Restaurant Review Metrics Without Misleading Averages

A useful restaurant review report keeps each platform's rating scale, feedback population, and time window visible. Compare recent feedback within the same source first. Bring common complaint themes together only after preserving those differences.

That approach helps a multi-unit team decide where to investigate without turning unlike signals into a misleading brand-wide score. It also separates three questions that need different evidence: what guests are saying, how the team is responding, and what happened across completed orders.

Keep each platform's rating in its original form

DoorDash's current merchant guidance describes Loved, Liked, and Didn't Like feedback. Its customer-facing Lifetime Rating remains on a 1-to-5 scale, combines historical stars with newer emoji ratings, and uses an undisclosed calculation. The selected-period feedback distribution and that lifetime score answer different questions. DoorDash merchant guidance

Google describes its review score as the average of ratings published on Google for that business. It also says an updated score can take up to two weeks to appear. Google Business Profile guidance

Keep these values as separate fields. A restaurant's Google average and DoorDash Lifetime Rating should not become one blended average simply because both are displayed out of five. Their calculation and feedback populations differ.

For US and Canadian teams, confirm the rating experience shown for each store rather than assuming an identical rollout. UK and European teams can apply the same measurement discipline to their own available review sources; the DoorDash example does not imply availability in those markets.

Write down exactly what is being counted

Before a regional review meeting, put these details beside every measure:

• Source and location: the platform and the exact restaurant
• Native value: the original category or rating, without a made-up conversion
• Population: all ratings, written reviews, public reviews, private feedback, or eligible replies
• Time basis: feedback received during the period, order date, or lifetime display
• Coverage: what the source actually supplied, when it was refreshed, and what is missing

DoorDash distinguishes public reviews from private feedback and offers a written-comments filter. Applying that filter changes the population being examined. A report based on written comments should say so. DoorDash feedback filters

Avoid counting the same feedback twice when one record appears in two views. Preserve a source record identifier when available. Keep private feedback and guest details inside the authorized operating workflow.

Build a small scorecard around the decision

Use three separate views rather than forcing every signal into a single score.

Recent guest feedback

Within each platform, show the native rating distribution, total ratings received, and selected dates. A negative-rating share needs its numerator and denominator beside it.

If an internal report groups categories into a negative-feedback bucket, document the mapping for each source. That bucket is an operating convention, not proof that a Didn't Like response is equivalent to a particular number of stars.

Keep lifetime or customer-facing display scores nearby as context. Do not treat their daily movement as a direct measure of this week's operating changes.

Team response work

Show how many eligible feedback records received a reply, how many remain open, and the age of the oldest eligible unanswered item. Define eligibility before calculating a response rate.

For example, a platform restriction, expired response window, or unavailable reply control can affect the queue. Mark those records separately from items the team could answer but has not. DoorDash currently describes private merchant replies and a seven-day response window, so a blanket claim that every reply is public would be inaccurate. DoorDash response guidance

Operational follow-through

Group written feedback into practical themes such as missing items, packaging, or pickup experience. Show the number of reviewed comments behind each theme and allow more than one theme per comment.

Use the wording to choose what to inspect. A complaint is evidence of a guest's reported experience; it does not by itself establish which part of preparation or delivery caused the problem.

For a broader store comparison, use the same source and period rules in your restaurant benchmarking process.

A worked example with visible denominators

Consider two hypothetical stores on the same platform during the same complete week. These are invented teaching numbers, not customer results.

• Store A receives 100 ratings, including 12 negative responses: 12 divided by 100 equals 12 percent
• Store B receives 20 ratings, including 4 negative responses: 4 divided by 20 equals 20 percent
• Store A has more negative responses in total; Store B has the higher observed share among people who rated

Both stores deserve a look, for different reasons. Store A has a larger volume of reported concerns. Store B's smaller sample can move sharply: one additional negative response, with everything else unchanged, produces 5 out of 21, or about 23.8 percent.

Now suppose Store A completed 1,000 orders. Its 12 negative responses represent 1.2 percent of completed orders only if those responses can be matched to distinct orders in that same order cohort. That figure still measures recorded negative feedback per order. It does not establish that only 1.2 percent of all guests had a problem.

A feedback-received date window and an order-date window may contain different people. If you cannot reconcile those populations, leave the per-order measure unavailable rather than calculating a precise-looking number from mismatched periods.

Use the report to choose one operating check

Start with one question: where should the team investigate next?

1. Check source coverage and whether the period is complete
2. Compare each store with its own prior equivalent period on the same platform
3. Read a small set of the underlying comments before naming a cause
4. Assign one observable check, such as reviewing the pack-out process for a recurring missing-item theme
5. Record the action date and review the same measure again after enough comparable feedback arrives

Show low-volume results with their counts. Do not create a universal minimum sample or automatic escalation threshold without considering severity and the cost of a missed issue. Urgent concerns should follow the restaurant's established escalation process regardless of the weekly average.

Where Voosh fits

Voosh's Reviews and Reputation Automation brings supported review sources into a unified inbox and supports store-level filtering, sentiment, response-time and volume tracking, plus AI drafts, manual replies, and templates.

Use that shared view to organize the work. Confirm the source fields, date filters, and channel-specific actions available for your connected accounts before using a custom scorecard. This article does not claim that Voosh already supplies every calculation described above.

Book a Voosh demo to review your current feedback workflow and the store-level views available to your team.