How to compare store performance fairly using customer reviews

Why store league tables mislead

Most attempts to compare store performance with reviews start with a table of stores sorted by Google rating. It feels objective, and it is usually wrong in four ways at once.

• Different sample sizes. A store with 25 reviews and a store with 2,500 both show one number. The first can move half a star in a week; the second barely moves in a year.
• Different formats and customers. A highway outlet, a mall flagship and a small-town franchise serve different people with different expectations.
• Different periods. An all-time average mixes this year's store with the one that existed three managers ago.
• No reason attached. The rating says a store is lower, not why, so the table creates arguments rather than actions.

The rest of this guide replaces the league table with a comparison that managers can trust and act on.

Step 1: compare stores with their peers

Group stores that customers would judge against similar expectations: same format, similar size, similar location type. For a national brand, add a regional or city-tier split if expectations differ. Comparisons happen inside a group; across groups you compare trends, not levels.

Step 2: use the same recent window

Use one period for every store, recent enough to describe today. Ninety days is a common choice for stores with steady review flow; quieter stores may need six months to collect enough reviews. Keep the window identical across the stores you compare, and keep the all-time figure for context only.

Watch for seasonality. Festive sales, wedding seasons and holiday peaks change both volume and complaints. Compare a store's festive month with last year's festive month, not with a quiet month.

Step 3: respect sample size

The fewer reviews behind a number, the less it tells you. A simple way to see this: if 4 of a store's 20 reviews in the window complain about billing, that is 20%. One more complaint makes it 24% (5 of 21); one more good review makes it 19% (4 of 21). The number is real but unstable.

• Set a minimum. Decide how many reviews a store needs in the window before it is ranked, and show stores below it separately.
• Show a range, not a point, for small samples. Statistical intervals (for proportions, the Wilson interval is a common choice) make the uncertainty visible: a store with a 20% complaint rate on 20 reviews has a much wider range than one with 20% on 500.
• Never drop small stores silently. List them as "not enough reviews" so nobody assumes they are fine.

Step 4: compare issue rates, not just ratings

The comparison that leads to action is by issue. For each store, work out what share of reviews mention each theme negatively: billing, waiting, staff behaviour, stock, cleanliness, returns. Then compare each store's rate with its peer group.

• A store with an average rating but double the peer rate of billing complaints has a specific, fixable problem.
• A store with a low rating spread evenly across many issues may need a broader management conversation.
• A store whose only outlier is parking may need nothing from the store team at all.

This needs every review classified into the same set of themes. Without a shared taxonomy, each store's problems are described in different words and cannot be compared.

Step 5: add direction

Compare each store's issue rates in the current window with the previous one. A store whose staff complaints doubled in three months needs a visit sooner than a store that has always been slightly below average and is improving. Present level and change side by side.

Presenting the comparison

• One row per store, grouped by peer group, with the review count in the window shown next to every rate.
• Highlight only what is outside the normal range for the group, not every number.
• Put the reason next to the number: the top issue and two or three representative reviews.
• Show stores below the minimum sample separately.
• Repeat the same view every month, so managers learn to read it.

The aim is not to name a best and worst store. It is to show each regional manager which issue to raise at which store, with the evidence in hand.

Common pitfalls

• Rewarding review counts. Volume reflects footfall and how hard a store asks for reviews, not quality.
• Comparing all-time ratings. Use the recent window.
• Ignoring the reviews themselves. A number without quotes invites the store to dispute it.
• Treating one viral review as a trend. Check whether it is repeated before acting.

A worked example

Take four mall stores in the same city over the last 90 days.

• Store 1: rating 4.5 on 820 reviews; billing complaints 4%, peer median 6%.
• Store 2: rating 4.6 on 31 reviews; billing complaints 13% (4 reviews).
• Store 3: rating 4.2 on 610 reviews; billing complaints 15%, up from 8%.
• Store 4: rating 4.3 on 540 reviews; billing complaints 7%, stable.

A rating league table puts Store 2 first and Store 3 last. The fair comparison says something different: Store 3 has a large, well-supported and rising billing problem and needs attention now; Store 2's figure rests on four reviews and goes on a watch list; Stores 1 and 4 are within the normal range. Same data, different decision.

Doing this with Falcon

Falcon classifies every public Google Maps review for each store into themes, sub-themes and issues with its sentiment. Experience signals show negative share by topic and month, findings built on small samples are flagged rather than ranked, and every number opens the reviews behind it, so a regional manager can compare stores on what customers actually said.