On a wet Tuesday in April, branch three was turning work away by noon. Its drying teams were out, and the phones had not stopped since the river neighborhoods started calling. The manager had six techs. He needed nine.
That same afternoon, branch one, forty minutes away, sent two techs home early. Its jobs were dry, its next intake was a single bathroom loss, and there was not enough honest work to fill a shift.
Both branches were staffed correctly — on the annual plan.
What replaces that plan in a multi-branch shop whose demand is weather is not a better forecast. It is a shorter rhythm for reading demand as it arrives, and a transfer rule agreed in advance so crews move on the rule, not on who gets the owner on the phone first.
This is a typical shape, not one firm: four branches, about 85 people in all. The owner built the flagship himself. Branches two through four are run by managers who were his best techs first. That worked until his span did not.
The plan was built for a month that never arrives
Each branch got a fixed roster sized to an “average” month. Peaks were supposed to be absorbed by overtime and temps. Valleys were supposed to average out.
Restoration does not average out on a branch map. A wet spring floods branch three’s territory while branch one stays quiet. A January freeze triples intake for nine days, then February starves everybody. The failures happen at the same time, in opposite directions: one branch turning down work and burning out its crews, another sending people home because the work is somewhere else.
Transfers were possible. They were just not a system. Every move needed the owner’s arbitration, so the imbalance got resolved, when it got resolved, by whoever called him first or loudest.
The owner named it plainly: he was the load-balancing system, and he was tired. His orientation — where the pressure actually was — lived in his phone log, not on a board his managers could see together. Tempo was relative, branch to branch, week to week. His roster pretended it was fixed. That is Boyd’s mismatch in ordinary clothes: the plan inside no longer matched the weather outside.
What takt actually means
Toyota’s answer starts with takt time: available production time divided by customer demand. Takt is the beat at which you need to complete one unit to meet that demand — no faster, no slower.
The old comparison is a conductor’s beat. It is not maximum speed. It is not a quota to beat. It is the pace that lets the whole line finish together, in time with what was asked for.
That distinction matters. For Taiichi Ohno, overproduction was the worst waste, because making more than demand, sooner than demand, hides every other problem under inventory and busywork. The aim is to exactly meet demand. Staffing a branch for a storm happening in another branch’s territory is the same error in restoration form: people busy, or idle, in the wrong place — a pacing problem called a staffing problem.
Takt is also one part of a larger discipline. In Toyota’s standard work, it sits alongside the work sequence and the standard in-process stock that sequence needs. Change the beat, and the other two get looked at again. Standard work itself is not a manager’s poster. It is the current best way, owned by the people who do the work — held by SDCA until a small, operator-originated kaizen, tested through PDCA, earns a new standard.
Most importantly here, takt was never set once a year. It is recalculated whenever demand changes — in source accounts, a formal review roughly monthly, with adjustment checks about every 10 days. When the number moves, the whole line rebalances. Not one station. The whole line. That rebalancing is the transferable form.
A rhythm check instead of a roster
The four managers started standing together for 20 minutes every other Monday — the shop’s version of that roughly 10-day recalculation.
No forecast speeches. Each branch posts three things on one shared board — a small obeya in practice, one room, one board, one argument about the work instead of four about territory:
- Real intake over the last two weeks, not the plan’s intake.
- Jobs in drying right now, and when they are expected to clear.
- Crew-hours available over the next two weeks, after time off and committed work.
That is genchi genbutsu applied to staffing: go and see the actual intake, drying rooms, and hours before you move a person. The board replaced the owner’s phone log as the shared orientation, and the mismatch was visible early enough to act inside it.
The transfer rule that replaced the owner
The rule was agreed on a calm Monday, and written down where the board lives:
When a branch’s posted rhythm is peaking — intake and drying load above what its own crew-hours can cover for the next two weeks — it can claim a two-person drying team from the branch with the clearest slack. Managers do not call the owner to approve it. The board is the approval.
Three clauses made it work where “just help each other out” never had.
First, a return date is set the day the team leaves. A loan is not an open-ended draft.
Second, the lending branch gets first claim on the next transfer. Reciprocity is not a favor banked in memory. It is part of the rule, visible on the board.
Third, the shusa for this work — the one chief responsible for flow across branches, the owner at first, then the senior branch manager in rotation — owns the rule itself, not each move. If the rule produces a bad move, the group fixes the rule at the next rhythm check. That is the bargain: follow the standard, and you have a real way to change it.
Within a season, turned-down work dropped, because help arrived before a branch was full. Overtime concentrated where the weather actually was. The owner stopped being the transfer desk, and went back to asking where the system, not a manager, was stuck.
The cost, and the honest limit
The cost was structural, not motivational. A transfer rule has to override a manager’s sound instinct to hoard his best techs before his own storm. That instinct is not selfishness — he will be judged on his branch when the water rises there. Lending anyway means carrying real risk for the whole shop.
Twice in the first season, that risk landed. A lending branch got hit while its team was away — once by a freeze burst, once by a commercial loss nobody had posted because it had not signed yet. The work got covered, but late and hard. The rule survived only because the reciprocity clause paid back visibly, in the next transfer, in front of everyone at the board. If it had not, the rule would have become a poster by summer, and the calls to the owner would have started again.
And then the limit, stated plainly: takt precision does not transfer to restoration.
In a factory, demand can be counted as a predictable stream of near-identical units, and divided into a beat you can build a line around. Restoration demand arrives as weather, in lumps, in different mixes — a freeze is not nine copies of a kitchen leak, and a drying day on a small job is not interchangeable with one on a large loss. The signal is faint and lumpy where the factory’s is clear and countable.
So borrow the form, not the precision. The cadence transfers — read real demand often, rebalance the whole system when it moves, let a written rule carry the move. The clarity does not. A board that pretends its two-week rhythm is a takt calculation is lying to itself with better arithmetic. What it has is a shared, current picture and a fair way to act on it. That is enough to change the work.
There is no average month. There is only the next two weeks, seen together, and a rule for moving toward the weather.
The practice
- This Monday, kill the average for one decision. Post, for every branch, real intake for the last two weeks, jobs in drying now, and crew-hours available for the next two — on one board, in one meeting, before anyone staffs from the annual roster.
- Write the transfer rule before you need it. Name the unit that moves (start with a two-person drying team), the trigger that allows a claim, and the return date set on the day it leaves. No rule, no loan — otherwise you are back to arbitration.
- Put reciprocity in writing. The lending branch gets first claim on the next transfer, recorded on the same board. If that payback does not happen visibly, expect the rule to fail.
- Recheck on a rhythm, not in a crisis. Twenty minutes every other Monday is enough to start. If the board shows the old pattern — one branch drowning, one idle, owner in the middle — treat that as the system’s andon: stop, and fix the rule together, not the loudest manager.
From the Boyd & Toyota series at Restoration Intel — what two old disciplines, tempo and standard work, look like inside a restoration shop.