PRACTICAL RESTAURANT REFERENCE

Restaurant customer feedback and repeat business: turn complaints into specific improvements

Classify restaurant feedback, respond constructively and measure a service improvement using clearly defined customer and order records.

By Anurag · Reviewed 17 October 2026

Customer feedback becomes useful when it leads to a specific question and an observable change. “The delivery was disappointing” is important, but it does not tell a manager whether to change a recipe, improve packing, clarify the menu or investigate handover. The first job is to preserve what the customer actually said, then connect it carefully with the restaurant's own evidence.

Repeat business is another useful signal, but it is not the same thing as a favorable review. A customer can enjoy a meal and have no immediate reason to order again. Another can return frequently while reporting a recurring problem. A practical review process keeps service feedback, operational outcomes and repeat-order measures connected without pretending that any one of them explains the whole relationship.

This guide describes a manageable learning cycle for a small restaurant, café, takeaway or delivery kitchen. All examples, customers, counts and prices are fictional. No repeat-order percentage is offered as a benchmark, and no response template guarantees a better rating or another sale. Platform features and policies vary by account and market; handle reviews and customer remedies through the current permitted channels.

Understand what feedback can and cannot show

A review records an experience as the customer describes it. It can reveal a problem the restaurant did not observe, such as a leaking container or unclear reheating information where such information was supplied. It can also be incomplete, mistaken or unrelated to the step the team first suspects. Treat it as evidence to investigate, not a complete technical diagnosis.

Separate three layers: the customer's statement, the restaurant's finding, and the action taken. “Meal arrived without the side” belongs in the first layer. “The side bag remained in the collection area” belongs in the second if verified. “Add a visible split-bag handover check” belongs in the third. Keeping these layers apart makes the reasoning understandable.

A single detailed complaint may warrant immediate action when the consequence is serious. A small set of ordinary complaints may need more observation before a pattern becomes clear. Avoid both extremes: dismissing feedback because only one person mentioned it, or redesigning the entire menu after one ambiguous comment. Severity, evidence and recurrence all matter.

Positive feedback also needs interpretation. A customer praising generous portions may be responding to an unintended serving variation. A popular item may still generate a high number of issues because it sells frequently. Look for the relevant operational context before deciding what the comment means for the business.

Map the places where feedback already arrives

List the channels the business actually uses: counter conversations, telephone calls, email, platform feedback, public reviews and any voluntary survey. Identify who sees each channel and how urgent messages reach the shift lead. A complaint is easy to miss when everyone assumes someone else monitors an inbox or merchant device.

Keep the original channel and visibility in the record. A private platform comment is different from a public review, and the permitted response may differ. Do not assume that because one service offers a public reply, another does too. DoorDash's current merchant help distinguishes private feedback from public store reviews and describes responses through its own account tools.

Include verbal feedback through a short factual note when useful. Staff should not need a lengthy form to record that several customers found a pickup sign confusing. Note the date, location, issue and whether the wording is a direct recollection or a summary. Do not invent quotation marks around a paraphrase.

Check for duplicate reporting. The same customer may contact support and later leave a public review about the same order. Both records can matter, but counting them as two affected orders inflates the incident measure. Link known duplicates using authorized references, while preserving the separate communication records and their visibility.

Record enough detail to make an improvement possible

A compact feedback record can include the date received, outlet, channel, relevant order date where known, menu item, issue category, brief description, urgency, owner and status. Add the operational finding and action when available. Keep an internal case reference so follow-up messages can be connected without repeatedly copying the customer's details.

Preserve the difference between missing information and a negative finding. An unknown order reference does not mean that no order existed. No recorded handover issue does not prove handover was correct. Use explicit values such as unknown, not checked or not applicable instead of leaving a blank that later readers may interpret differently.

Keep source wording available through the approved system, but avoid spreading personal information into broad team reports. A weekly operational summary usually needs the issue and action, not a customer's full name, address or contact details. Summaries can be useful while using less sensitive information than the original communication.

Make maintenance practical. A record that requires twenty fields before someone can log a simple problem may discourage reporting. Start with the minimum needed to route and investigate the case, then add findings as work proceeds. The goal is a dependable learning trail, not the appearance of a comprehensive database that staff cannot keep current.

Use categories that point toward action

Choose categories based on the restaurant's actual operating steps. Order completeness, preparation, packaging, menu description, portion consistency, collection experience and payment questions can be useful starting points. Keep a brief definition for each. “Food quality” alone may be too broad if it combines taste preference, temperature complaints and a suspected safety problem.

Allow more than one issue tag when the customer describes distinct problems, while counting the order only once in an affected-order measure. A leaking drink and a missing side can belong to the same order. Item-level issue counts and unique affected-order counts answer different questions and should not be mixed in the same total.

Review an “other” category periodically. Repeated notes may reveal a topic worth separating, such as confusing collection instructions. Conversely, two categories may be so similar that staff use them inconsistently. Refine the definitions with the people logging cases, then record when the classification changed so historical comparisons remain understandable.

Keep severity separate from topic. Packaging feedback can range from an inconvenient torn outer bag to a serious concern needing immediate escalation. A category should help route the issue, not decide its seriousness automatically. Give staff a clear path for concerns that fall outside ordinary service recovery.

Escalate serious concerns through the proper process

Suspected allergen errors, contamination, illness, threats or other serious concerns should follow the business's established escalation and local requirements. A feedback spreadsheet is not a substitute for that process. Staff need to know whom to contact immediately and what factual information to preserve, rather than waiting for the weekly review meeting.

Do not diagnose the customer's condition or dismiss a safety concern because other customers have not complained. A public response is rarely the appropriate place to investigate detailed personal or medical information. Use the authorized private route and the responsible food-safety or management process. Get appropriate official or professional advice where the situation requires it.

Separate immediate protection from later root-cause work. If an issue could affect other orders, the relevant team must assess the current product or process promptly under its validated controls. Offering a future discount does not address that immediate question. Preserve relevant records without altering them to support a preferred explanation.

Once the incident is handled, identify any learning appropriate for routine operations. That may involve menu information, supplier details, staff training or escalation clarity. Share only what the team needs to know. A serious case should not become an identifiable anecdote circulated widely as a training example without proper consideration.

Respond to the specific concern

A useful response acknowledges what the customer reported, explains the next appropriate step and avoids promises outside the restaurant's control. Keep it short enough to be read. A generic paragraph about always providing excellent service can feel unresponsive when the customer has described a particular missing item or confusing collection experience.

For an ordinary missing-component complaint, a permitted private reply might say that the team is checking the order and ask the customer to use the relevant support route for the remedy available through that channel. Adapt the wording to the facts. Do not claim to have refunded an amount unless the authorized process confirms that action.

When replying publicly through a service that permits it, avoid exposing order references, addresses, staff schedules or private messages. Acknowledge the concern and direct detailed follow-up appropriately. The objective is to show a constructive response, not to win an argument in front of other readers.

Have someone review replies involving uncertainty, significant allegations or unusual remedies. Staff can use approved templates for structure, but should check that the wording matches the actual case. A copied apology for cold food is not helpful when the complaint concerns a menu description, and an automated-sounding promise can create a new expectation the business cannot meet.

Keep service recovery independent of review behavior

A restaurant may decide that a remedy is appropriate under the relevant rules, but should not make it conditional on a positive review or removal of criticism. The customer response and the review policy are different matters. State the remedy through the proper channel, with its terms, without asking the customer to change their public account of the experience in exchange.

Consider the economics of discretionary offers without treating them as automatic. An offer can reduce contribution and may not result in another order. It can still be an appropriate service decision, but should be authorized and recorded. Do not let every employee invent a different compensation promise during a busy shift.

Distinguish an offer issued from an offer redeemed. An issued code may never be used, may have restrictions, or may be funded differently under a platform arrangement. Use actual account information when analyzing the financial effect. Avoid treating all issued face value as either confirmed expense or confirmed future revenue without the relevant basis.

Review whether recovery solved the original operational problem. If the same container keeps leaking, sending repeated offers leaves the cause intact. A good response supports the affected customer while the operating team investigates why the failure happened. Those two responsibilities need named owners and a way to see when both are complete.

Request genuine feedback without screening for praise

Follow the rules of the service where a review would be posted. Google's policy prohibits incentivized reviews and selectively soliciting positive reviews, while allowing requests for genuine unincentivized experiences. Do not route only satisfied survey respondents to the public review link while sending dissatisfied respondents elsewhere to keep their criticism private.

A neutral request should allow the customer to describe their actual experience. Avoid suggested wording that tells them what rating or praise to provide. Do not ask staff, family or other people without a genuine relevant experience to manufacture reviews. A higher visible rating obtained through manipulation is not useful evidence about the restaurant's service.

Keep private feedback channels available for all customers as an additional route, not as a barrier placed in front of negative public feedback. A person may prefer to discuss an issue privately, but that should be their choice. Never threaten, shame or pressure a customer because their experience differs from the restaurant's view.

Check new marketing tactics against current platform policy before adopting them. A merchant marketing article may suggest a tactic that does not fit another service's review rules. Treat the policy owner as the authority for its own platform. The restaurant's goal is honest feedback and improved service, rather than an optimized pathway that suppresses unwelcome information.

Protect customer information while learning from it

Use the minimum information needed for the decision. An operational discussion about lids generally needs the product, date, packaging type and reported issue, not the customer's identity. Keep personal details within the authorized case-handling system and share summaries with the people responsible for improvement.

A public browser-local worksheet does not need customer names, addresses or merchant credentials. Manual counts and non-identifying categories can support most examples in this guide. Do not assume that a platform export authorizes every possible reuse, cross-channel match or external upload. Follow the account's terms and the business's applicable privacy obligations.

If repeat-order analysis uses customer-level data, it needs an authorized stable identifier and an appropriate purpose. A first name, similar address or matching order preference is not a reliable identity link. Avoid trying to identify the same person across separate platforms through guesswork. Where the necessary authorized data is unavailable, use the platform's own aggregate reporting or state the limitation.

Decide how records will be accessed, corrected and retained through the business's approved process. This guide does not supply a universal legal retention period. A small, well-governed record is more useful than an uncontrolled archive of screenshots that nobody can search safely or explain.

Look for patterns without inventing a cause

Group feedback by issue, item, outlet, channel and relevant service period. A pattern may emerge around one container, one bundle or a particular handover point. Compare the comments with operational evidence such as preparation changes, packaging deliveries and order volume. The timing can suggest a hypothesis, but does not establish it on its own.

Consider exposure. A popular dish may attract more complaints in total while having a lower issue rate per eligible order than a rarely ordered dish. Conversely, a low complaint count can hide a serious problem when sales are also low. Use the corresponding order population when it is available, and state when it is not.

Read representative comments rather than relying only on a category total. “Cold” might describe a cold side dish the customer expected to be hot, a genuine temperature concern, or a delayed journey. Those possibilities require different investigation. Keep the original meaning visible while reviewing the operational facts.

Allow disconfirming evidence. If leaks occur across several packaging batches and different dishes, the explanation may not be one faulty supplier lot. If a problem appears only after a ticket-layout change, inspect the information flow. A useful review is willing to revise its first explanation rather than gathering only examples that support it.

Choose a manageable improvement

Prioritize using consequence, recurrence, evidence and the practicality of a change. Serious concerns require their own immediate response. For routine service issues, choose one problem the team can describe clearly and influence directly. “Improve loyalty” is too broad to assign to a packer; “make the separate dessert visible at handover” identifies a specific operating action.

Estimate the effort as well as the possible benefit. A ticket configuration fix may remove repeated confusion more effectively than a long retraining program. A packaging trial may need purchasing and kitchen approval. Name the people whose work is affected before changing the routine, so the proposed improvement fits the actual operation.

Avoid selecting only the loudest complaint. A strongly worded review can draw attention, but quieter repeated issues may affect more orders. Equally, a rare but severe problem should not be dismissed because it contributes little to a numerical total. A short written explanation of the priority helps the team understand the decision.

Define what remains outside the current change. Improving handover identification will not necessarily solve food texture after a long journey. A narrow scope makes evaluation more honest and prevents one intervention from being credited for every favorable movement in the business.

Turn the complaint into a testable hypothesis

A useful hypothesis connects an observed issue with a plausible process step. For example: “Separate chilled desserts are sometimes missed because the final handover position shows only the hot bag.” The hypothesis explains what evidence to inspect and suggests a change. It is more useful than “staff need to care more,” which is difficult to test and often unsupported.

Write the proposed change and expected observation before implementation. A visible split-order marker should make the incomplete state clearer. The relevant observation is whether staff use it and whether defined omissions decrease among eligible split orders. Overall restaurant sales would be a weak primary measure for this particular test.

Record the start date, affected items, responsible person and any other simultaneous changes. If a new menu, new packaging and extra staffing all begin together, a later improvement cannot easily be attributed to the marker alone. Sometimes several changes are necessary, but the evaluation should acknowledge that limitation.

Include a check for unintended effects. A label that delays handover, covers important information or detaches during handling can introduce a new problem. Ask the team to report these observations. A trial succeeds when it improves the actual process, not merely when its main metric looks favorable.

Use matching denominators when measuring results

Define the numerator and denominator together. In a fictional period, twenty-four submitted reviews among four hundred completed orders represent a six percent review submission share. That is a description of feedback participation, not proof that the remaining customers were satisfied or dissatisfied.

Suppose six of those reviews report a problem, each concerning a different eligible order. Six divided by twenty-four gives twenty-five percent of submitted reviews. Six divided by four hundred gives one point five percent of completed orders with an observed complaint in this simplified example. These percentages use different populations and answer different questions.

Do not label the review share as an order-defect rate or the observed complaint rate as the true rate of all customer problems. Customers choose whether to report, channels capture different feedback, and some issues remain unknown. The measure should describe the evidence actually available rather than imply complete observation.

Keep definitions stable across the comparison. If the first period includes telephone complaints and the second includes only platform reviews, the change may reflect collection method. Record additions to channels and category rules. Where definitions change materially, start a new series or explain the break rather than drawing a smooth improvement line through incompatible data.

Compare observation periods fairly

Allow enough time for the relevant feedback to arrive. An order placed yesterday may not yet have a review or resolved support outcome. Mark the latest period provisional when reporting is incomplete. A mature earlier period will otherwise appear worse simply because its customers had more time to respond.

Use comparable service periods where practical. Weekend dinner, weekday lunch and a holiday event can have different order mixes and customer expectations. Review volume, channel, items and staffing context alongside the issue measure. A favorable before-and-after result remains observational unless the design supports a stronger conclusion.

Show counts with rates. Moving from two incidents to one can produce a large relative percentage change while involving very little evidence. A restaurant does not need complex statistics for every operational decision, but should recognize uncertainty. Avoid declaring a permanent solution after a short quiet run when the original problem was rare.

Combine the numbers with direct observation. If staff never used the new process, an improved rate is weak evidence that the process helped. If usage was consistent but a particular exception still failed, refine the design around that exception. Measurement should support learning about the work, not replace watching how the work is performed.

Define repeat business explicitly

“Repeat customer rate” can describe several different calculations. It might mean the share of this month's customers who had ordered previously, the share of a first-order cohort returning within a defined window, or the share of orders placed by returning customers. These measures have different denominators and can move differently even when based on the same transactions.

Choose the definition that fits the question. If the business wants to understand whether newly acquired customers return, a first-order cohort with a specified follow-up window is useful. If it wants to understand today's sales mix, an order share from previously observed customers may be more relevant. Name the measure in plain language.

Define qualifying orders and identity scope. Decide how cancellations, full refunds, test transactions and duplicate records are treated. State whether the analysis concerns one outlet, one channel or an authorized combined dataset. A customer returning through another channel may be invisible to a single-platform record, and that limitation should remain visible.

Do not call absence from a limited dataset proof that a person never returned. They may have dined in, ordered through another service or purchased without an identifiable account. The metric describes observed behavior within the defined records. It is not a complete account of the customer's relationship with the restaurant.

Work through a fictional return cohort

For one original illustration, define the cohort as first-time observed customers whose first qualifying order occurred in a specified entry period. Include only customers whose full thirty-day follow-up window has elapsed. The measure is the share of that eligible cohort with another qualifying completed order within thirty days of the first.

If one hundred twenty customers are eligible and thirty-six make a qualifying return order, divide thirty-six by one hundred twenty and multiply by one hundred. The result is thirty percent. Count each customer once in the numerator, even if one customer places several return orders. A separate order-frequency measure can describe that additional activity.

This thirty percent is a fictional result under a chosen definition. It is not a universal retention formula or a target for other restaurants. A business serving occasional celebrations may have a different natural purchase rhythm from a nearby daily lunch café. Comparing them without context can produce an unfair conclusion.

Keep the cohort's acquisition context. A first order driven by an unusual promotion may attract a different mix from ordinary trade. If the next cohort changes, inspect the offer, season, channel and customer occasions before attributing the difference to a service response. A return calculation needs a business explanation, not just a percentage.

Handle incomplete follow-up windows honestly

A customer whose first order happened five days ago has not had thirty days to return. Including that person as a non-returner in a thirty-day measure depresses the rate artificially. Either wait until the full window has elapsed or use another clearly defined analytical approach. Do not mix incomplete and complete windows without explanation.

Set the observation cutoff before calculating. For each customer, compare the first-order date with that cutoff and include only fully observed windows under the chosen method. Timezone and day-boundary rules should be consistent. A report extracted at the start of a day may not contain all events someone assumes belong to that day.

Preserve the distinction between “not yet observable” and “observed without a return.” Both can look like an empty return-date cell, but their meanings differ. A simple eligibility field makes the denominator easier to audit and prevents a spreadsheet from silently treating every blank as a failure.

When presenting recent performance, show mature cohorts separately from provisional ones. This allows the team to see current volume without claiming that the newest customers have completed a follow-up period. A transparent lag is more credible than a constantly changing rate whose denominator nobody can explain.

Respect the limits of customer identity data

Repeat analysis requires an authorized stable way to recognize the same customer within the intended scope. If the export does not provide that information, do not manufacture it by matching names, partial addresses or meal preferences. Similar names can belong to different people, and the same person can use different details.

Use native aggregate reports where they provide the needed insight without exposing individual records. Understand the provider's definition and date range. A platform's “new customer” may mean new to the store on that platform, not someone who has never visited the restaurant. Keep that wording when sharing the result with the team.

A local manual workbook should not become a hidden cross-channel customer-tracking system. Enter only the non-sensitive counts or authorized pseudonymous references needed for the task. Follow applicable rules and account terms for any customer-level analysis. This guide does not establish a legal basis or permission for a new marketing use.

If reliable identity is unavailable, focus on measures the business can support: repeat visits voluntarily identified through an appropriate program, native channel summaries, issue recurrence, menu consistency and customer comments about their experience. “Not measurable from these records” is an honest conclusion and more useful than a precise but unreliable retention figure.

Interpret channel and menu mix before comparing

Customers may use different channels for different occasions. Someone discovering a restaurant through delivery may later collect directly or dine in. A fall in observed repeat orders on one channel can coexist with a stronger overall relationship that the dataset cannot see. Avoid turning a single-channel metric into a claim about every customer.

Menu mix can affect feedback. A new item with unfamiliar ingredients may generate more questions than an established staple. A promotion can increase the share of complex bundles and expose a packing weakness. Compare relevant item and order groups before assuming a change reflects a general decline in service.

Outlet comparisons need consistent bases too. A newly opened location has fewer previously observed customers and may naturally have a different returning-customer share. A mature outlet near offices serves a different pattern from one near event venues. Use shared definitions and explain structural differences before ranking teams.

When a mix change explains part of a movement, show it rather than adjusting the result invisibly. The team can then decide whether the new mix requires a different process. A delivery-heavy menu may need more packaging attention, while a collection-heavy location may need clearer waiting and handover information.

Cost a recovery or return offer without promising loyalty

A future offer can support a service decision, but its value needs a clear funding and redemption basis. Distinguish a restaurant-funded discount from another arrangement. Check what happens to sale proceeds, applicable fees and included costs when the offer is used. Do not assume that every amount displayed to a customer is retained by the restaurant.

In a fictional fulfilled order, sale proceeds are eighteen currency units. Ingredients cost eight, packaging costs two, assumed channel-related variable cost is three, and a restaurant-funded offer reduces proceeds by two. Subtracting those stated amounts leaves three before other costs outside the example. This is contribution under the stated scope, not net profit.

A later order from an offer recipient does not prove that the offer created the order. Some customers would have returned anyway, while others may use the offer once without further purchases. Compare appropriate groups and context if evaluating effectiveness, and keep uncertainty visible. Avoid promises that a small compensation amount will recover a customer for life.

Keep the offer independent of review behavior. A customer should not have to leave praise, remove criticism or post publicly to receive an otherwise appropriate remedy. Make the operational fix the center of the response, and evaluate offers through actual outcomes and costs rather than optimistic assumptions about goodwill.

Follow a fictional café improvement from complaint to review

Willow Café receives several comments that collected drinks are difficult to identify. Staff initially suspect that customers are taking the wrong cups. Reviewing the counter shows another explanation: drink labels face the preparation area, while customers see identical cup lids. The problem is not established by the comments alone; observation reveals how the presentation contributes.

The café tests an outward-facing order identifier and a clearly marked collection position. Staff confirm the identifier before handing over the complete order, using the business's normal process and avoiding unnecessary personal information. The change is recorded with its start date and affected service periods.

The chosen measures are observed drink-identification incidents among eligible collection orders and staff use of the new handover step. The team also watches waiting time and congestion so the fix does not simply move the problem elsewhere. The latest feedback period remains provisional until normal reporting has caught up.

After a reasonable run for its own volume, the café reviews the records and staff observations. Fewer incidents would support continuing the process, but would not prove a permanent causal effect on repeat business. The café can separately examine mature return data if authorized information exists. Clearer handover is already a worthwhile operating result without inventing a lifetime-customer claim.

Make the team review useful and safe to contribute to

Use a short recurring review with a stable structure: what customers reported, what the evidence showed, what changed and what remains open. Select a manageable number of issues. A meeting that reads every review aloud can consume time while leaving the actual operating decisions unclear.

Invite the people doing the work to explain constraints. A packing issue may reflect insufficient space, unreadable tickets or incompatible packaging rather than a lack of attention. Staff should be able to report a problem without being blamed for making the numbers look worse. Otherwise the record can improve while the service does not.

Assign an owner and next review point to each chosen action. “We should improve communication” is not an action someone can verify. “Test the revised pickup sign during the next relevant service periods and record whether customers still queue at the kitchen door” has a concrete purpose and an observable result.

Close actions with evidence. Record whether the change was adopted, what was observed and whether another step is needed. Keep unresolved causes visible rather than repeatedly reopening the same discussion from memory. The review should accumulate practical knowledge about the restaurant, not merely display a stream of ratings.

Avoid common interpretation mistakes

Do not treat silence as universal satisfaction. Many customers do not review, and the reasons vary. Equally, do not assume every unfavorable review represents the same underlying failure. Read the specific account, compare the facts and choose a proportionate response. A single total cannot replace that judgment.

Avoid changing several definitions while claiming continuous improvement. Adding telephone complaints, revising categories and changing the eligible order population can alter the rate even when service is unchanged. Document those changes and explain how comparisons are affected. Honest discontinuities are preferable to a smooth chart built from incompatible periods.

Do not compare a customer's return behavior with a staff member's reply count as if they measure the same thing. Fast responses may improve case handling, but the content, remedy and underlying service still matter. A target that rewards sending any reply can encourage superficial messages without meaningful resolution.

Finally, resist turning every comment into a new discount. Some feedback calls for a clearer description, a recipe review, a better packing method or a direct explanation. Choose the response that addresses the actual issue. A learning process earns its value through more reliable execution and clearer decisions, rather than through a growing list of offers whose effect nobody can verify.

Audit one feedback case from receipt to learning

Choose a closed ordinary case and ask another authorized colleague to reconstruct it. They should find the original concern, the relevant order context where available, the factual investigation, the customer response and the operational action. If those stages exist only in one person's memory, the case is not yet a dependable learning record.

Check whether the final status says what actually happened. A reply sent is not necessarily a remedy completed, and a remedy completed is not necessarily an operating cause resolved. The record may legitimately show that the customer response is complete while a packaging trial continues. Separate statuses make that situation clearer than a single ambiguous tick.

Inspect the summary used in the team meeting. It should preserve the useful issue without exposing unnecessary customer information or presenting an unconfirmed cause as established fact. If the summary says “courier error” while the investigation found no evidence about handover, correct the summary before it shapes future decisions.

Finally, follow the case into the improvement record. Someone should be able to see whether a change was tested, adopted or rejected, and why. This small audit checks whether feedback is producing accumulated knowledge. It can reveal missing connections more effectively than adding another dashboard that counts comments without showing what the restaurant learned from them.

Sources and platform scope

Research checked on 17 October 2026. Uber Eats delivery-feedback guidance supports structured operational learning; its customer-journey guidance provides qualitative context for return experiences. DoorDash's merchant review help describes provider-specific visibility and responses. Google's review policy governs Google review requests. Metric definitions and fictional cases here are original illustrations, not reproductions of proprietary scores or promises about retention.