The P&L tells you what happened. PMIX tells you what is about to happen. Most operators are reading the wrong report at the wrong cadence, which is why they always find out about operational problems a week too late.
Product mix, PMIX for short, is the breakdown of what guests ordered as a percentage of total items sold. It is generated free by every modern POS. Toast prints it. Micros prints it. Square for Restaurants prints it. Almost none of the operators I work with actually read it. They read the monthly P&L, they read the weekly labor report, they read the daily sales flash. PMIX sits in a folder nobody opens.
That is a mistake. In sixteen years of multi-unit ops across 21 franchise units and $54M+ of operating scope, the single most consistent early warning signal I have found is a PMIX shift of 3 points or more in a category, week over week, sliced by daypart. That signal precedes almost every meaningful operational failure by five to ten days.
Why PMIX moves before the P&L does
The P&L is a summary. It aggregates a lot of ticket-level noise into a monthly rollup that smooths out the exact volatility you need to see. By the time a P&L line has moved 2 percent, the underlying operational cause has been running for two to four weeks.
PMIX is ticket-level. Every ticket contributes to it the moment it closes. A change in the operating environment shows up in guest behavior within 48 to 72 hours, and PMIX captures that shift on the same-day timeline. The lag between "something changed on the floor" and "the number moved" is measured in days, not weeks.
Here is the mental model to hold. The P&L is a rear-view mirror. PMIX is a side mirror. You use both, but the side mirror tells you what is about to come into view.
Read attach rates, not top sellers
The most common mistake operators make with PMIX is looking at the top sellers list. The top sellers list barely moves. Your best-selling burger will still be your best-selling burger next week. The signal is not in the ranking. The signal is in the attach rate.
Attach rate is the percent of guest checks that include a given category. Four attach rates worth tracking at minimum:
- Appetizer attach. Percent of tickets with an appetizer. Healthy full-service benchmark: 38 to 48 percent. Fast-casual: 15 to 22 percent.
- Dessert attach. Percent of tickets with a dessert. Healthy full-service benchmark: 22 to 30 percent. Fast-casual: 6 to 12 percent.
- Beverage attach. Percent of tickets with a beverage beyond water. Healthy full-service: 68 to 82 percent. Fast-casual: 55 to 68 percent.
- Modifier attach. Percent of tickets with any paid modifier (add cheese, extra protein, side upgrade). Healthy: 30 to 45 percent across formats.
When any of these moves 4 points or more against a four-week trailing baseline for the same daypart, something operational is happening. The number will tell you where to look.
The three signals and what they mean
Appetizer attach drops 4 to 6 points
This is a service pace signal. Guests order appetizers when the greet is fast, the water hits in under two minutes, and the first ten minutes feel attentive. When appetizer attach drops without a menu change, the answer is not the menu. It is the pace.
Look at greet time on the shifts where the drop is largest. Look at server section sizes. Look at whether an expo position is being covered by whoever is closest instead of a dedicated body. In a fast-casual environment, look at whether the front counter is being double-tasked with online orders in a way that slowed the greet.
Dessert attach drops but check average holds
This one is subtle and it is my favorite tell. If dessert attach drops but check average holds, the entrees are getting bigger. That is portioning drift on the entree line. A cook is plating heavier than spec, guests feel full, dessert order stops happening, check average holds because the entrees themselves are absorbing the dessert dollar.
This shows up in food cost as a 1 to 2 point rise 10 to 14 days later. Catch it at the PMIX stage by pulling the entree line off spec sheets, reweighing portions on a shift, and re-training the closing cook that week. Cheap fix, small window.
Beverage attach drops on specific days
If beverage attach drops but only on Thursdays, the answer is almost never the beverage program. The answer is that the bar is short-staffed on Thursdays, or the bar closing shift is being pulled from onto the floor because someone called out. Guests order fewer drinks when the drinks arrive slow. Look at bar labor by day of week and cross-check the drop.
Fig. 1 · Protein-bowl attach rate. Inflection at week 9. P&L moved at week 11.
Case: three Hana units, one wok station, eight days
Hana Group runs 21 franchise units across six states inside Walmart, Sam's Club, Whole Foods, and Target footprints, roughly $36M in P&L. Franchise means the units are close-cousins to each other in equipment, menu, and format, which makes cross-unit PMIX comparisons unusually clean.
In one region, three units in the same three-state cluster showed the same PMIX pattern in the same week: protein bowls dropped from a 26 to 27 percent share of items to 22 to 23 percent. Grain bowls, which do not pass through the wok station, held share. Salads picked up two of the four points the protein bowls lost.
Same shift across three units, same week, only on items that used one station. That is not a menu preference change. That is not a marketing change. That is a signal, and the signal points at the shared piece of equipment.
When the same category shifts across multiple units in the same week, it is almost never a guest preference story. It is a supply-side story. Look at what those units share.
We pulled the maintenance logs. Nothing flagged. We sent a regional field lead to one of the three units on the next visit day. Wok station recovery temperature was running 40 degrees below spec. The heating element was starting to fail. Same part, same manufacturer, same install date across all three units. Roughly a year in, all three were on the same failure curve.
We replaced elements in all three units the following week. Protein bowl attach came back to 26 percent within nine days. The P&L on those units was going to show a 1.2 point food cost rise and a 0.8 point revenue drop the following month, because slower recovery means longer ticket times and shrinking wok-item share. We caught it eight days ahead. That is what PMIX is for.
P&L snippet: what the miss would have cost
Per unit per month, if we had waited for the P&L to reveal it:
- Revenue impact: -0.8 percent on $180,000 monthly = -$1,440 revenue
- Food cost as percent of sales: +1.2 points = -$2,160 in margin
- Labor variance: +0.4 points (slower ticket times) = -$720
- Total per unit per month: roughly -$4,300
- Across three units, one month of delay: -$12,900
- Across three units, if it took the P&L a full quarter to surface it: roughly -$39,000
A weekly PMIX read caught a problem that would have cost $39,000 across the quarter for one region. Multiply that across a 21-unit portfolio and PMIX becomes the highest-return report on the field team's desk.
How to build the weekly PMIX read
The report itself is simple. What matters is the cadence and the slicing.
- Pull PMIX weekly, every Monday, for the trailing seven days. Same time, same person, same format. This is a discipline, not an ad-hoc report.
- Slice by daypart. Lunch and dinner have different attach rates and different failure modes. Weekend brunch has its own. If your POS will not slice this natively, export the raw ticket file and pivot it in a spreadsheet.
- Compare to a four-week trailing baseline for the same daypart. Not last week. Not last month. Same-daypart trailing four weeks. That is the baseline that filters out noise and shows real drift.
- Flag any category-level attach shift of 4 points or more. That is the alert threshold. Anything smaller is inside the noise band and will resolve itself. Anything larger is a hypothesis to investigate.
- Cross-unit comparison in a multi-unit environment. The single strongest signal is when the same shift appears in multiple units simultaneously. That almost always points at a supply-side cause: an equipment cohort, a vendor change, a menu-guide update, a shared training issue.
Building the muscle across a region
A single unit reading PMIX weekly is useful. A region reading PMIX weekly is transformative. The cross-unit comparison is where the signal gets loud.
The regional PMIX read has one extra column the single-unit version does not: unit-versus-unit variance for the same category, same daypart, same week. When one unit's dessert attach is running 8 points below its three peer units in the same market, the answer is almost never guest preference. Guests in the same trade area buy roughly the same mix. The gap points at something happening inside that one unit: the server team, the closing rotation, the plating standard, the freezer temperature on the dessert case.
Set a standing agenda item in the weekly regional call: three PMIX flags per week, one hypothesis per flag, one field verification assigned by name and due date. That is the whole meeting slot. Twelve minutes if the general managers come prepared. The units that get in the habit start solving problems before they escalate, and the units that do not stand out immediately.
PMIX is a team sport at the regional level. One person watching it catches half of what a whole region watching it catches.
What PMIX will not tell you
PMIX is not a diagnosis. It is a hypothesis generator. The number will point you at a category, but you still have to walk the unit within 48 hours to confirm what the shift actually means. I have seen operators try to fix a PMIX signal from behind a laptop and land on the wrong cause. It is a field job.
PMIX will also lie to you on a low-volume day. Sunday night at a fast-casual with 40 tickets is not a statistically useful PMIX sample. Aggregate at the weekly level for the low-volume dayparts, then compare weekly rollups instead of daily.
And PMIX cannot see comps and voids unless your POS categorizes them. In many older POS setups, comps drop out of the mix entirely and you lose the signal. Verify with your POS admin that comps are being counted in the mix denominator, not excluded.
The point
The operators who spot problems before the P&L are not smarter than the ones who do not. They are reading a different report at a different cadence. PMIX weekly, by daypart, with attach rates against a four-week trailing baseline. That is the whole discipline. It takes an hour a week per region and it is worth more than any dashboard that costs money.
Guests are the honest signal. They are telling you what is wrong on the floor every time they place an order. PMIX is how you hear them. Read it before the P&L makes you.