How 204 Restaurant Groups Improved Food Cost Control in their First 18 Months on Rosnet

The problem

The average restaurant spends more on food than their own recipes say they should. Not by a trivial margin, but by roughly 9 percentage points when they first start tracking it accurately. The question operators and their ownership groups actually care about is whether that gap closes, how fast, and whether the improvement reflects real operational change or just better accounting.

This is what five years of inventory data from 204 restaurant groups, 5,281 locations, and 145,000 inventory periods shows.

The quick answer

• New Rosnet customers begin with food cost running about 8.7% above theoretical expectations.
• By month 17 on the platform, that gap closes to about 6.7%, a 2+ percentage point improvement.
• The improvement is behavioral: kitchens waste less and portion more accurately. Recipe targets do not shift upward to manufacture the result.
• The pattern holds at every operator size, from 5-location groups to 500+ location enterprise systems, and is validated by major franchise brands.

How the analysis works

The analysis uses AvT Ratio, Actual food cost divided by Theoretical, as its primary measure. Theoretical food cost is what a location should spend based on its recipes and what the POS recorded as sold. Actual is what the inventory count shows was spent. The ratio normalizes for location size, menu mix, and sales volume, so a quick-service unit doing $8,000 in weekly food revenue and a casual dining location doing $80,000 are directly comparable.

A ratio of 1.00 means actual matches theoretical exactly. A ratio of 1.087 means actual food cost is running 8.7% above what recipes predict, roughly $870 per week in unexplained variance for every $10,000 in theoretical food cost. The dataset covers July 2021 onwards, deliberately excluding 2020 and the first half of 2021, where COVID-driven supply chain disruptions and staffing instability distorted inventory patterns. Each location was anchored to its own first month on Rosnet and tracked through its first 17 periods, giving an improvement curve independent of calendar timing or start date.

Finding 1: The improvement curve is real, steady, and predictable

The average AvT Ratio at Month 1 is 1.0895, meaning locations begin with food cost running about 8.7% above theoretical. That ratio declines month over month throughout the early-tenure window.

Error bars (plus or minus 1 standard error) are tight at each tenure milestone, which means the improvement trend is broadly shared across locations rather than driven by a small subset of high-performing outliers. The trend is nearly linear. Most improvement happens in the first several months, followed by gradual, continued tightening as locations build consistent inventory habits.

What 2 percentage points is worth in dollars

Here is the simple math: take your total monthly food revenue across all locations and multiply by 0.02. That is roughly what a 2 percentage point improvement puts back every month.

Swap in your own numbers. The 2+ percentage point figure is the average at Month 17 across the full dataset, and operators who start with more variance typically see bigger gains. The revenue figures here are illustrative. These are not projections someone built in a spreadsheet. They are what the math looks like when you apply observed results to real portfolio sizes. Thrive Restaurant Group's leadership estimated their theoretical food cost tracking alone was worth over $1 million in annual food cost savings, which lines up with exactly what this math predicts at their scale.

"Purchasing, inventory, scheduling, all reporting is in Rosnet. It is the one-stop shop we need to make decisions and run better restaurants." Area Director, Neighborhood Restaurant Partners

If you want your actual number rather than an estimate from a table, Rosnet can run a preliminary AvT gap analysis for your portfolio before implementation. Most groups find more variance than they expected.

Finding 2: The improvement is behavioral, not an accounting adjustment

When AvT improves, it can happen two ways: actual food cost falls, or theoretical targets shift upward. The second scenario is recipe inflation, operators adjusting their expected cost upward rather than reducing actual waste. The analysis tested this directly by tracking Actual's share of total food cost over the same 17-month window. If improvement came from Theo increasing, Actual's share would stay flat or rise. If improvement came from Actual falling, Actual's share would decline.

Actual's share declines consistently over tenure across every size group. Theoretical cost stays stable or increases slightly, which means recipe targets are not being relaxed. Operators are genuinely wasting less, portioning more accurately, and managing inventory more tightly. The gap closes because kitchens get better, not because the target moved. For a board or ownership group that has seen software vendors overstate ROI, this is the answer. It is not a reporting trick. It is behavioral change, measured across 145,000 inventory periods at 5,281 locations.

Finding 3: Every operator size improves

The analysis segmented operators into four groups: Small (1 to 5 locations), Medium (6 to 20), Large (21 to 75), and Enterprise (76+). The finding is straightforward: every group improves during their first 17 months on Rosnet. No size group is an exception. What differs by size is what that improvement looks like in practice.

Large and Enterprise operators start closer to their theoretical food cost targets because they tend to have stronger processes already in place. Their improvement is steady and gradual, less dramatic in absolute terms but highly consistent. Think of it as fine-tuning an operation that is already reasonably tight.

Small operators often have more variance to begin with, so the absolute improvement is larger. Their month-to-month numbers are noisier because one bad inventory period at one location moves their whole average, but the underlying downward trend is just as real.

Medium operators show the most modest improvement trajectory of any group. The analysis suggests this reflects a common structural challenge at this scale: they have grown past the informal controls that work at a few locations, but have not yet built the centralized oversight that larger operators rely on. Medium operators who invest early in count discipline and above-store visibility tend to improve faster than those who do not. The bottom line is simple: if your group is 5 locations or 500, the data shows the same direction. How dramatic the improvement looks depends on where you started.

Finding 4: This is not just a fix for struggling operators

A reasonable skeptic reading the first three findings would ask: is this only working because some operators were doing things badly? If the improvement were mainly a catch-up effect, disorganized groups finally getting their counts right, you would expect mature, well-run brands with disciplined operations to show little to no improvement.

Wendy's is the test case. It is one of Rosnet's largest brand relationships: hundreds of locations, multiple franchise groups, and a high baseline of operational discipline. If anyone should be the exception, it is them. They are not. Wendy's locations follow the same improvement curve as their Large and Enterprise peers when compared on a size-adjusted basis. The trajectory is nearly identical, steady, low-volatility gains over the first 17 periods.

For Enterprise Wendy's operators specifically, the data shows minimal change in Actual Share over the tenure window. That is exactly what you would expect from a brand already operating near its optimum. At that level of maturity, the value is not a dramatic swing in improvement, it is early warning when something drifts and visibility into where. Stability in a tight operation is a signal worth paying for. The practical takeaway: if your concern is "we already run a tight operation, so there probably is not much here for us," the Wendy's data says otherwise.

What the data does not claim

This analysis is intentionally bounded to the first 17 months of each location's tenure, where data integrity is cleanest. Long-tenure analysis beyond 18 months requires additional historical data to distinguish truly new adopters from locations that entered the dataset mid-tenure, and that validation work is ongoing. The analysis does not attribute all improvement to a specific Rosnet module. Whether improvement is driven mainly by recipe management accuracy, inventory count discipline, or above-store reporting varies by operator. What the data establishes is that the improvement pattern is consistent, the mechanism is behavioral, and the trend holds at every organizational scale.

What operators ask most often

Does this hold for my concept?

The pattern holds across QSR, fast casual, casual dining, and fine dining operators in the dataset. Concept type affects starting AvT Ratio and absolute dollar variance per period. The directional improvement during early tenure is consistent across all of them.

What drives the fastest early improvement?

Locations that establish a consistent count cadence early, same day and same time each week, show faster convergence toward theoretical expectations. The inventory discipline that produces accurate Actual numbers is also the discipline that reduces waste. The measurement system and the operational behavior it enables are the same thing.

How does our current performance compare?

If you have historical food cost data and POS sales, Rosnet can run a preliminary AvT analysis for your portfolio before implementation. Most groups who have not previously tracked AvT find more variance than they expected.

See your own numbers

See what your food cost gap looks like, and what closing it would be worth. Talk to a Rosnet team member.