RESTAURANT OPERATIONS · September 2026 · ~11 min read
Comps, voids, and what they tell you about the floor
A void removes an item before it was made. A comp removes the charge after the guest already has the food. That difference is the whole diagnostic. Voids point at ordering and POS problems, comps point at a service or kitchen failure that already cost you the product. Most operators watch the dollar total and miss both signals.
On this page
- 01What does a month of comps actually cost you?
- 02Why does the reason code matter more than the amount?
- 03What patterns should you actually look for?
- 04Who should be allowed to comp?
- 05What changes when there are two locations?
- 06Is it cheaper to fix it afterward instead?
- 07Where does this reading break down?
- 08What to do this week
- 09When you do not need this
- 10Sources
- 11Related reading
- 12Questions about your comp log?
The total is the least interesting part of the report. A month of comps is a record of every moment your restaurant failed a guest badly enough that somebody gave money away to fix it. That is expensive research you already paid for.
Comps are not a cost line. They are a defect log with a dollar sign on it.
01What does a month of comps actually cost you?
More than the total on the report, and the gap between those two figures is the part worth sitting with.
Here is the arithmetic with numbers you can swap for your own. Take a full service restaurant doing $1.2M in annual net sales. Pull sixty days: net sales of $196,000 and comps and voids together of $4,180.
That $4,180 is where most operators stop reading. It looks small next to $196,000, which is exactly why nothing happens about it.
Now follow the dollar. The guest ate the food, so the food cost was already paid. Nothing about comping it made the kitchen cheaper. Had that check been collected, you would have the money and your costs would be identical. A comped dollar comes straight off the bottom line, not off the top.
Annualized, $4,180 over sixty days is roughly $25,400 a year. The National Restaurant Association describes the pre-pandemic shape of a typical independent restaurant as food taking about 33 cents of every sales dollar, labor about 33 cents, everything else about 29% of sales, and a pre-tax margin near 5%. At a 5% margin, $25,400 of pre-tax profit is what a restaurant doing roughly $508,000 in annual sales produces in a year.
So a $1.2M restaurant with an unremarkable comp report is giving away the entire annual profit of a half million dollar restaurant, and calling it good service.
Now split the same $4,180 by reason instead of by dollar rank:
| Reason | 60 day total |
|---|---|
| Long ticket time | $1,690 |
| Wrong item | $920 |
| Generic manager comp | $930 |
| Quality | $640 |
That table is a work order. Ticket time and quality are the kitchen, and together they are $2,330 of the $4,180. Wrong item is the handoff between the floor and the line. The generic code is the one that should not exist, because $930 of your defect log is currently unreadable.
Annualize the top line alone. $1,690 over sixty days is about $10,280 a year. Cut it in half by fixing one station's throughput and you have found $5,140, which at a 5% margin is the profit on roughly $102,800 of sales. Nobody had to sell anything.
Run this on your own sixty days before you talk to anyone about it. The conversation changes when the number is annual.
02Why does the reason code matter more than the amount?
Because the amount tells you the size of the problem and the reason tells you where it lives.
A month sorted by dollars gives you a number to be unhappy about. The same month sorted by reason gives you a work order. Long ticket time is a kitchen or expo problem. Wrong item is a communication problem between the floor and the line. Quality is a recipe or a station.
Which means the reason codes have to be worth something. Most restaurants have too many, or have one called manager comp that absorbs most of the entries and explains nothing.
Cut the list to five or six codes that map to a fixable cause, then require one every time with no default. A code that is preselected is a code that is never chosen.
The food that walked out on a comp also left the building without a matching sale, which is one of the places food cost variance actually hides. If you have been hunting a gap in the walkin and your comp volume is high, you may have been looking in the wrong room.
03What patterns should you actually look for?
Four, in order of how often they turn out to matter.
By shift. Sort by day and daypart. If one shift produces most of them, you have a shift problem, not a restaurant problem. That is a staffing, manager, or volume question, and all three are fixable.
By item. One dish showing up repeatedly is telling you something concrete. It is hard to fire, it takes longer than the guest expects, or it is not what people think they ordered. An item comped often costs more than its recipe says, which matters when you decide which prices to move without losing covers.
By employee. Read this one last and carefully. High comps on one server is often a bad section, a slow station, or someone never taught what to say when food is late. Sometimes it is a person buying goodwill with your food. The report cannot tell the difference. Two shifts on the floor can.
By hour. Comps clustered at one time usually mean capacity. The kitchen falls behind at a predictable point, tickets stretch, and the floor pays for it in dessert. That is a prep or staffing fix, not a service lecture.
Voids sort differently. They cluster around training and menu structure. A high void count on a modifier heavy item usually means the POS build is confusing, not that the staff is careless.
04Who should be allowed to comp?
Someone on every shift, with a number, and a record.
The two failure modes are opposite and both common. In the first, only the owner can comp, so when the owner is out the floor has no tool and the fix arrives as a bad review two days later. In the second, everyone can comp anything, the number grows quietly, and nobody can reconstruct why.
The workable version is a shift lead with a per table cap and a per shift cap, a required reason code, and a name on every entry. The manager reviews the log at close, not the following week. A comp explained at close is information. The same comp on Thursday is a memory.
Then let people use it. A floor that is afraid to comp hands you a clean report and a slow bleed of guests who never came back, and those appear in no system you own.
05What changes when there are two locations?
The comp report stops being a management tool and becomes the only window you have.
At one location you comped it yourself or heard about it. At two, half of those events happen in a building you are not in, and the log is the record. Reason codes matter more once you grow, and that is one of the specific things that changes when you go from one location to two.
Compare locations as a share of sales rather than in dollars, and compare the reason mix. A store with a different reason profile has a different problem at the same total. One is losing money to ticket times, the other to wrong items, and those are two different conversations.
Be careful comparing months without accounting for volume, because a busy month produces more comps for structural reasons. Same reasoning as why seasonality distorts month over month comparisons.
06Is it cheaper to fix it afterward instead?
No, and this is the part of the category where restaurants get sold something worse than the problem.
A comp is recovery at the table, while the guest is still in the room and the cost is a dessert. The alternative product being marketed to restaurants is recovery after the fact, and a meaningful share of that market is running fraud. Joy Hawkins of Sterling Sky, who has documented this firsthand, describes review removal companies charging businesses thousands "often through shady tactics. One common trick is abusing DMCA takedown notices with false claims, which Google often approves automatically." She documented a forum thread critical of one such company being removed from Google on a notice claiming the post had stolen content from a news article about an earthquake in Haiti. Google approved it. Traffic to that page went to zero.
Google's own published position on removals is narrower than the sales pitch: "You can report any review, but only those that violate Google policies are eligible for removal... Do not report a review just because you disagree with it or dislike it. Google doesn't get involved in conflict between businesses and customers." There are legitimate grounds, and they are specific. A review from a former employee or a competitor is removable on a named conflict of interest ground. A review from a guest who waited fifty minutes for an entree is not.
A comped dessert at the table is the cheapest reputation management that exists. Everything downstream of the front door costs more, works less, and in part of the market carries legal exposure that gets sold to you as a service.
07Where does this reading break down?
Three places, and knowing them keeps you from managing a number that is telling you nothing.
When the comps are marketing. Chef sends, industry courtesy, a table you are deliberately investing in. Those are spend, not defects, and they belong in their own code so they stop polluting the diagnosis. If they sit inside your comp total, your defect log is wrong by whatever that number is.
When the person entering the code benefits from the code. A server choosing between wrong item and long ticket time is choosing between their own error and the kitchen's. This is not dishonesty, it is incentive. The correction is a manager approving the entry with the reason, not the reason arriving unsupervised.
When you want a benchmark and there is not one. There is no published comp percentage for full service restaurants that survives checking. The National Restaurant Association's Restaurant Operations Data Abstract, 2025 edition, collected financial and operating data from more than 900 operators nationwide, and its segment cuts by check average and sales volume sit behind a paywall. The Association also states plainly that its data is "not intended to represent standards or goals for individual restaurants." So compare your comp rate to your own last six months, by reason, and ignore anyone who quotes you an industry number without a sample size.
08What to do this week
Pull the last sixty days of comps and voids with reason, employee, item, and time. If your POS cannot produce that, the limitation is itself useful information.
Sort by reason. If one code holds more than a third of the entries and that code is a generic one, your codes need rebuilding before the data means anything.
Then sort by item and pick the top one. Go watch it get made and delivered during a busy service. The answer is almost always visible in ten minutes.
Rewrite your comp authority in one sentence with a dollar cap and post it where the floor can see it. Review the log at the end of every shift for two weeks.
Be honest with yourself
When you do not need this
If your comp volume is small and stable, look at it once a quarter and spend your attention elsewhere. Not every number deserves a weekly meeting.
If you run counter service where the guest sees the food before paying, most of this does not apply. Your defect log lives in refunds and remakes, and the same sorting logic works there.
And if comps spiked in a month when something specific happened, a menu change or an equipment failure, do not build a program around it. Fix the cause and look again in thirty days. Reacting to one bad month is how restaurants end up with rules nobody remembers the reason for.
Sources
- National Restaurant Association, "Elevated costs continue to pressure restaurant profitability". 8 July 2026. Trade association estimates built from government data and operator surveys. Source of the cost structure and the 5% pre-tax margin used to convert comp dollars into a sales equivalent.
- National Restaurant Association research reports. Home of the Restaurant Operations Data Abstract, 2025 edition, released August 2025, built on financial and operating data from more than 900 operators nationwide. Paid publication, trade association research. Cited here for the absence of a public comp benchmark and for the Association's own caveat about using its figures as standards.
- Sterling Sky, Joy Hawkins, on the review removal industry. Practitioner documentation, first hand, updated 1 August 2026. Source of the DMCA takedown abuse account. Sterling Sky sells local SEO services, including legitimate policy based review removal.
- Google, "Remove reviews from your Business Profile". First party policy documentation. Source of the removal eligibility language and the conflict of interest ground.
Related reading
- Reading your POS data for something other than sales. where the comp and void reports live, and the four other reports worth pulling in the same sitting.
- The operating controls that hold under a Friday night rush. how to write the comp authority rule so it survives a full dining room instead of collapsing into it.
- Waste tracking without a full inventory program. the kitchen side defect log, which explains the food that never reached a guest at all.
- Review response as customer service, not damage control. what to do about the guest whose comp came too late, once they are writing in public.
Questions about your comp log?
Email me at eric@seod.com with one month of comps broken out by reason code and shift. Strip the names if you prefer, I do not need them. I will tell you whether what you have is a kitchen problem, a floor problem, or a menu problem, and which single item I would go stand and watch.
Sixteen years running multi unit restaurants and more than $54M in annual P&L, and the comp log was the most honest report in the building every time. Guests are reacting to what they experience, which includes what they expected before arriving, and expectations get set early by things like whether your photos match the room.
Otherwise, there is more on running the operation.