One counter, one float, one named cashier, a sealed period, and an amount of physical money that either matches or does not. Everything difficult about cash is contained in that last clause — and the only way to make the difference meaningful is for the expected figure to be computed from the sales that actually belonged to that session, rather than typed by whoever is closing up.
Why cash is different
A card payment leaves a trail at the acquirer. A mobile money payment leaves one at the operator. Neither can quietly become three hundred and fifty shillings less on the way from a customer's hand to a safe. Cash can, and it does — through error far more often than through theft, which is exactly why the control has to be arithmetic rather than suspicion.
The failure is almost never a missing amount. It is a missing period: nobody can say which stretch of trading the money in the drawer was supposed to correspond to, who was standing at it, or what the float was when they started. Once those three facts exist, a variance is a small conversation. Without them it is unanswerable, so nobody asks, and the answer becomes “cash is always a bit out”.
Four moments
There is no fifth. No editing the float after trading has started, no adjusting the expected figure to make the variance go away, no reopening a closed session to tidy the number. The shape is deliberately rigid, because a cash control with a back door is a cash control that gets used through the back door.
Somebody selects the counter, counts the starting cash, and records it. The session is stamped with who opened it, the moment it opened, and the cashier it belongs to. A note can be attached — and your own custom fields, if your close-out procedure needs a shift label, a branch code or a supervisor initial.
This is the part that has to be automatic. A cash payment taken at that counter, by that cashier, while that session is open, is bound to the session as it is recorded. Nobody assigns it afterwards, and nobody can reassign it — which is what makes the expected figure something other than an opinion.
A busy counter should not be holding a day of takings, so cash is periodically removed to a safe. Each drop records the amount, the moment, the person who made it, and a note. Every drop reduces what should be in the drawer at close — so a till holding less than the day's sales is explained rather than suspicious.
The cashier counts what is physically there and enters that one figure. Everything else — the cash sales total, the drops total, the expected amount and the variance — is computed at that moment and frozen onto the session, so it cannot drift afterwards as new records arrive.
One typed figure
A reconciliation where the expected amount is entered by hand is not a reconciliation — it is two numbers somebody chose. Here the expected figure is derived, and one detail in the derivation is worth calling out because most systems get it wrong.
Every one of those figures is frozen onto the session at close. That matters more than it sounds: a variance computed live from a query would silently change as late payments, refunds or corrections landed afterwards, and a manager reviewing last Tuesday would see a different number from the one the cashier signed off. The close writes them down.
The awkward part, published on purpose
Two open sessions on one physical till would accumulate sales against two floats, and neither would reconcile. So a counter can hold exactly one open drawer. Combined with the rule that a sale must belong to your session, that produces a genuine trap at a shift change — and rather than pretend it does not exist, here it is, with the way out.
The arriving cashier tries to sell: refused, because there is no open session belonging to them. They try to open a drawer: refused, because the counter already has one. Two refusals, neither of which explains the other, and a queue forming.
This is a real failure and it was ours. The rules were both correct; the messages were both useless, because neither named the holder or stated the exit.
The refusal says who is holding the counter and what to do about it — count the cash and close that session — and it says explicitly that you can do this yourself from the cash drawers screen. Because that was always true: any user in the workspace may close a session, not only the cashier who opened it.
Which is the honest resolution: the constraint stays, because it is right, and the arriving cashier is told precisely how to satisfy it in one sentence rather than deducing it from two refusals.
It also means an abandoned drawer is closable by a supervisor with a physical count, so the variance lands on the session it belongs to rather than being absorbed into the next shift. That is the outcome that actually matters: a shift change should not silently merge two people's accountability.
The drawer register
One screen listing sessions newest first, with the computed snapshot on each row so the float, the sales, the drops, the expected figure and the variance are visible without opening anything. The filters are the four questions a manager actually asks.
Open or closed. "Show me every drawer still open" is the first thing anybody wants at the end of a day, because an open drawer means the branch close is provisional.
One till at a time, across weeks. This is how a persistent problem at counter 3 becomes visible as a pattern rather than a run of individually forgivable small variances.
One person's sessions. Used far more often to defend somebody than to accuse them — a cashier with forty clean closes and one bad one has a story that a single incident does not.
From and to, on the opening date. For a period review, a payroll query or an auditor asking about a specific fortnight.
Every session matching the filters you have set, with its float, cash sales, drops, expected, counted and variance. The file for a period review or a request from an auditor.
A single session with its arithmetic, its drops listed individually, and the cash sales that made up the total. The file you attach to an incident, or hand to the person whose count was short.
Cash drawers is one of the eighteen datasets that can be posted on a timer — daily at 07:00 to whoever reviews yesterday, in your own timezone, with every delivery attempt recorded.
The same all-drawer and single-drawer exports are available to the API, gated on the same permissions, for a branch dashboard or a finance system that wants the figures without a person in the loop.
Saved filters work here too, so “open drawers at the Kisumu branch” is a named chip above the list rather than four dropdowns you set every morning.
The straight answer
Those are different products, and the second one involves safes, banking runs and denomination counts. Here is exactly where the line falls.
What AWRA OpsHub does today
More we can add to your workspace
Where we point you to a specialist
Denomination counting and a safe-to-bank ledger are the two items in that middle column that turn this from a till control into a cash office, and they are one design conversation rather than two builds — the hard part is your banking procedure, not the arithmetic. A blind close is the cheapest single improvement for anywhere with a real shrinkage concern. Tell us how your cash actually moves and we will come back with a written spec, a timeline and a price.
Where it sits
The branch-level view over the same sessions: payment method breakdown, cashier and counter contribution, variance tolerance and the daily close.
Open featureThe till itself — the sales, the payments and the counters a drawer session is bound to.
Open featureWhere cash sits alongside card, mobile money and bank transfer in one consolidated register.
Open featureTurning "open drawers at Kisumu" into a chip, and then into a 07:00 email.
Open featureThe POS shift summary and sales summary reports the drawer figures roll up into.
Open featureWhat is recorded around a close, and how the ordering of events can be shown to be intact.
Open featureWhat changes
Not because people stop making mistakes, but because a mistake now has a shift, a name and a figure attached to it within a day. Small variances get explained while somebody still remembers the transaction, which is the only time they can be.
One side of the subtraction is computed, so the difference is not two chosen numbers.
A 1,000 note against an 850 sale puts 850 in the drawer, and the arithmetic knows it.
Every mid-shift drop is on the record with its amount, time and person.
One drawer per counter, closable by anybody, so the variance lands on the shift that made it.
Filter by counter or cashier across weeks and a persistent problem stops looking like bad luck.
Everything is frozen at close, so last Tuesday reads the same today as it did then.
Help Center
Run inventory, procurement, assets, sales, and field work with approved AWRA guidance for setup, migration, integrations, security, pricing, and support.
