POS Shift Reconciliation: Closing a Till Without an Argument
The end-of-shift till count is where most Kenyan retail disputes happen, and almost all of them are arithmetic nobody wrote down. The float, the drops, the expected figure and the variance — and why the count must come before the number.
Every Kenyan retailer has had this conversation. It is 8pm, the cashier is counting, the number is short, and nobody can establish whether the shortage is theft, a wrong change given at 2pm, a sale rung up incorrectly, or the float never having been what everybody assumed it was at the start of the day.
The argument is not really about the money. It is about the fact that four different explanations are all equally available, which means the conversation cannot be resolved and both parties leave it unhappy. Fixing that is mostly a matter of writing four numbers down in the right order.
A shift is a closed accounting period
The mental shift that solves this is treating each cashier's shift on each till as its own small accounting period with an opening balance, movements and a closing position — rather than as a vague stretch of time during which sales happened.
Which means a shift has four numbers, and each one has to be recorded at the moment it exists rather than reconstructed at the end.
| Number | When it is recorded | What goes wrong without it |
|---|---|---|
| Opening float | At the start, by whoever hands over the till | Nobody can agree what was in the drawer to begin with |
| Cash sales for the shift | Continuously, by the system | The expected figure is somebody's recollection |
| Cash drops | Each time money is removed to the safe | Cash legitimately removed mid-shift reads as a shortage |
| Counted cash | At close, by the cashier, physically | The count gets adjusted to match the expectation |
The third row is the one most Kenyan retailers miss, and it produces the most unnecessary accusations. In a busy shop a supervisor removes cash to the safe two or three times a day so the drawer is not holding a large sum. If those removals are not recorded as drops against the shift, the till comes up short by exactly the amount that was correctly and honestly removed — and now somebody is being questioned about a variance created by good practice.
A till that closes cleanly
Illustrative, in KES. Without the drops recorded, the same shift reads as 60,150 short and the conversation becomes an accusation. With them, it is a 150 shilling question somebody can probably answer — wrong change on a single sale.
Most till "shortages" in Kenyan retail are cash that was legitimately moved to the safe and never recorded as moved. The accusation is manufactured by the missing record, not by the missing money.
Count before you see the expected figure
This is the single most important discipline in till reconciliation and it costs nothing to adopt. The cashier counts the physical cash and commits that number before anybody looks at what the system expected.
The reason is not suspicion of cashiers. It is that a person who can see the target will, entirely unconsciously, count towards it — recount until it agrees, or decide a note was miscounted. That is human, universal, and it destroys the only signal the exercise produces. A blind count that comes up 150 short tells you something real; a count taken with the expected figure visible tells you almost nothing, because it almost always agrees.
This is a practice, not a lock
AWRA records the opening float, the cash sales, the drops, the counted cash and the resulting variance per session — but counting before looking is a discipline your supervisors have to keep rather than something the system physically prevents. Say it out loud when you set the process up, because the value of every variance figure you produce afterwards depends on it.
Per cashier and per till, not per day
A daily reconciliation at branch level tells you the branch was short. A shift-level reconciliation tells you which till, which cashier and which four-hour window — and that difference is what makes a variance actionable rather than absorbable.
It also changes what a pattern looks like. One cashier consistently 200 short is a training issue about giving change. One till consistently short regardless of who works it is a hardware or process issue. Shortages clustered on Saturday afternoons is a staffing and supervision issue. None of those patterns are visible in a daily branch total, and all three have completely different responses.
Daily, at branch level
- The branch was short by some amount
- Nobody can attribute it to a person or a period
- The number gets absorbed because it cannot be investigated
- Absorbing variances teaches everyone they do not matter
- Patterns are invisible inside a single daily figure
Per shift, per till
- A named cashier, a specific till, a bounded window
- The question is small and usually answerable
- Variances get explained the same evening
- Explaining variances teaches everyone they are watched
- Patterns by person, till and day surface within weeks
Mobile money and card belong in the same close
A cash-only reconciliation is only part of the job in Kenya, where a large share of takings arrive by mobile money and card. Those tenders do not sit in the drawer, so they cannot be counted — but they still have to be reconciled against system sales, and they have their own characteristic failure.
The failure is timing: a mobile money payment received at 7:55pm may settle the next morning, and a card batch may settle a day later, so a naive comparison shows a shortfall that is purely a settlement lag. Reconciling each tender separately against system sales, and tracking settlement separately from receipt, is what stops that lag being investigated as a loss every single evening. The branch-level daily discipline including M-Pesa is covered in the retail daily close.
Returns and discounts change the expected figure
Two things routinely make a till appear short when nothing is wrong. A cash refund reduces the cash in the drawer without reducing sales, and a discount reduces the sale value. Both are legitimate; both need to be recorded as what they are rather than absorbed into a variance.
A return recorded properly carries the original sale, a reason, the refund tender, whether the goods were restocked and who processed it — which means a cash refund is visible in the shift reconciliation as a cash outflow rather than as an unexplained shortage. Discounts are recorded against the sale and the line with a description and an amount, so a heavily discounted shift is legible rather than mysterious. Worth knowing honestly: discounts at the till are recorded and attributable, not gated by an approval step — if you need authorisation before a discount is given, that has to be a supervisory practice rather than something the till enforces.
What we do and do not do
What AWRA OpsHub does today
- Cash sessions per counter and per cashier, with opening float, opened/closed timestamps and who opened and closed them.
- Cash drops recorded individually with amount, time and who removed the money.
- Expected versus counted cash, with the variance computed and stored on the session.
- Notes on the session, so an explanation lives with the variance rather than in somebody's memory.
- Returns linked to the original sale, with reason, refund tender, restock decision, exchange support and who processed it.
- Discounts recorded against the sale and the line with a description and amount.
- Every sale moving stock, so the shift reconciles on goods as well as on cash.
What it does not do
- The blind count is a discipline, not an enforced lock — the system records the counted figure but does not physically prevent someone seeing the expectation first.
- No discount approval gate at the till. Discounts are attributable after the fact; pre-authorisation is a supervisory practice.
- No cash-drawer hardware control — we do not open or lock drawers, and we do not supply tills.
- No automatic settlement matching for card batches or mobile money against bank credits; those are reconciled from the records rather than matched by an integration.
Because the count is not physically blinded, the credibility of your variance data rests on the supervisory habit. It is worth stating explicitly in your till procedure.
The habit that makes it stick
-
Hand over the float formally
The outgoing and incoming cashier both present at handover, float counted and recorded. Most disputes start with an opening balance nobody agreed.
-
Record every drop as it happens
Not at the end from memory. A drop recorded three hours late is indistinguishable from a drop that never happened.
-
Count blind, then reveal
Cashier commits the counted figure, then the expected figure is shown. This is the whole integrity of the exercise.
-
Explain the variance the same evening
While the shift is reachable and the day is remembered. A variance examined next week is unanswerable and will be absorbed.
-
Read patterns weekly, by cashier and by till
Not to punish, but because a repeated 200 shortage is a change-giving training gap and a repeated till problem is not a people problem at all.
Our take
Record the float at handover, record every cash drop as it happens, and have the cashier count before seeing the expected figure. Those three turn the 8pm argument into a small arithmetic question that is usually answerable — and they cost nothing beyond insisting on the order of operations.
See a till that closes without an argument
Cash sessions per cashier and counter, opening float, recorded drops, expected versus counted with the variance stored, and returns that show as cash out rather than shortage.
Explore shift reconciliationFrequently asked questions
Why does our till come up short so often?
In most Kenyan retailers the largest single cause is cash legitimately removed to the safe during the shift and never recorded as a drop. The till is then short by exactly the amount that was correctly removed, and somebody gets questioned about a variance created by good practice. Record every drop as it happens, with the amount and who removed it, and a large share of apparent shortages disappear immediately.
What is a blind count and why does it matter?
The cashier counts the physical cash and commits that figure before anyone looks at what the system expected. It matters because a person who can see the target counts towards it — recounting until it agrees, or deciding a note was miscounted — entirely unconsciously. A blind count that comes up short tells you something real; a sighted count almost always agrees and therefore tells you almost nothing. Note that this is a discipline your supervisors keep, not a lock the system enforces.
Should we reconcile per shift or per day?
Per shift and per till. A daily branch figure tells you the branch was short but attributes it to nobody, so it gets absorbed — and absorbing variances teaches everyone they do not matter. Shift-level reconciliation gives you a named cashier, a specific till and a bounded window, which makes the question small and usually answerable. It also reveals patterns that are invisible in a daily total: one cashier consistently short is a training issue, one till consistently short regardless of operator is not a people problem at all.
How do mobile money and card fit into the close?
They cannot be counted because they are not in the drawer, so reconcile each tender separately against system sales rather than lumping them together. Their characteristic problem is settlement lag — a payment received at 7:55pm may settle the next morning — so track settlement separately from receipt, otherwise you investigate a timing difference as a loss every evening. There is no automatic settlement matching against bank credits; this is reconciled from the records.
Can the system stop a cashier giving an unauthorised discount?
Not at the point of sale. Discounts are recorded against the sale and the line with a description and an amount, so they are fully attributable afterwards and a heavily discounted shift is legible — but there is no pre-authorisation gate at the till. If you need approval before a discount is given, that has to be a supervisory practice. What the records do give you is the ability to see discount behaviour by cashier over time, which is usually where the real problem shows up.
