Cuedis

Voids, refunds and shrinkage: spotting unusual till activity

Every till has voids and refunds. A customer changes their mind, a cashier scans the wrong item, a product turns out to be faulty. They are a normal part of retail, and almost all of them are honest.

But voids and refunds are also where shrinkage likes to hide, whether from mistakes, sloppy procedures or, occasionally, dishonesty. Because they are normal, they are rarely looked at closely. This post explains how to read them fairly, so you can spot the patterns that deserve a closer look without treating every cashier as a suspect.

Know the terms in your system

Systems use different words, but these are the usual ones:

  • Void (line void): an item removed from a sale before it is paid for.
  • Bill void or cancelled sale: a whole transaction cancelled before payment.
  • Refund or return: money paid back after a completed sale.
  • Price override: the cashier changes an item's price at the till.
  • Manual discount: a discount typed in rather than applied by a promotion.
  • No-sale: the cash drawer opened without a sale.

Each of these is recorded with a time, a till and usually the cashier's login. That record is what makes fair analysis possible.

What normal looks like

Before looking for anything unusual, learn what normal is for each till, shift and person. Some reasons voids and refunds differ perfectly innocently:

  • New staff void more while they learn the system.
  • Busy counters have more of everything, including mistakes.
  • Some products are returned more often: electronics, clothing in the wrong size.
  • Some shifts handle the returns desk.
  • After a festival or sale, returns go up for everyone.

So the useful measure is not the raw number, but the rate: voids per hundred transactions, refund value as a share of sales, for each person and till, compared with their own history and with colleagues in similar shifts.

Patterns worth a closer look

These are the patterns loss-prevention people commonly look for. None of them proves anything on its own, but each is worth a quiet look:

  1. A rate well above colleagues on similar shifts, especially if it started at a particular point in time.
  2. Refunds without a matching earlier sale, or refunds of items the shop rarely sells.
  3. Refunds clustered at quiet times: end of shift, closing time, when fewer people are around.
  4. Many voids just under a supervisor-approval limit. If approval is needed above $50, a run of voids at $48 is worth noticing.
  5. Voids immediately after payment-type changes, or bills voided and re-entered at a lower value.
  6. Price overrides on the same few items, especially high-value ones.
  7. No-sale drawer openings well above the usual number.
  8. A sudden change in one person's pattern. Someone whose rate was always normal and has recently tripled.

Looking at it in your own data

A simple first step is the void or refund rate per cashier per week. Here is an example for MySQL, assuming a sales table with a cashier, a date, a total and a status:

SELECT cashier_id,
       YEARWEEK(sale_date, 3) AS week,
       COUNT(*) AS bills,
       SUM(status = 'VOID') AS voided,
       ROUND(100 * SUM(status = 'VOID') / COUNT(*), 1) AS void_pct
FROM sales
WHERE sale_date >= CURDATE() - INTERVAL 12 WEEK
GROUP BY cashier_id, YEARWEEK(sale_date, 3)
ORDER BY cashier_id, week;

The SQL Server version uses CASE instead of adding up a true/false value:

SELECT cashier_id,
       DATEADD(week, DATEDIFF(week, 0, sale_date), 0) AS week_start,
       COUNT(*) AS bills,
       SUM(CASE WHEN status = 'VOID' THEN 1 ELSE 0 END) AS voided,
       ROUND(100.0 * SUM(CASE WHEN status = 'VOID' THEN 1 ELSE 0 END) / COUNT(*), 1) AS void_pct
FROM sales
WHERE sale_date >= DATEADD(week, -12, CAST(GETDATE() AS date))
GROUP BY cashier_id, DATEADD(week, DATEDIFF(week, 0, sale_date), 0)
ORDER BY cashier_id, week_start;

Look for people whose rate is consistently far from their colleagues', and for any sudden jump in one person's own rate.

Be fair, and be careful

Data like this is about people, so treat it with care:

  • A number is a reason to look, never proof. There is almost always an innocent explanation worth hearing first: training, a faulty scanner, a difficult customer, a manager asking them to process returns.
  • Compare like with like. The returns desk and the express till are not comparable.
  • Fix procedures before blaming people. If voids above $50 need approval, make sure approval is actually happening, and that the limit makes sense.
  • Limit who sees it. Till-by-person analysis should be visible to the owner or manager who needs it, not the whole company.
  • Mind personal data. In Singapore, the PDPA applies to staff data too. Use it for the purpose it was collected for, and keep it secure.

Watching it without watching people

The best way to deal with shrinkage is to make it visible early and routinely, before it becomes a confrontation. A short weekly note along the lines of "refunds at till 3 on Sunday closing shifts are three times the usual rate since September" lets a manager tighten a procedure quietly.

That is how Cuedis handles till activity. It learns the usual rate of voids, refunds and overrides for each till and person, and mentions it in the morning briefing only when a pattern stands out, with the records behind it. Who can see what is controlled by teams: you decide, for example, that cashiers never see colleagues' figures. Personal details such as phone numbers and IC numbers are not read at all unless you switch them on for that business.

Voids and refunds will always be part of retail. Knowing what is normal for each till and each person is what lets you spot the rare pattern that is not, early, quietly and fairly.