What actually broke when I went from one location to many

The first location was never the hard part. One kitchen, one team, one person carrying the whole picture in their head. I knew what was in the walk-in without opening a spreadsheet. I knew which line cook was having a bad week before service started. I knew what we were short on because I had felt it on the pass the night before. That is not a system. That is proximity. And proximity is the single most expensive thing to lose, because you do not notice it leaving.

The second location is where the illusion cracked. Not immediately - the first six weeks were fine, because I was physically there almost every day, doing to the new site exactly what I had always done to the old one. The trouble started once I stopped being able to be in both buildings. The new site quietly began running on my memory rather than on anything written down, and my memory only travels as far as I do.

The escalation trap

The thing that nearly buried me personally was not cost or quality. It was my phone. As the site count climbed I was fielding decisions all day that I had no business being involved in. A delivery short two cases. A dishwasher not showing. Whether to comp a table. Whether to run the special another night. None of it was hard. All of it came to me.

I spent a long time treating that as a manager quality problem. It was not. I had trained it. Early on, when a manager made a call I disagreed with, I overturned it, usually without much ceremony and sometimes in front of their team. Every one of those moments taught the whole management group a very rational lesson: checking with Willie is free, and deciding alone is risky. They were not weak. They were responding correctly to the incentives I had built.

What fixed it was not a motivational conversation. It was writing down, per role, which decisions belonged to that role outright, which needed a heads up after the fact, and which genuinely needed me before the fact. That third list turned out to be much shorter than my behaviour had been implying. A handful of items. Everything else was mine only because I had never said it was not.

The uncomfortable part was that once I published the list I had to honour it, including the calls I would have made differently. A decision right is worth less than a decision owned, at least until the owner has enough reps to be trusted, and they cannot get reps if you keep taking the bat.

The report nobody trusted

I want to sit on this one, because it is the mistake I see most often in other people's operations and it is the one I made most confidently. I wanted visibility across sites, so I built the reporting first. Weekly food cost by site, variance to theoretical, the whole thing. It looked serious. It was, functionally, fiction.

One site counted the walk-in on Sunday night, another on Monday morning after a delivery landed. Transfers between sites got recorded by whoever remembered. Two sites had different names for the same product from the same supplier. Waste was logged at one site and eaten at another. Every one of those is a small inaccuracy. Stacked, they moved the number by more than the operational differences I was trying to measure.

So when the report showed a site up on food cost, the site leader could explain it away in about ninety seconds, and honestly, they were often right. Once a number can be argued with, it stops being a management tool and becomes a debate topic. It took me far too long to accept that the sequence has to run the other way: agree definitions, fix the counting, standardise the product list, and only then build the report. A crude number everyone agrees on beats a sophisticated number everyone disputes.

Doing it in that order is what eventually made the reporting real for us. Definitions first, then MarketMan across all 14 sites so every site counts the same product the same way. Food cost now holds between 28 and 35 percent depending on the model, each one steady against its own line rather than one average hiding four different businesses. Food waste came down 14 percent over nine months and margin improved about 20 percent. The software was the last step, not the first, and that sequencing was the whole difference.

I have written the longer version of how I eventually got food cost to a number I could defend in reduce restaurant food cost, and the production side of it - what centralising actually solved and what it quietly made worse - in central kitchen operations.

What I would do differently

If I were opening location two again, I would not start with a manual. Manuals written before the second site exists are guesses. I would start by writing down, honestly, every thing at location one that only works because a specific named person is present. Not the job description. The actual list. Who decides the order guide. Who notices when the prep list is wrong. Who knows which supplier substitution is acceptable and which is not. That list is your real exposure, and it is usually five to fifteen items long, not a hundred.

Then I would open the second site deliberately understaffed on heroes and deliberately overinvested in the boring parts: one product list, one counting method, one prep sheet format, one shared definition of what done looks like at close. Boring, portable things travel. Talent does not travel, because talent stays in the building it is standing in.

I would also accept something I resisted for years, which is that the second site will do some things better than the first. My instinct was to enforce sameness downward from the original location. That was ego dressed up as consistency. The better move is to make differences visible and deliberate rather than quiet, so the good ones can spread and the bad ones can be caught. The site that quietly does it their own way is not the problem. The problem is that it is quiet.

And I would stop measuring managers on outcomes alone. Outcomes at a single site are badly contaminated by luck, location, and staffing. I would measure whether the work is controlled: is the count happening, is the handover happening, is the decision being made at the level it belongs to. Control leads outcomes, and it leads them by long enough to matter. If you only watch outcomes you find out too late to do anything about it.

None of this is complicated. It is just unglamorous, and it competes for attention with a hundred things that feel more urgent on any given Tuesday. That is the actual reason most groups stay stuck. Not ignorance. Sequence and attention.