A 13-minute playbook for independent owners. Categorize every comp, install the authorization matrix, and run the weekly review that distinguishes the comps you want from the ones eating your margin.
Owners running this routine cut their comp rate 0.4 to 0.7 points.
Read This First
Comps and discounts aren't a single category. They're at least seven: service recovery, VIP, marketing promotions, employee meals, training comps, manager discretion, error corrections. Each one has a different purpose, a different ROI, and a different abuse vector. Lumping them on a single line of your P&L hides everything.
The benchmark moves with concept. Fine dining lives at 2.5%, full service at 1.5%, fast casual at 1.0%, QSR at 0.6%. If you don't know yours, you can't tell the difference between a healthy comp habit and one that is losing you money.
This playbook gives you the categorization, the authorization, and the cadence.
I take days to figure this stuff out. This is game-changer type stuff. I don't even see my numbers until end of month right now.
Andrew Foster
Owner, Caplingers Fresh Catch
What you will have by the end of this page
01A comp loss number specific to your concept and revenue, with a benchmark you can defend.
02The 7-Category Comp Taxonomy you can map onto your POS this week.
03An Authorization Matrix calibrated to your concept and span of control.
04The Manager Override Audit checklist for catching the patterns POS reports miss.
05The Weekly Comp Review routine. 15 minutes, every Monday.
Quick Win, 60 Seconds
The Comp Recovery Calculator
Concept-aware. Five inputs. Three buckets you can act on.
$
%
Your estimated annual loss:
$100,540
Comp variance vs. benchmarkbiggest
1.7 pts above full service benchmark of 1.5%
$76,500
Unauthorized comps
15% of comps issued without proper auth, 65% recoverable
$14,040
POS user sprawl
$800 / year per user with comp authority above necessary tier
$10,000
Your routing
You're well above the concept benchmark. Start with Chapter 2 (Categorize Every Comp) and Chapter 5 (Weekly Comp Review). The variance is in the comp mix, not the volume.
For reference: total comps issued at this volume = $144,000/yr
Benchmarks: fine dining 2.5%, full service 1.5%, fast casual 1.0%, QSR 0.6%. Your concept type drives the variance calculation. Adjust as your data improves.
Bring your questions to a free working session. No system access needed for the call.
Heads up: this playbook gives you the routine. It does not run the routine for you every morning. Marty does. Keep reading if you want the routine. Skip to the bottom if you want it run for you.
The Playbook
Six chapters. Thirteen minutes. One system.
Each chapter ends with something you've actually done.
CHAPTER 01
Set the concept benchmark
A 3% comp rate is healthy for fine dining and a five-alarm fire for QSR. Without the concept-anchored benchmark you can't read your own data. The benchmark isn't a target. It's the line that tells you whether the next conversation is "tune this" or "fix this."
Within a category, the benchmarks vary by concept maturity. New restaurants run hotter on training and service-recovery comps. Established restaurants should sit closer to the concept median. The first six months are a different game from year three.
Concept benchmarks (% of sales)
Concept
Benchmark
Healthy range
Fine dining
2.5%
2.0% to 3.5%
Full service
1.5%
1.0% to 2.2%
Fast casual
1.0%
0.7% to 1.5%
QSR
0.6%
0.4% to 0.9%
Your turn
Pick your concept. Pull your trailing 90-day comp % of sales. Compare. The gap is your target for the next 90 days. Anything inside the healthy range is a "tune," anything above is a "fix."
Takeaway
The benchmark isn't a goal. It's the line between tuning and fixing.
CHAPTER 02
Categorize every comp into seven buckets
Service recovery, VIP, marketing promotion, employee meals, training, manager discretion, error correction. Every comp belongs in exactly one bucket. Most POS systems let you tag this. Most owners don't because nobody set up the tags.
The categorization is the diagnostic. Restaurants with healthy comp habits run heavy on service recovery and VIP, light on manager discretion. Restaurants that lose money on comps run heavy on manager discretion with no narrative.
The 7-category comp taxonomy
Category
Purpose
Authorized by
Service recovery
Fix a service failure
MOD
VIP
Loyal regular, hosted guest
GM, Owner
Marketing promo
Active campaign / coupon
System (POS rule)
Employee meal
Per policy
System (POS rule)
Training
New hire, soft-open
GM
Manager discretion
Anything else
Manager (logged reason)
Error correction
Wrong item rung in
MOD
Your turn
Audit the last 30 days of comps. Tag every one into a bucket. Anything in "manager discretion" without a logged reason is your work item this week.
Takeaway
If half your comps are "manager discretion," half your comps are uncontrolled.
CHAPTER 03
Build the authorization matrix
Who can comp what, with what dollar limit, requiring what justification. By role. The matrix is one page. Print it. Pin it next to every POS terminal. The act of writing it down forces the decisions you've been avoiding.
The single highest-leverage line is the "manager discretion above $X" rule. Set the threshold low enough that a daily review covers it, high enough that you're not approving every birthday dessert.
Authorization matrix, full-service example
Role
Service recovery
VIP
Discretion
Server
≤$15 single item
no
no
MOD
≤$50
≤$25
≤$25 logged
GM
≤$200
≤$150
≤$100 logged
Owner
unlimited
unlimited
unlimited
Your turn
Draft your matrix in 20 minutes. Walk it through one server, one MOD, one GM. Have them tell you where they violated it last month. The violations point to where the limits are calibrated wrong.
Takeaway
The matrix isn't bureaucracy. It's the cover that lets staff make the right call fast.
CHAPTER 04
The Manager Override Audit
Every POS records when a manager swiped to authorize a comp. Most reports show you the comp. They don't show you the pattern. The pattern is what matters: same manager, same server, same time of week, same item.
Pull the override log monthly. Sort by manager, then by server. Patterns jump out. A single manager-server pair accounting for 40% of overrides isn't a coincidence. It's a relationship that's costing you money.
Override audit, monthly
Pull last month's manager override log.
Group by manager. Identify any manager with override volume > 25% of total.
Within that manager, group by server. Identify any pair > 30% of that manager's volume.
Pull the underlying comps for the flagged pair. Categorize. Look for repetition (same item, same time, same justification).
Have the conversation. The conversation is rarely "we caught you." It's usually "let's tighten the workflow."
Your turn
Run steps 1-3 this week. Most owners find one pair worth a conversation. The cadence keeps the pattern from re-emerging.
Takeaway
One manager + one server = a pattern, not a coincidence.
CHAPTER 05
The Weekly Comp Review
Fifteen minutes every Monday. Pull the prior week's comps. Three views: comp % vs. benchmark, comps by category, comps by user. Anything off-trend gets a one-line investigation note. That's it.
The review isn't about catching theft (that's the audit). It's about keeping the categorization clean and the benchmark visible. Owners who run this for 60 days find their comp % drops 0.4 to 0.7 points without any other intervention.
Weekly comp review template
View
This wk
Last 4 wk avg
Note
Comp %
2.1%
1.8%
Holiday catering, expected
Manager discretion
28%
17%
Investigate
Top user (Tara M.)
$340
$210
Likely VIP weekend
Your turn
Run last week's review tomorrow. Flag any line that drifted 30%+ from the trailing 4-week average. That's the work item for the week.
Takeaway
Run the review for 60 days, your comp % drops half a point. Cadence wins.
CHAPTER 06
Quarterly POS user cleanup
Every quarter, pull your full POS user list. Inactive users with comp authority are an attack surface and a sprawl tax. Former employees still in the system. Servers promoted to MOD whose old server account never got revoked. Each one is $800/year of preventable risk.
The cleanup takes 30 minutes. Pull the user list, cross-reference with active payroll, disable anyone not on payroll, downgrade anyone whose role no longer matches their authority tier. Done.
Quarterly POS cleanup checklist
Export the full POS user list.
Cross-reference with active payroll list.
Disable any POS user not on active payroll.
For each remaining user, verify role → authority tier mapping.
Downgrade any mismatch (over-authorized).
Document the cleanup in a one-line changelog.
Your turn
Block 30 minutes next quarter (or this week if you've never done it). The first cleanup usually disables 4 to 8 users and downgrades 2 to 4 more. After that it's a routine.
Takeaway
Active payroll = active POS user. Anything else is sprawl.
Print this cheat sheet
The Cheat Sheet
One page. Print it. Pin it.
Everything in this playbook, condensed.
Concept Benchmarks
Fine 2.5% / Full 1.5% / Fast casual 1.0% / QSR 0.6%. The line between "tune" and "fix."
7-Category Taxonomy
Service recovery / VIP / Marketing / Employee meal / Training / Manager discretion / Error correction. Every comp = one bucket.
Authorization Matrix
By role × category. Server / MOD / GM / Owner. Discretion threshold low enough for daily review, high enough not to bottleneck.
Override Audit
Monthly. Group by manager, then by server. One pair > 30% of a manager's volume = conversation.
Weekly Review
15 min Mondays. Comp % vs. benchmark / by category / by user. 30%+ drift = work item.
Quarterly Cleanup
POS user list × active payroll. Disable non-payroll. Downgrade over-authorized. 30 min.
What this playbook can't do
You now have the system. Marty runs it for you.
A playbook can't:
Tag every comp into the 7-category taxonomy automatically
Watch comp % vs. concept benchmark every day, not every Monday
Surface manager + server pattern flags before they harden
Run the quarterly POS user cleanup on a recurring schedule
Marty does all of it.
Marty is the analyst sitting next to your GM. We plug into your POS, run this routine every morning before you open, alert the moment a number breaks, and review the rest of your operation in the background. It starts with a 2–3 store Cash Recovery Plan pilot. If the Cash Recovery Plan doesn't surface anything worth fixing, we refund the plan fees in full. You keep the work.