How each figure is actually worked out.

What it accounts for

• Landed cost, not invoice price. For an imported line, freight, duty and clearance are part of the money tied up, and using invoice price alone understates it.
• Stock on order is counted separately, never folded into this figure. Projected cover after inbound is reported as its own number.
• Case size and minimum order quantity are read before any suggestion to reduce an order, because a reduction that cannot be made is not a recommendation.
• Shelf life and ageing are read. Twenty weeks of ambient rice and twenty weeks of a short-dated line are not the same exposure.
• A line marked as a deliberate bulk buy is excluded, with the reason recorded.

When it produces no figure

If the stock file is stale, or sales history is too short to give a stable velocity, MarginLoop lowers confidence or suppresses the recommendation rather than producing a number from a bad input.

What it accounts for

• Supplier terms are versioned with effective dates. What a business was entitled to last year depends on the agreement that applied last year, not the current one.
• MarginLoop reports that a condition appears to have been met and shows the evidence. It does not assert that a supplier will agree, and it never contacts the supplier.
• A finding of this kind is labelled REBATE TO CHECK while it is a question, and REBATE VERIFIED once the credit note appears in your purchase ledger. MarginLoop reads that transition from the data; nobody is asked to confirm it.
• Rebates are paid on many different mechanics: automatically, quarterly, annually, retrospectively, by credit note, or only on submission. MarginLoop does not assume yours.

When it produces no figure

Where an agreement's terms cannot be read unambiguously, MarginLoop says the terms are unclear and shows what it did read, rather than guessing at a tier structure.

What it accounts for

• This is a timing effect, not a loss and not a bad debt. Nothing here is overdue, which is precisely why a receivables report does not show it.
• Every input is a figure from the affected accounts themselves: their own invoice values, their own invoicing intervals, their own twelve-month average. No annual total is divided by 365, so nothing here assumes invoicing is spread evenly across the year.
• Account identifiers and dates are sufficient. No customer names, addresses, phone numbers or email addresses are needed at any point.
• Seasonal accounts are read against their own seasonal pattern where there is enough history to establish one.

When it produces no figure

With fewer than a handful of invoices, an account has no stable baseline, and MarginLoop reports insufficient evidence rather than a drift figure.

What it accounts for

• The same product can legitimately carry different prices by date, supplier, promotion, pack change or landed cost. A price difference is not by itself evidence of a mistake.
• A confirmed match is remembered as a rule, so the same question is not asked twice.

When it produces no figure

Below the confidence threshold, MarginLoop asserts no figure at all. The example on the home page sits at 78% and deliberately carries no number.

What it accounts for

• Where a UK observance depends on local moon sighting, the date can move by a day, and that is stated rather than hidden.
• Wedding season is treated as a broad, regionally variable pattern rather than as a fixed date, and read from your own order history.
• Product-category assumptions are not hardcoded. What sells at Diwali for one account is not what sells for another.

When it produces no figure

Without at least one prior cycle in your history, MarginLoop has nothing to align and says so, rather than substituting a generic seasonal curve.