Three of the largest restaurant operators in America spent nine years and billions of dollars answering the same break, every decision was correct, and all three operations are in worse condition than when they started — because a transactional architecture only stocks one class of answer, and none of them ever read the ceiling on the road they chose.
McDonald’s put $6 billion of kiosks and remodels into its United States restaurants beginning in 2018, splitting the cost 55 percent to the company and 45 percent to the franchisee, and is now retraining two million employees on service beginning October 5, 2026, in the largest training exercise in company history. Starbucks spent from 2022 forward on an equipment system built to let it run stores with fewer people, then reversed it and put labor back. Burger King committed roughly $700 million through 2028 and has decided to move its ordering kiosks along the walls instead of standing them in front of the lobby entrance. Every one of those moves was a repair. Not one of them raised the level of operation. The money is real, the competence is real, and the level of operation in 2026 is the level of operation in 2017 with more equipment in the building and a shorter applicant line at the door.
The Arithmetic Nobody Ran
Start with the number, because the number closes the argument before the reasoning begins.
McDonald’s announced a $6 billion investment with its franchisees over two years, Experience Of The Future, new decor and self-order kiosks. About 3,000 of 14,000 United States restaurants were converted, with 4,000 more scheduled that year. The company paid 55 percent of each remodel. The operator paid the rest. On a single unit that is the operator writing a check against a brand decision, financed, amortized, sitting on his balance sheet through 2026.
What was it bought against. Labor. Wages climbing, applicants thinning, turnover eating the schedule. That is the break, and it was a real break.
Here is the part that decides everything: the labor never came out. Nation’s Restaurant News reported flatly that McDonald’s was not saving on labor from the kiosks. CNN reported in 2024 that the kiosks never replaced the workers. RJ Hottovy put the mechanism in one sentence — the labor saved from kiosks is often reallocated to mobile ordering and delivery, and the kiosks “have created a restaurant within a restaurant.” The hours moved from the counter to the pickup shelf, the delivery staging area, and the person standing beside the kiosk helping a Guest who cannot find the substitution screen.
Then read the same finding from inside an operation, spoken by the operator who ran it. Brian Niccol, on Starbucks’ second-quarter call in April 2025: “In recent years, we have been decreasing labor within our stores, under the impression that equipment could compensate for the labor reduction.” And: “we’ve realized that this assumption did not hold true based on the outcomes we observed.”
So the ledger on the first repair reads: money out, labor unchanged, break still open.
Now the second entry. McDonald’s internal training document, 2026, describing its own restaurants: many Guests say their experiences “seem rushed, impersonal, sterile and cold.” That condition did not exist in that language in 2018. It is damage the repair produced. The counter position came out, the interaction that lived in the position came out with it, and a second operation grew inside the first one pulling the same cast in four directions.
Two purchases, one level of operation, and a ledger that shows spending where the operator believes it shows building.
There Are Only Two Work Classes, And The Answer Declares Its Class Before You Buy It
Every problem in a restaurant gets answered in one of exactly two classes of work, and the difference is not effort, cost, sophistication, or how well the answer performs.
[Repair Work] solves the problem inside the frame the problem lives in. The finished state is the state that existed before the break. Replace the broken plate. Change the dead bulb. Re-fire the wrong order. Find the rack-stack pattern that breaks plates every Tuesday and fix that instead — smarter, upstream, more thorough, and still repair, because it holds the assumption that plates are the correct delivery vehicle. [Innovation Work] exits the frame and leaves the operation standing higher than it was when the thing broke. Four moves: recognize the frame, question the frame, imagine the operation without it, build toward the new one.The question that sorts them is where the answer leaves the operation. At the state that existed before the breakage, or past the state it was standing at when the thing broke.
And here is the discipline almost nobody runs: that question gets asked before the answer is bought, not after the fix holds. The answer declares its class on inspection. A machine standing where a person stood preserves the frame completely — same counter, same transaction, same throughput logic, one fewer human in the sequence. It was readable as repair in 2018 by anyone who ran the question, in the vendor’s demo, before a purchase order existed.
Nobody ran it. Not because they were careless. Because the answer worked.
A Transactional Architecture Only Stocks Repair
This is the physics, and it is not a criticism of the three operators in this piece.
Road 1 capability is throughput, speed, cost, and conversion. That is method capability, and method transfers — which means Road 1 friction is genuinely reducible by purchase. A transactional operation buying a better system is doing exactly the right thing. The kiosk raises throughput and it raises the check. Shake Shack’s chief executive said out loud why the machine earns its place: it guarantees the upsell gets presented, which “is not always a priority for employees when you’ve got 40 people in line.” The machine never forgets to ask. It never has a bad shift.
So on a transactional road, every problem that arrives has a correct purchase attached to it. Labor pressure arrives, and there is a cheaper way to execute the same transaction sitting in a catalog. The purchase performs. Throughput goes up, the check goes up, the architecture rewards exactly what it was built to reward, and the counter resets to zero.
That is why the frame question never gets asked. Not suppression, not stupidity. There is no signal anywhere in the operation telling the operator he is on a loop, because every individual move on the loop is correct and every one of them pays.
And a transactional architecture does not ask what the operator believes. It asks what gets measured, scripted, and rewarded. If those three things are all extraction-oriented, the architecture is transactional regardless of what the operator intends. An operator can sincerely want a relationship with every Guest who walks in and still be running a building where the only things measured at the station are ticket times, average check, and labor as a percentage of sales.
Road 1 is a legitimate choice. At 14,000 units it may be the only coherent one available. Nobody is on trial for choosing it. The physics does not punish the choice.
Returning To Zero Is A Declining Position
Here is what accumulates while the repairs are being executed competently.
[Repairman Syndrome] is the operating disposition in which the operator runs the operation as an endless sequence of repairs, returning it to zero rather than building past it, with no architectural pass on why the fires keep starting. The distinguishing feature is not the volume of repair. The stream of things to fix never ends in a restaurant, and answering it is the job. The distinguishing feature is that nothing runs after the repair holds.The conundrum inside it is that competence makes it worse. An operator who is genuinely excellent at fixing things gets rewarded in the short term for precisely the behavior that traps him. Every skillful repair is more evidence — to him and to everyone around him — that repairing is what he does. His standing in the building is made of it.
Scale that to a brand and it is the same mechanism with a press release attached. Experience Of The Future. Reclaim The Flame. Make It Golden. Back To Starbucks. Each one announced, each one executed, each one resetting the counter, and the standing of the operation with its operators and its market is made of the announcing.
Then the physics that makes all of it expensive: returning to zero is not neutral. The same level of operation, held for years through competent restoration, is a declining position. Costs rise. The Guest’s alternatives improve. The cast’s tolerance shortens. And the operator experiences the decline as a hard market rather than as the accumulated result of his own read.
Nine years. Correct decisions throughout. Worse operation.
The Operator Who Reopened The Diagnosis And Landed In The Same Store Of Answers
Starbucks got further than anyone else in this piece and it is worth being precise about how far.
Niccol arrived in September 2024. The Siren system — equipment introduced in 2022 to speed drink production and offset fewer people in the store — was slated for broad rollout in October 2024. In January 2025 it was cut to the top quartile of stores by sales. By April 2025 it was limited to very targeted stores with high drive-through volume. In the same quarter he said the assumption had not held, and he put labor back: a pilot at five stores, then 1,500 to 2,000 stores by May, then about 3,000 by the end of 2025.
What that quarter cost while it was being figured out: United States comparable sales down 2 percent, transactions down 4 percent, average ticket up 3 percent, earnings per share of 41 cents against roughly 40 percent down year over year, margin down 590 basis points on the fifth consecutive quarter of decline.
He reopened the diagnosis. He named his own wrong assumption in public, which almost no operator at any scale does. He got the read right.
And the answer he reached for was to restore the staffing level that existed before the equipment went in. That is [Repair Work], correctly scoped, honestly chosen, and it returns the operation to the state it was standing in before the Siren build — which is the state that produced the labor decision in the first place.
He also did something the other two did not: he designed. The green apron service model was built deliberately, scoped, resourced, and rolled to more than 2,000 company-operated stores by the end of fiscal 2025. Service is a Road 1 output. It is executable, it is method capability, it transfers, and better-executed service is a real gain on a transactional road. He is buying what his architecture actually sells, by design rather than by reflex.
And he still caps. That is the beat that matters more than any number in this section. The best-executed, most honest, most deliberately designed version of this move still lands at service, because the architecture set the ceiling before the design work started. Design gets you the best available output on your road. It does not move you to the other road.
The One That Moved The Fixture
Burger King is the smallest move in the piece and the most legible, because the reasoning is visible in the money.
Reclaim The Flame launched in September 2022 at $400 million, later raised toward $700 million through 2028, with $550 million earmarked for Royal Reset — remodels, kitchen equipment, technology. The refresh was structured as $50 million from the brand against a comparable operator co-investment. Remodel funding runs through royalty credits, with greater funding available to operators who elect a higher royalty rate. The goal is 85 to 90 percent of roughly 7,000 restaurants in new image by 2028. Patrick Doyle put the total United States commitment above $2 billion and said franchisees would make billions of dollars of accelerated investments.
That is an operation mid-reinvestment with an image deadline and operator capital already committed. It needs every dollar of check the building can produce.
So read the decision. Burger King’s United States president says what is happening in his restaurants plainly: Guests want a friendly face, and “that’s evaporating in the fast food space.” The same company’s chief executive said in 2023 that kiosks eliminate stressful interactions at the counter. The fix is not to bring a person back to the point of ordering. It is to move the kiosk along the wall so the lobby reads warmer, keep the check lift the machine produces on every order, and staff a friendly face somewhere else in the room.
The machine stays because the machine earns. The feeling gets staffed back in beside it. Nothing about the frame moved. A fixture moved.
The Violation Is Not The Kiosk
Now the part that indicts, and it is not the equipment purchase. The kiosk was coherent. Buying throughput on a throughput road is correct, and on Road 1 it is the right move at the right time against a real break.
The violation is asking a transactional architecture for a relational output and putting a budget behind the request.
Read what the 2026 answer is actually chasing. McDonald’s internal document wants greetings that connect, authentic hospitality, an experience that does not read as sterile. Those are not throughput items. Those are outputs of a relational architecture, and no amount of capital moves an output from one road to the other.
And read what is being bought to go get it: a script for energetic greetings, pushed to two million people, timed against a queue. That is service, executed. Service can be scripted, measured, and trained, which is exactly why it is a Road 1 instrument. Hospitality is produced, and it cannot be scripted — which is why the existence of the script is the tell.
Notice where the only false statement in the entire nine-year sequence lives. Not in the operation. The operation is honest: it executes service, it always did, and it is about to execute service better. The false statement is in the announcement, where scripted service is called hospitality. The word is doing work the building cannot.
So the third repair is coherent too, on its own road. Service quality degraded when the counter position was removed, and training service back in is the correct Road 1 instrument for that. It will partially work, the way the kiosk partially worked, and it will hold for exactly as long as the training budget does — because nothing structural carries it once the budget stops.
Same law reaches all three operations from opposite entry points. Starbucks chose its road, designed inside it, and reached the ceiling. McDonald’s chose its road, defaulted rather than designed, and is buying an output the road does not carry. Burger King chose its road, protected the instrument that earns, and staffed the shortfall.
Three operations. Independent decisions. Company-operated, franchised, and both. Identical class of move. When three different operating structures under three different management teams produce the same answer, the answer was not a management decision. It was the architecture.
What The Missing Read Would Have Cost
Nothing. That is the hardest part of this to accept.
Every operator in this piece knew his numbers cold. Prime cost, ticket times, transaction counts, labor as a percentage of sales, unit-level cash flow by daypart. None of that was missing. What was missing was a read that costs no money and appears on no ledger: what does my architecture produce, and what will it never produce no matter what I spend.
Run that read in 2018 and the kiosk decision comes out the same way, with one difference that changes nine years. We are a throughput operation. Throughput is what we sell and what we are good at. The kiosk raises throughput and raises the check, so buy it. And it will not produce a Guest who is known, because nothing on this road does, so we are not going to spend the next decade expecting that and calling the shortfall a service problem.
Same purchase. Different expectation. The loop never starts, because the loop is powered by expecting a return the structure cannot pay.
The limitation half is the half that gets skipped, every time. An operator will read his architecture for what it enables all day long — that read flatters him and it funds proposals. Reading it for the ceiling means saying out loud, in a room with his own numbers in front of him, that there are outcomes his operation will never produce. Nobody wants that sentence. It is the only sentence that stops the loop.
And it sits upstream of design. Choice first, understanding second, design third. Skip the second and the design executes faithfully toward an output that was never available, which is 2,000 stores of excellent work running at a ceiling nobody named.
The Diagnostic
Four reads. Run them against your own building this week, on paper, with your own numbers.
The classification read. Take the last five things you bought for the operation over $1,000. Point-of-sale module, scheduling platform, new equipment, a consultant engagement, a remodel line item. For each one, write where it left the operation: at the state that existed before the problem appeared, or past the state it was standing at when the problem appeared. Do not grade the purchase on whether it worked. Repair works. That is not the question. If all five land in the first column, you are running one work class, and you have been for as long as you can remember.
The ceiling read. One sheet of paper, two lists. Outcomes my architecture produces. Outcomes my architecture will never produce at any spend. If the second list is empty, you have not run the read — you have run the flattering half. The read is not finished until something painful is written in the second column.
The expectation mismatch read. Go to your last three budget decisions and name the output each one was chasing. Then check the output against the second list. Any dollar chasing something on the second list is a permanent loss, and it is usually the line item you are proudest of.
The vocabulary read. Find every place your operation uses the word hospitality — training material, job description, the wall, the new-hire packet, the last thing you said at a pre-shift meeting. Ask what is actually being described. If it is a greeting, a phrase, a timing standard, or a step in a sequence, that is service, and it is fine. It is just not what you are calling it, and what you call it is what you will keep expecting.
The Closer
The three operators in this piece are not cautionary examples of bad judgment. They are the proof that judgment is not the variable. Six billion dollars, the deepest counsel available in this business, nine years of competent execution, and the loop ran anyway — because every answer the architecture had in stock was a repair, and nobody read the ceiling before they went shopping. If that is what happens with those resources, the independent’s inability to think past the next fix is not a resources problem, its an architectural one.
What You Do Monday Morning
Pick the single largest operating check you have written in the last three years. The equipment, the platform, the remodel, the build-out. Put the invoice on your desk.
Write two sentences on it. The break this was bought to close. Whether that break is closed today.
If the break is still open, the purchase was repair and it held — which is what repair does — and the item has been sitting open on your operation ever since with a paid invoice attached to it. That is the loop, visible, in your own handwriting, on one piece of paper. Then run the frame question on it, late but for the first time: does the broken part belong in this operation at all.
Digging Deeper
Positions on the record:
- 1,000 Restaurants Bought The Same Operation Last Month — https://hacksterism.jeffreysummers.com/1000-restaurants-bought-the-same-operation-last-month
- The Only Asset Your Competitor Cannot Buy This Afternoon — https://physics.jeffreysummers.com/the-only-asset-your-competitor-cannot-buy-this-afternoon
- You Did Not Build A Restaurant, You Rented A Trend — https://hacksterism.jeffreysummers.com/you-did-not-build-a-restaurant-you-rented-a-trend
- Demand You Create Is The Only Demand You Own — https://physics.jeffreysummers.com/demand-you-create-is-the-only-demand-you-own
- The Industry’s Editorial Class Just Endorsed A Case Study Reduction Of The Year — https://hacksterism.jeffreysummers.com/the-industrys-editorial-class-just-endorsed-a-case-study-reduction-of-the-year-nobody-ran-the-causal-read/
- Forms, Templates & Other Nonsense — https://jeffreysummers.com/docs/forms-templates-other-nonsense/
Terms used in this piece: Repair Work, Innovation Work, Repairman Syndrome, Repairman Conundrum, Transactional Duct Tape, Transactional Fix, Two Roads, Transactional Architecture, Relational Architecture, Customer Architecture, Guest Architecture, Restaurant Architecture, Reward Structure Architecture, Transactional Reward, Relational Reward, Incentive Recursion, Friction, Read Friction, By Design Or By Default, Default Gravity, Default Architecture, Designed Architecture, Architectural Coherence, No Static Achievement, The Law Of Forward Motion, Static Decline, Lost Opportunity Tax, The Hack Roster, Transactional Pushers, Sameness Machine, Measurement Lock-In, Cover Blindness, Visibility Trap, Operator’s Visibility Problem, The Operator’s Read, Causal Read, Uncertainty Capacity, Positioning Capital, Constraint Inheritance, Ordinary By Design, Value Congruence, Million Dollar Mediocrity, Relational Innovation, Likeability Trap, Structural Scale, Profit Foreclosure, The Vocabulary Theft, Guest Experience, Five Fundamentals. Definitions in the Knowledge Base: https://kb.jeffreysummers.com/
The architecture taught in full, fundamental by fundamental: https://physics.jeffreysummers.com/
The Road 1 arbitrage prosecuted where it lives in the wild: https://hacksterism.jeffreysummers.com/
How my thinking shapes the work: https://jeffreysummers.com/
