The Brand Health Score, fully disclosed: how PlaceOptimizer turns checks into one number

This post is a complete, no-secrets explanation of how the PlaceOptimizer Brand Health Score is calculated: the exact point weights per severity, the per-listing formula, the brand-level deductions, and one worked example computed step by step. If a scoring model is going to influence how you prioritise fixes across dozens of stores, you deserve to see the whole formula, not a marketing number. So here it is.

Before anything else, one statement we want impossible to miss: the Brand Health Score is a PlaceOptimizer diagnostic metric. It is not a Google metric, not a Google ranking factor, and Google does not endorse it or recognise it in any way. It is our internal way of summarising audit findings into something a marketing lead can track month over month. Google's actual local ranking is driven by relevance, distance, and prominence (Improve your local ranking on Google, accessed July 2026), and no third-party score changes that.

The two-level model

The score is built in two stages. First, each listing gets a per-listing score from 0 to 100. Second, those listing scores are averaged into a brand number, and brand-level deductions are applied for problems that only exist across the estate. Both stages are deterministic — the same inputs always produce the same score.

Per-listing scoring

Every listing starts at 100 and loses points for each issue the audit raises, weighted by that issue's severity:

The formula is:

listing score = 100 − (15 × critical count) − (7 × warning count) − (2 × info count)

The result is floored at 0, so a badly broken listing lands at 0 rather than going negative.

What counts as Critical, Warning, and Info

The severities come from the audit itself, applied to the fields Google exposes through its Places API (New) (accessed July 2026), such as rating, review count, opening hours, place types, and business status. Missing phone, missing address, no business hours, a rating below 3.5 stars, zero photos, and a permanently-closed flag are Critical. A missing website, hours set for fewer than seven days, a rating between 3.5 and 4.0, fewer than 10 reviews, and a thin photo gallery are Warning. A short description, a low-but-not-tiny review count, missing secondary categories, and website- or phone-format drift are Info. The full list, organised by the seven audit categories, is in our companion post on what the audit checks today.

One mechanic is worth repeating because it affects the arithmetic: the rating, review-count, and photo-count checks are severity ladders. Each ladder contributes at most one issue. A listing with a 3.2-star rating trips the Critical rating tier and nothing else from the rating ladder — it does not also get charged the Warning tier. This keeps a single weak signal from being double- or triple-counted.

A worked per-listing example

Consider one listing with a valid phone and address, but with these issues:

• Website missing → Warning (−7)
• Description only 20 characters → Info (−2)
• Hours set for 5 of 7 days → Warning (−7)
• Rating 3.8 stars → Warning, the 3.5-to-4.0 tier (−7)
• 32 reviews → Info, the 10-to-50 tier (−2)
• Unanswered-review risk (3.8 stars with only 32 reviews) → Info (−2)
• 6 photos → Info, the fewer-than-10 tier (−2)

That is 3 Warnings and 4 Info issues, with the primary category correct and the listing operational.

score = 100 − (7 × 3) − (2 × 4)
      = 100 − 21 − 8
      = 71

This listing scores 71. Notice that the 3.8-star rating produced exactly one rating-ladder issue (the Warning), not two, because ladders fire a single tier.

Brand-level scoring

Once every listing has a score, the brand number is built like this:

1. Average the listing scores and round to the nearest whole number. This is the brand base.
2. Apply brand-level deductions for estate-wide problems:

3. Clamp the result to the 0–100 range.

These deductions exist because some damage is invisible at the single-listing level. A store can score a healthy 85 on its own and still be one of four differently-named pins for the same brand — a fragmentation problem you only see when you compare listings. The duplicate-candidate deduction is driven by the proximity detector described in our audit post; because those are candidates and not confirmed duplicates, the penalty is a modest −5, a nudge to investigate rather than a verdict.

A worked brand example

Take a ten-location brand whose individual listing scores are:

82, 78, 88, 71, 64, 90, 76, 83, 69, 79

Their sum is 780, so the average is 780 ÷ 10 = 78.0, which rounds to 78. That is the brand base.

Now suppose the audit found two estate-wide problems: the business name appears in several drifted variations across the ten pins (name fragmentation), and the proximity detector surfaced one duplicate-candidate cluster. Categories were consistent and the estate's average rating was healthy, so those two deductions do not apply.

brand score = 78 − 10 (name fragmentation) − 5 (duplicate candidates)
            = 63

The brand scores 63. Two clearly-motivated deductions took a middling-but-not-terrible base down into "needs attention" territory, which is exactly the signal a marketing lead wants: the individual stores are mostly fine, but the estate has structural consistency problems worth a cleanup project.

Why these particular weights

The weights encode a priority order, not a scientific constant. A Critical issue costs roughly twice a Warning and more than seven times an Info because a missing phone number genuinely hurts a customer more than a thin description does. The low-average-rating deduction exists because reviews are a genuine prominence signal: Google states that review count and score factor into local ranking (Improve your local ranking on Google, accessed July 2026), which is why an estate skewed toward sub-3.5-star pins loses brand points. The brand deductions are heavier for problems that compound across an estate (name fragmentation at −10) and lighter for problems that are candidates awaiting confirmation (duplicates at −5). We tuned these to make the ranked issue list actionable, and we may adjust them as we learn from more estates — which is another reason to treat the number as a directional diagnostic rather than an absolute grade. If you want the mechanics behind how much review count really moves rankings in Indian metros, we cover it in do Google reviews move local rankings in India.

Limitations, stated plainly

A single number always loses information, and this one is no exception. A few things to keep in mind:

• Small estates behave differently. The name-fragmentation deduction compares the count of distinct names against a fraction of the total location count. On a very small estate this threshold is sensitive, so a two- or three-store brand can trigger the deduction more readily than a large one. Read the underlying consistency finding, not just the score, when your estate is small.
• The score reflects automated checks only. It does not know whether your reviews are answered, whether your photos are fresh, or whether a flagged category is actually wrong for a non-furniture vertical. Those are the "manual verification recommended" items from the audit, and none of them move the score today.
• Duplicate candidates are candidates. The −5 duplicate deduction fires on proximity, not on a confirmed duplicate. A human should verify before any merge.
• It is not comparable across brands as a league table. Two brands with the same score can have very different underlying problems. Use the score to track one brand over time, not to rank unrelated brands against each other.

And, once more because it matters: this is a PlaceOptimizer diagnostic. Google neither produces nor endorses it, and improving it is not the same as improving your Google ranking, though the two tend to move together because both reward complete, consistent, well-maintained listings.

Frequently asked questions

Is the Brand Health Score a Google ranking predictor?

No. It is a PlaceOptimizer diagnostic that summarises our own audit findings into one number. Google does not produce, recognise, or endorse it, and it does not predict where you rank in the local pack. Google's actual ranking runs on relevance, distance, and prominence. Use our score to track one brand's listing quality over time, not to forecast rankings.

Why did my score drop when nothing changed on my listings?

The most common cause is a brand-level deduction firing on a comparison, not on a single listing. If a new pin was discovered with a drifted name, or the proximity detector surfaced a duplicate candidate, the estate-wide checks trigger even though each individual listing looks unchanged. Read the underlying consistency and duplicate findings, not just the headline number.

Can two brands with the same score be compared?

Not as a league table. Two brands can both score 63 for entirely different reasons, one from low ratings, another from name fragmentation. The score compresses a ranked issue list into a single figure, so always read the issues beneath it. It is built to track a single brand month over month, not to rank unrelated brands.

Does a perfect 100 mean my listings are flawless?

No. A 100 means your public data passes every automated check. It does not certify that your reviews are answered, your photos are fresh, or your holiday hours are correct, because those are manual verification items the score does not measure. See what the audit checks today for the full automated-versus-manual split, and how the audit runs for the end-to-end flow.

If you want to see the checks that feed this score, read what the audit checks today or run the free audit on your own estate. To verify findings by hand, the GBP audit checklist walks the same seven categories in order. If your audit came back thin for a city, why some city audits return no data explains why the score sometimes has nothing to compute.