You counted thirty dozen eggs into the fridge on Monday. On Friday the shelf says there should be eleven, and you can find nine. Two are gone with no sale, no note, and no one to ask — because the whole point of the stand was that nobody had to be there.
A wrong number on its own tells you nothing. It doesn't say whether two cartons were lifted, miscounted on Monday, sold while the till was offline, or moved to the Saturday booth and never written down. On a staffed register you'd ask the person who was standing there. At an unattended stand there is no person — so the only thing that can answer the question is a record of every change the count went through. Not the number. The history behind it.
This is a guide to reading that history: what a stock ledger actually logs, how to walk it backward to the thing that broke your count, and the handful of habits that stop most drift before it starts.
A count is a conclusion. A ledger is the evidence
Most small sellers track stock as a single number per product and update it by hand, or let it ride and recount when it feels wrong. That number is a conclusion — "I think I have eleven" — with nothing underneath it. When it's wrong, you have no way to tell which of a dozen ordinary events bent it, so the only fix is to recount and start the guesswork over next week.
A ledger flips that. Every event that can move stock is written as its own line, and the on-hand number is just the running total of those lines. In TallyTill each movement carries four things:
- What kind of event it was — a sale, a refund, a physical count, a harvest, a delivery received, a move between locations, a manual adjustment, or an import.
- A signed quantity — minus three because three sold, plus forty because a crate came in, or a hard "set to" when you counted.
- Where it happened — which stand, booth or fridge, because the same product has its own count at each place you sell from.
- Who or what did it — the device that rang the sale, the Portal user who keyed the count or move, or the system itself when it reconciled an offline sale.
The on-hand you see is never a standalone figure you have to trust. It's the sum of those lines, newest first, and any one of them can be opened and read.
Walk the ledger backward to the cause
When a number looks wrong, you don't recount the whole stand. You open that one product's history and read down from the top. The cause almost always announces itself by where in the timeline the error sits and what kind of line sits next to it.
That last step is the one that matters. The ledger's job isn't to catch a thief — it can't; it only sees events that were recorded. Its job is to subtract away everything that has an innocent explanation, so the piece you're left with is small, specific, and tied to one product at one place in one window of time. "Something's wrong with my inventory" becomes "six jars of honey went missing from the North Road fridge between Tuesday and Thursday" — a problem you can actually act on.
The movement types, and what each one tells you
Every line in the history is one of these. Knowing which is which is most of the skill of reading it.
| Line | What it means | What it tells you when the count is off |
|---|---|---|
| Sale | An item was rung up and paid for | Draws down the location it sold at. A missing-looking drop that matches real sales is not a problem — it's the system working |
| Refund | Money (and stock) given back | Adds stock back. A refund without a matching earlier sale is worth a look |
| Count | You did a physical count | A hard reset to what you counted. If a count was taken during an outage, it can be the source of the error, not the fix |
| Harvest / Receive | New stock came in | Lands at a location, and can carry a lot label and a pack date. Over-received or box-vs-unit errors start here |
| Move | Stock went from one place to another | Both sides change at once. A move you forgot to record is the classic "it's wrong at the stand but fine at the booth" |
| Adjustment | A manual correction | Visible and dated on its own line, so a "set to" you made at 11 p.m. is never invisible later |
| Import | Stock loaded in bulk | Usually only at setup or when switching from another system. A bad import shows as one big dated line |
Because receive lines can hold a lot and a pack date, you can also narrow the whole history to a single lot — useful when a batch of eggs or a run of a prepared item needs rotating out first, or when you simply want to know when the stock currently on the shelf was packed. (That's an operational record for your own rotation, not a substitute for your state's egg, dairy or meat labeling rules — follow those on their own terms.)
"Who did it" is only as honest as your logins
The attribution is where a ledger earns its keep on a shared operation — and where most setups quietly break it. Portal actions like counts, moves, adjustments and catalog edits are tied to the Portal account that made them. Sales and the stock they draw down are tied to the device that rang them and the location it's set to. The system tags its own entries, like an offline sale reconciled after the fact.
The failure almost everyone makes is the same one every inventory guide warns about: one shared login for everybody. When five people manage stock through a single account, the "who" column says the same name on every line and answers nothing. If more than one person touches your catalog or keys counts, give each their own Portal login. It costs nothing and it turns "who set this to zero?" from an argument into a line you can read.
For an unattended stand this matters less than you'd think, because there's no staff — the device is the accountable party, and each till draws down only its own location. The attribution you lean on there is "which fridge, which day," and that's exactly what the ledger gives you.
The one cause that isn't anyone's fault: the signal dropped
In a field in July there is usually no signal, and spotty internet is the single most common reason a count looks wrong on a small stand. A till that goes quiet for an hour hasn't stopped selling — and if it stopped counting while it sold, every number it touches drifts until you notice.
TallyTill's tills keep selling and keep counting locally while they're offline, then reconcile those movements to the cloud when the connection comes back. The sales don't vanish and the count catches up on its own; they land in the history as real, dated lines, tagged as reconciled by the system. The practical rule that follows: don't take a physical count on a till you know is offline. A count is a hard reset, and resetting to a number that doesn't yet include the morning's unsynced sales bakes the error in. Let it reconcile first, then count.
Stop most of the drift before it starts
You can't make shrink impossible, but you can make the recordable causes rare, which shrinks the unexplained remainder to something small enough to actually chase.
- Let sales deduct themselves. The count should move when a sale closes, not when someone remembers to update a spreadsheet. A number kept by hand is a number that's already wrong.
- Receive at the unit you sell. If you buy a case and sell singles, record the singles. Over-received and box-vs-unit errors are the leak that starts at the loading dock and isn't found until the next count.
- Set each till to its real location. The booth tablet is the booth's till; the fridge's tablet is the fridge's. A till pointed at the wrong place drains a count that was never actually low.
- Do corrections as counts or adjustments — not quiet edits. Both are logged, dated and attributed. A correction that leaves a trace is one you can explain next week; one that doesn't is a future mystery.
- Give real people real logins. The "who" column is worth exactly as much as the number of distinct accounts behind it.
None of this requires a warehouse system, and TallyTill isn't one — there are no purchase orders, supplier lead times or multi-echelon transfers here. It's an honest, per-place stock ledger for the stands, booths, fridges and pickup sheds that make up a small operation, built so the count explains itself when you're not there to watch it.
Start with a ledger you can read
If your stock lives in your head or a spreadsheet, the first wrong number of the season costs you an afternoon of recounting and no answer. A ledger costs you nothing extra — it's the byproduct of ringing sales up instead of dropping cash in a box — and the first time a count drifts, it hands you the cause instead of a shrug.
You can run it on a phone or tablet you already own, at one stand or several. Starting is free — no credit card. Card sales carry a small per-sale fee and cash and checks carry none; current rates are on the pricing page.
Create your free account, add a product or two, and ring your next sale through it. The line it writes — signed, placed, timed and attributed — is the first entry in a record that will, the day your count goes wrong, actually tell you why.