Store health scores: what a single number can and cannot tell you

In a chain with two hundred stores, the question a regional manager asks every morning is this: "which store should I look at today?" Going through reports one by one takes time; a health score exists to answer that question in seconds.

But a single number loses information by nature. What the score says has to be designed as carefully as what it doesn't.

What does a score solve?

The score's real function isn't ranking but directing attention. A regional manager can't give equal time to fifty stores; the score helps them choose where to go.

That distinction matters, because the idea of ranking quickly turns into a competition — and competition distorts measurement.

Which indicators belong in it?

The choice of inputs decides the score's fate. A set that works includes one indicator from each of several dimensions:

• Operations — on-time completion rate for field tasks.
• Audit — the last audit score and recurring findings.
• People — staff turnover and time-to-fill for open roles.
• Communication — read and acknowledgement rates for critical announcements.

Whether commercial performance (revenue, basket size) belongs in the score is debatable. Include it and the score turns into a sales report while operational signals disappear; exclude it and managers don't treat the score as "the real thing". The practical answer is to show commercial data alongside the score rather than inside it.

Weighting and transparency

How the score is calculated should be as open as the score itself. A store manager who sees the number must be able to see how much each indicator contributed; otherwise they don't know what to improve and the score becomes a verdict.

Keeping the weights simple also helps. Giving four indicators equal weight is usually more defensible than a complex formula, and it ends the argument.

Anomalies matter more than absolute values

A store scoring 72 tells you little on its own. The real signal is change: a store that has been at 85 for three months dropping to 72 is a clear reason to visit.

That's why the most valuable element of a score screen isn't the ranking table but the decline alerts. A store that is always low is already known; a store that has just fallen is new and open to intervention.

Don't turn the score into a penalty

This is the most common and most destructive mistake. The moment the score is tied to bonuses and performance reviews, stores start improving the score rather than the work.

The consequences are predictable: temporary arrangements before an audit, tasks marked complete without being done, problems going unreported. The score rises, the real situation worsens, and management goes blind.

We covered the same trap on the goals side — a similar mechanism is described in our OKR article.

Comparing with self-assessment

The gap between a store's own checklist and the regional manager's audit is an indicator in itself. A store that systematically rates itself higher may be misunderstanding the standard — that calls for training, not blame.

Tracking that gap is the most practical way to measure whether the field and head office share the same language.

Visibility: who should see what?

A store manager should see their own score and its components — otherwise they can't improve. A regional manager sees their whole region. Making the ranking of the entire chain public is usually a bad idea; it demotivates the stores at the bottom.

Visibility restrictions must be enforced in the data layer; hiding a score in the interface doesn't protect the data.

The data-quality trap

A score can be no better than the data feeding it. If field tasks aren't closed consistently and audits happen irregularly, the score is noise.

That's why the score module should be switched on after field tasks and audit processes have settled. A score introduced on day one formalises a false picture and never regains credibility.

Measurement

Three questions for evaluating the score itself:

• Predictive power — do stores whose score falls actually turn out to have a problem?
• Action rate — how many decline alerts turned into a visit or an intervention?
• Data completeness — in how many stores are the score's inputs complete?

The second is decisive: a score that raises alerts nobody looks at is just screen decoration. We cover sending work to the field in our field tasks article and managing store data in our store directory article.