Retail Operations Dashboard: How to Compare Stores Fairly

You run twelve stores and the chain's sales figure is on your desk every morning. It tells you how the chain did. It cannot tell you that one store has been losing margin for a fortnight, or that your biggest store is the weakest of its format. A retail operations dashboard exists to answer those questions, one store at a time.

Stores differ more than most reports let on. Across US retail, a store at the 75th percentile of labour productivity is about 150% more productive than one at the 25th, a gap that has barely moved in decades¹. Inside one chain the rules are the same, yet in a study of two large retail companies, store managers alone explained 25% to 35% of the variation in store productivity².

What a store dashboard is for

Its job is narrow: show which stores need attention today, and give a first clue why. Anything beyond that is a report. Management surveys call this monitoring: tracking a few performance indicators, reviewing them often and acting on what they show. In the UK, the Office for National Statistics finds that firms in the top quarter of management scores have 43.3% higher turnover per worker than those in the bottom quarter³, and the use of KPIs is the practice British firms score lowest on⁴.

Three people read the dashboard differently. The owner looks for the stores that moved the chain's numbers. The operations manager looks for the store to call. A store manager looks at their own store against its usual week. One view serves all three if every number is per store and every number has a comparison next to it.

The five numbers to show for every store

• Sales, against the same weekday last week, so a Monday is compared with a Monday.
• Gross margin as a percentage, so a store that sells more by discounting shows up.
• Transactions, the number of receipts, which shows whether fewer people bought.
• Average ticket, sales divided by transactions, which shows whether they spent less.
• Distance from the store's own normal, with a flag when it leaves the usual range.

Items per basket, stock at cost and shelf availability are worth adding when your POS and ERP record them per store. Without its comparison, any of these is only a number.

Why the chain average hides the problem

An average sits in the middle of the stores by construction, so it always looks reasonable. In the chain below it looks healthy. The store at the bottom has been there for weeks, and nobody has asked the one at the top how it does it.

The low store needs a visit. The high one may be doing something the others should copy, or may be booking its costs late. The average says neither.

Compare each store with stores like it

Ranking every store on raw sales mostly ranks their size: the mall store wins, the kiosk loses, and nobody learns anything. A fairer comparison takes three steps.

1. Group stores by format and size: mall, high street, neighbourhood, kiosk.
2. Compare each store with the median of its group, so one strong store does not set the bar for the rest.
3. Keep it like-for-like. Leave out stores that were not open for the whole of both periods, and mark the ones that were refitted or moved. The like-for-like guide covers where to draw the line.

Against its own format, the biggest store falls below the middle and the smallest kiosk rises near the top. Those are this week's two conversations: what the mall store is missing, and what the kiosk is doing right.

Choose the baseline before you read the number

A number means something only next to the right comparison, and each choice misleads in its own way:

• Same weekday last week. The simplest fair comparison for daily sales, because shopping follows a weekly pattern. Forecasters call it the seasonal naive method: the forecast for a day is the last value from the same season⁵. It breaks on holidays and paydays.
• The median of the same weekday over the last four weeks. Steadier, and one odd day does not move it.
• Same period last year. Useful across seasons, but calendars shift, and a store refitted since then is a different store.
• Plan or budget. Shows whether the target was met, not whether the store is normal; targets are often last year plus a percentage.
• The peer median. The quickest way to tell a weak store from a weak week for the whole format.

Pick one baseline per number and keep it. A dashboard that switches baselines between columns makes every store look odd somewhere.

A bad week, or a real decline?

Every store has noisy weeks. Statistical process control, the method factories use to watch a production line, offers two simple rules. A point outside limits set three standard deviations from the centre line calls for a look at its cause⁶. A run of consecutive points on one side of the centre line is a shift even when none crosses a limit; the classic Western Electric rule uses eight⁷.

The single bad week usually has a cause the store manager already knows: a closed day, a broken till, a delivery that never came. The run of weak weeks is the one that goes unnoticed, because no single week looked alarming.

From flag to action

Read by exception, the dashboard takes minutes, because most stores need nothing today. The routine that works is short:

• Every morning: the stores outside their normal range, each with the number that moved.
• Every week: each store against its peer group, and any run of weeks on one side of normal.
• Every month: each format as a whole against last year, like-for-like.

When a store's sales drop, look at how its staff hours are spread across the day before reaching for a discount. In 191 apparel stores of one chain, conversion and sales fell as staff worked through more customer encounters in a day. On the researchers' model, rearranging the existing schedule could lift a store's sales by 2% without adding staff⁸.

How marql builds this view

marql reads sales and stock from a chain's POS and ERP, syncs through the day and runs a full pass each night. Its Stores view puts every store's sales, change, gross margin, average ticket and share of the chain side by side, against the previous period or the same period last year. Each store's margin is set against the median of stores of the same format, and a gap of more than two points is marked in the row.

Anomalies are judged against each store's own weekly pattern over the last eight weeks. A store is flagged only when it sits about three standard deviations from its normal level and the change is material: 15% of sales, or 2.5 points of margin. Period reviews are like-for-like, using only stores present in both periods, and the morning brief names the stores that moved the chain's numbers most. See the analytics view, and the Impact Ledger for how an action on one store is measured against comparable stores. If you are weighing a BI tool instead, the comparisons set out when each option is the better fit.