The Till With a Name On It: Counters, Cashiers & the Handover
A till is not a place in your shop. It is a record with a name, a warehouse, a location and exactly one person attached to it — and almost every argument about a short drawer traces back to one of those four fields being wrong.
The shop has three tills. On paper that is three tills. In the records it might be one counter used by four people, or five counters where two have never been touched, or three counters all pointing at the same warehouse when one of them is actually at the branch in Nakuru. Nobody notices until a drawer is short and the question "whose till was this?" turns out to have no answer.
This is the least glamorous part of running a point of sale and the part that decides whether every other control works. Cash accountability is not a policy you write on the wall. It is a chain of records: this counter, this person, this drawer, this shift, this float, this count. Break any link and the chain holds nothing.
The rule that everything else rests on
A counter in AWRA OpsHub belongs to one named user. A sale can only be completed on a counter assigned to the person logged in — if it is not their counter, the sale is refused rather than recorded against the wrong name. This is deliberate, and it is the reason a shortage can always be attributed to a person rather than to "the evening shift".
What a counter record actually holds
A counter is four fields, and each one answers a different operational question. Getting them right at setup takes ten minutes per till and saves the same argument every month for years.
The four fields on a counter, and what each one decides
Counter name
What appears on the receipt and in every report that groups sales by till. Name it after the physical position — "Front Till 1", "Pharmacy Counter" — not after the person currently standing there. People move; tills do not.
Assigned user
The one user who can ring sales on this counter. Not a role, not a group — a single named user. When someone else needs to sell here, the assignment changes, and that change is a deliberate act rather than a shared login.
Warehouse
Where the stock leaves from when this till sells. The available quantity checked at checkout is the quantity at this warehouse, so a counter pointed at the wrong branch will happily refuse sales of stock sitting a metre away.
Location
The bin, shelf or sub-location within the warehouse, where you track at that depth. If it is blank, the default location for the warehouse is resolved instead — which is usually right and occasionally the source of a puzzling stock movement.
The assigned user is the field people get wrong most often, because the instinct is to create one counter per till and let everyone use it. That instinct is exactly what makes a shortage unattributable.
One drawer, one counter, one shift
A counter is the till. A cash drawer session is a shift at that till. The two are different objects on purpose: the counter outlives everyone who works it, and the drawer session is a closed period with an opening figure, a closing count and a name against both.
Only one drawer can be open on a counter at a time. That single constraint does more for cash discipline than any amount of training, because it makes the overlap that hides shortages structurally impossible — you cannot have two people accountable for the same physical cash at the same moment.
-
Open with a declared float
The cashier opens the drawer and states the opening float. That number is recorded against them with a timestamp. If the float is wrong at open, every figure downstream is wrong by the same amount — so it is counted, not assumed.
-
Sell
Every sale on that counter attaches its payment to the open drawer session. Cash, M-Pesa and card are all recorded, but only cash affects what should physically be in the drawer.
-
Drop cash mid-shift
When the drawer gets heavy, cash goes to the safe as a recorded drop with an amount, a note, a time and a person. The expected drawer figure falls by exactly that amount. A drop that is not recorded looks identical to a theft.
-
Count, then close
The cashier counts the physical cash and enters the figure. The system computes the expected amount, records the variance, and locks the session with the closing time and the person who closed it.
-
Hand over by closing, not by swapping
The next shift opens a new drawer with a new float. There is no "take over the drawer" — because a period with two people in it cannot be reconciled to either of them.
The handover is the control
Most retail cash losses are not dramatic. They are the accumulated residue of handovers where nobody counted, so a small shortage on Monday was carried by Tuesday, absorbed by Wednesday, and by Friday belongs to nobody. A clean handover is five minutes and it terminates that mechanism.
What each side of a shift change is responsible for
The outgoing cashier
Closes a period they can defend.
- Records any cash already sent to the safe as a drop, before counting.
- Counts the physical drawer and enters the figure — before looking at the expected number.
- Explains a variance in the closing note while the shift is still fresh in memory, not next week.
- Closes the session, which stamps the time and their name on the count.
The incoming cashier
Opens a period they are willing to own.
- Counts the float they are being given, rather than accepting a stated figure.
- Opens a new drawer session under their own login with that counted float.
- Refuses a float that does not match what is physically there — a disputed opening figure is cheap to fix now and impossible to fix at closing.
- Rings every sale under their own counter assignment for the whole shift.
What actually crosses the counter at handover
- The physical cash, counted by both people.
- A closed session with a recorded variance, which is now a historical fact.
- A new session with a counted opening float, which is now the new baseline.
Notice what does not cross: responsibility for the previous shift's variance. That is the entire point.
Why you have two variance numbers, and why they disagree
This catches people out, so it is worth being precise. There are two places a cash variance appears, they are computed differently, and both are correct for their own purpose.
A drawer close, worked through
The drawer figure counts cash by payment date inside the session window, and nets change given against cash received so a 1,000 note tendered for a 700 sale contributes 700, not 1,000. Cash refunds paid out during the shift reduce the expected figure too, because the money genuinely left the drawer.
The shift summary report answers a different question — "how did this shift go?" — and so it slices differently. It windows on when the sale was made, across a start and end time you type in, and sums the opening floats of every drawer session that overlaps that window. The cash figure you compare against is typed into the report form rather than saved.
| Drawer close | Shift summary report | |
|---|---|---|
| What it is | A permanent record of one drawer session | A live report over a time window you choose |
| Cash counted by time of | The payment | The sale |
| Opening float used | That one session's float | Every overlapping session's float, summed |
| Counted cash | Saved on the session forever | Typed into the form, not saved |
| Variance | Stored, with the closing note | Recomputed each time you load the page |
| Tolerance | None — any variance closes | Flags anything outside a tolerance you set (default 100) |
| If a drawer is still open | Not applicable | Marked provisional, because cash is still moving |
So when the two numbers differ, the usual cause is a sale rung near the boundary — made at 17:58 and paid at 18:02 — or a report window that does not match the actual drawer session. Neither number is broken. Use the drawer close as the accountability record and the shift report to look across counters and cashiers.
What we do and do not do
What AWRA OpsHub does today
- Counters with a name, an assigned user, a warehouse and an optional location.
- A hard rule that sales can only be rung on a counter assigned to the logged-in user.
- One open cash drawer per counter, opened with a declared float under a named cashier.
- Cash drops with amount, note, timestamp and the person who made them.
- Expected cash computed as float plus cash sales less drops, with change netted off, and the variance stored on the session.
- A refusal to sell at all when no drawer is open for that counter and cashier.
- A drawer register you can filter by status, counter, cashier and date, and export to CSV or PDF.
- A shift summary across counters and cashiers, with a variance tolerance flag and a provisional marker while drawers are open.
- POS sales that cannot be deleted — reversals go through returns, so the trail survives.
What it does not do
- Any approval or block on closing a drawer with a large variance — it records the number and closes.
- A blind-count enforcement: nothing technically stops a cashier seeing the expected figure before entering their count, so this stays a supervision habit.
- Denomination breakdowns — you enter one counted total, not a note-and-coin tally.
- Multiple cashiers sharing one open drawer, or a formal "take over this drawer" handover action.
- Physical cash-drawer hardware control — we do not fire a kick-out signal to open the till.
- Automatic reconciliation of the safe itself; drops record cash leaving the drawer, not the safe's own balance.
The absence of a variance approval step is a choice we would revisit if asked, but the discipline it replaces is cheap: a supervisor who reads the drawer register daily catches more than a threshold ever will.
Settle these before you open tomorrow
Four questions with answers you can act on today
How many counters exist, and how many people can ring a sale on each?
The answer you will usually get
"Three tills, everyone uses whichever is free."
What to do about it
That is one counter per till with a rotating assignment, or shared logins. Create a counter per till, assign it to the person on that till, and change the assignment when the rota changes.
Who counted the float this morning?
The answer you will usually get
"It was the same as last night."
What to do about it
An uncounted float means the first variance of the day is inherited, not earned. Count at open, every open, and record the counted figure.
Where does the cash go when the drawer gets heavy?
The answer you will usually get
"The supervisor takes it to the safe."
What to do about it
If that movement is not recorded as a drop, the drawer is short by exactly that amount at close and the cashier carries it. Record every drop at the moment it happens.
When was the last time anyone read the variances across a week?
The answer you will usually get
Silence, or "we look when it is big."
What to do about it
A pattern of small consistent shortages on one counter is more informative than one large one. Read the register weekly, by cashier and by counter.
None of this requires new hardware or a new policy document. It requires that the four fields on each counter are right, that a drawer is opened and closed by the person accountable for it, and that somebody reads the variances. The related mechanics are covered in POS shift reconciliation, the stock side in POS and inventory as one record, and what happens when the connection drops in offline POS in Kenya.
Cash accountability that names a person
Counters assigned to named users, one open drawer at a time, recorded drops, and a stored variance on every close — so a short till is a conversation with one person rather than an argument with a shift.
See POS in AWRAFrequently asked questions
Can two cashiers share one till?
Not in the same drawer session, and not on the same counter assignment. A counter belongs to one user, and only one cash drawer can be open on a counter at a time. If two people genuinely work one physical till across a day, they hand over by closing one session and opening another — each with its own counted float and its own recorded variance. That is what makes a shortage attributable to a person rather than to a shift.
What happens if a cashier tries to sell on someone else's counter?
The sale is refused. The counter lookup at checkout requires the counter to be assigned to the logged-in user, so a sale cannot be recorded against a till that is not theirs. This prevents the most common source of unattributable cash: a sale rung under whoever happened to be logged in on that machine.
Do we have to open a cash drawer even if we only take M-Pesa?
Yes — sales are blocked until a drawer is open for that counter and cashier, regardless of tender. It is a shift record, not just a cash record: it is what ties every sale, including card and M-Pesa, to a named person and a defined period. If you take no cash, open with a zero float and the expected cash will simply be zero.
Does the system stop a cashier closing a drawer that is short?
No. The variance is computed, recorded permanently on the session alongside the closing note, and the drawer closes. There is no approval step or threshold block on the close itself. The shift summary report does flag variances outside a tolerance you set — it defaults to 100 — but that is a flag to read, not a gate. If you want an approval gate on large variances, say so; it is a small change, not a redesign.
Why does the shift report show a different variance from the drawer close?
Because they window on different events. The drawer close counts cash by payment date within that one session and uses that session's float. The shift report counts by sale date within a start and end time you type, and sums the floats of every session overlapping the window. A sale made at 17:58 and paid at 18:02 lands on different sides of the two boundaries. Use the drawer close as the accountability record and the shift report for comparing counters and cashiers.
Can we record cash going to the safe during a shift?
Yes, as a cash drop — amount, note, timestamp and the person who made it. The expected drawer figure falls by exactly that amount, so the cashier is not left carrying money that has already left their custody. Drops are the single most commonly skipped step and the single most common cause of a "short" drawer that was never short.
