How estimates work
StockBrik never asks you to log stock movements. Instead it infers your stock levels from data Brikly already has, and is upfront about how solid each number is. This page explains the machinery in plain terms - no maths degree required.
The equation
For every tracked ingredient, the estimate is:
Estimated stock = last count
+ deliveries since (from matched invoices)
- sales since (from the best available source)
± adjustments
Anchor at the last counted quantity, add what came in, subtract what went out, correct for what you told it about. That is all an estimate is.
Deliveries in
Every invoice processed through Brikly with matched line items feeds StockBrik. When Tuesday's flour invoice is matched, those 2 x 16kg bags land in the flour estimate - quantities, pack sizes, and units all come from the same matching work invoice processing already does for your costings. Supermarket receipts put through Brikly count exactly the same way. Purchases that never reach Brikly are the one blind spot, which is what the Delivery (not invoiced) adjustment is for.
Not every matched line should add to stock, though. Each matched line carries an Adds to stock toggle, on by default: switch it off and the purchase keeps its price history and accounting treatment but never enters your stock levels. That is the right tool for a one-off purchase of something you do track - a case of the soft drinks you sell, bought for the staff fridge. If you never want estimates for an ingredient at all, switch off tracking for the ingredient instead (its price history still builds quietly in the background, ready if you track it later). And skipping a line remains the bluntest option: a skipped line touches neither your costings nor your stock.
Sales out
Sales deplete stock through your CostingBrik recipes: sell a victoria sponge slice, and StockBrik subtracts the flour, butter, sugar, and eggs that slice contains. Where the sales figures come from varies by ingredient - see the source labels below.
Counts anchor
Estimates accumulate small errors over time; that is the nature of inference. A completed stock count wipes the slate: the counted quantity becomes the new starting point and the drift is discarded. Counts are not an admission the system failed - they are a designed-in part of how it works.
The source labels
Each ingredient's estimate is depleted by the best sales signal available for that ingredient, and the Stock Levels page labels it plainly:
| Label | What it means |
|---|---|
| Based on POS sales data | A POS is connected and the items sold are linked to recipes. The strongest signal: actual sales, every day, automatically. |
| Based on your daily sales log | You log portions sold on the Daily Sales page. Nearly as strong, as complete as your logging. |
| Based on estimated weekly sales | No per-day sales data, so StockBrik uses the weekly sales figures on your CostingBrik recipes - a steady average rather than actual days. |
| Based on stock count history | Usage inferred from how fast the ingredient depleted between your stock counts. |
| Based on ordering patterns | Usage inferred from how often, and how much, you buy the ingredient on invoices. The broadest fallback. |
Different ingredients can sit on different sources at the same time - coffee beans on POS data, the cleaning chemicals on ordering patterns - and an ingredient moves up automatically the moment better data exists for it.
Confidence: the three badges
Every estimate carries a badge - Estimated, Approximate, or Count recommended - never a raw percentage. Two things drive it:
- Time since the last count. Confidence drifts down as an estimate ages, because small errors accumulate. A count restores it.
- The quality of the source. Better sources earn higher confidence ceilings. An ingredient depleted by real POS sales can hold Estimated for far longer than one running on ordering patterns, which will never climb as high no matter how recent the count.
So the two levers for trustworthy numbers are simple: feed StockBrik better sales data (connect a POS, or keep the daily sales log complete), and count the important ingredients when the badge asks you to.
StockBrik gives you a well-informed estimate, not a weighbridge reading. Treat Estimated numbers as reliable for day-to-day decisions, treat Approximate with a little care, and treat Count recommended as exactly that. Regular counts are what keep the whole system honest - skip them for long enough and any inference system, this one included, will drift from reality.
What makes estimates better
- Process every invoice through Brikly - the deliveries side is only as complete as your invoice flow
- Connect a POS and link items to recipes, or keep the Daily Sales log consistent
- Record the leaks with adjustments when you see them
- Count when asked - a Count recommended badge is StockBrik telling you precisely where your time is best spent