Guest experience metrics for restaurants: what to count each week

Count seven things each week: how often the wait you quote matches the real wait, kitchen ticket times, drive-thru times if you have a lane, order errors per 100 orders, the share of calls answered, new review ratings and your reply rate, and how many guests come back. Start by checking which records your POS, waitlist, phone log and review pages actually provide.

These counts describe recorded parts of a guest visit, not every guest's experience. Some accounts lack the needed fields; use a small manual log where practical and mark missing data unknown. This guide explains the counting rules and ends with a weekly sheet you can copy.

Which guest experience numbers can you measure?

Metric	How to count it	Where the data lives
Wait-quote match	Parties seated within five minutes of the wait you quoted, divided by all waitlist parties	Waitlist or reservation system
Kitchen ticket time	Tickets over your time limit, divided by all tickets	Kitchen display system reports
Drive-thru time	Average time from order point to exit, by daypart	Drive-thru timer
Order errors	Distinct orders with a recorded wrong or missing item, divided by all orders in the same week, multiplied by 100	POS void and refund reports, plus a paper log
Answered-call rate	Calls answered by a person or a phone service, divided by all incoming calls	Phone system call log
Review rating and reply rate	Average stars on new reviews; public replies on those same reviews divided by that review count, at a stated follow-up date	Google Business Profile, Yelp and other review sites
Repeat-guest share	Distinct supported guest or account records with a visit this week and in the 60 days before the week began, divided by all distinct supported records with a visit this week	Reservation guest records or a loyalty program

The five-minute margin, the ticket limit and the 60-day window are our suggestions. Pick your own, write them at the top of the sheet, and keep them the same from week to week.

How do you count wait times?

Record the quote at the host stand, then the time the party sits. If you use Toast Tables, its Bookings report lists quoted against actual wait time for each booking, along with turn time, party size and source, and you can filter it by date, party size and source and download it. Our waitlist app guide compares other waitlist tools.

Compare party sizes separately so the overall figure does not hide a change in the mix of parties.

How do you track kitchen ticket times?

Toast's platform guide distinguishes prep station fulfillment time from whole-order timing. Tickets by fulfillment reports station times; Ticket details measures the interval from the first item firing to the last item being fulfilled. Fulfillment times are available only for KDS devices. Confirm which report and timestamps your account provides before comparing weeks.

Printed tickets carry no bump time. Without a KDS, have the expo write the time on 20 tickets during the peak and work out how many ran past your limit. Our kitchen display guide covers the screens themselves. For a lane, our drive-thru timer guide explains what each timer stage measures and gives a worksheet for timing cars by hand.

How do you count order errors?

For each week, count distinct orders with a recorded wrong or missing item, divide by all orders in that same week, and multiply by 100. An order with a complaint, a remake and a refund counts once in this measure. Keep those three event counts on separate lines; adding them would count one affected order more than once. Use an order reference to match records within the restaurant, mark unmatched reports separately, and never put guest details in a public report. Record the channel, cause and review cutoff; if a later complaint changes an earlier week, label the revision.

Voids give a partial count. Toast's Sales Summary report has a Void Summary with the void amount, order count and item count, and links to the Voided Orders report. Staff also void for reasons that have nothing to do with the guest, and a fix handed over free may leave no void at all. Our phone order accuracy guide gives an error log for phone orders, and our drive-thru order accuracy guide gives an audit for the lane.

What is your answered-call rate?

Divide answered calls by all incoming calls to the restaurant's number. Most business phone systems keep a call log. RingCentral's admin call log, for example, shows whether each call was connected, accepted or missed, and lets an admin filter, download or schedule the log by email.

Count every incoming call attempt, redials included, and count a transferred attempt once. If you also group attempts by displayed caller number within an hour, keep that separate from the call-answer rate: a number does not establish a unique person, and shared, blocked or missing numbers limit the grouping. Keep missing identifiers unknown. Separate staff-answered, service-only, voicemail-only, unanswered and routing-unknown attempts by hour; retain unknowns in the total. Our call volume guide walks through that count with a worksheet.

If a phone service answers some calls, keep its count apart from staff. Maple's Voice and Pro plans include call analytics, and our product page suggests counting answered calls separately from completed requests and staff handoffs. Check whether the caller’s intended request was completed; answering alone does not establish that.

How should you track reviews?

Each week, keep a fixed list of new reviews for each site, with the rating and date recorded when you collect them. Average the stars for that list only. For reply coverage, divide reviews on that same list with a public reply by the total reviews on the list. Record when you checked, and compare weeks at the same follow-up interval. Track replies submitted, pending or rejected separately from public replies; replies to older reviews do not enter the new-review numerator.

Google's help page on reading and replying to reviews says you must verify your Business Profile before replying. Google checks replies before publishing approved ones; review usually takes up to 10 minutes but can take up to 30 days. Customers can still change their reviews. For guest-visible coverage, check that the reply is public rather than counting it when submitted. A submission count is a separate measure of staff activity.

Keep the count honest. Google's Maps content policy bars businesses from offering incentives for reviews, from discouraging negative reviews and from asking only happy guests for reviews.

How do you measure return visits?

Define the supported guest or account record you can match across visits; a name alone does not establish the same person. Count each matched record once per week, and check for a prior qualifying visit in the 60 days before that week began. Record exclusions such as cancelled visits and no-shows, and leave unmatched visits unknown rather than treating them as new guests. Toast Tables' Guests report, per the same reports article, lists every guest in the waitlist and reservation system with the number of bookings and covers, leaving out cancelled bookings and no-shows. Shared accounts, duplicate records, cash visits and changed cards can limit identity matching. Treat the result as a measure of recorded repeat visits, not a count of all returning people.

What goes on the weekly review sheet?

Every figure below is made up. Fill in your own counts using the same definitions and record each review cutoff. The two reply-coverage examples assume the same follow-up interval after each review week; pending replies stay outside the numerator. Compare with past weeks only when the records and definitions are comparable.

Metric	Last week (made up)	This week (made up)
Wait-quote match	Last week: 131 of 172 parties, 76.2%	This week: 144 of 180 parties, 80.0%
Tickets over 15 minutes	Last week: 81 of 1,190 tickets, 6.8%	This week: 62 of 1,240 tickets, 5.0%
Order errors per 100 orders	Last week: 24 distinct affected orders in 1,790 orders, 1.34	This week: 18 distinct affected orders in 1,850 orders, 0.97
Answered-call rate	Last week: 571 of 702 calls, 81.3%	This week: 548 of 640 calls, 85.6%
New reviews, average stars	Last week: 20 reviews, 4.4 stars	This week: 23 reviews, 4.3 stars
Review reply rate	Last week: 12 of 20 reviews have public replies, 60.0%	This week: 19 of 23 reviews have public replies, 82.6%
Repeat-guest share	Last week: 101 of 395 matched guest/account records, 25.6%	This week: 118 of 410 matched guest/account records, 28.8%

Give each line an owner, such as the host lead for waits and the kitchen manager for tickets. In the weekly meeting, pick the one line that got worse and agree on one change to test. In this made-up week, the rating slipped while every other line improved, so the manager would read the three lowest reviews before changing anything.

Published by Maple, which sells AI phone answering and ordering to restaurants. This AI-assisted guide combines Toast, RingCentral and Google help pages with an original metric table and weekly sheet. Every figure in the sheet is made up, and the thresholds are our suggestions. The guide reports no survey or study and makes no claim about any restaurant's results.