, over-discounting and habitual returns.
- ₹68,200Below-cost saleOrder #4471
- ₹41,500Excessive discountRao Traders
- ₹32,300High-return customerMehta Stores
Illustrative — sample scan.