Where the productivity systems examples agree, and where they do not

Six named systems dominate operations improvement work, and they do not agree. Toyota's production system, Goldratt's theory of constraints, Deming's quality management, the Balanced Scorecard, Lean Six Sigma and OKRs each answer "where does productivity come from" differently.

Read side by side, they agree on measurement and disagree on the target. That split is the useful part, because it tells you which system to reach for.

What to take away

Where they agree

Every one of the six starts with a count. Toyota counts takt time and defects. Goldratt counts the constraint's throughput. Deming counts variation. The Balanced Scorecard counts lead and lag measures. Lean Six Sigma counts defects per million opportunities. OKRs count key results.

The agreement is narrower than it looks. They agree that a number must exist before a change can be judged. They do not agree on which number, or whose.

They also agree on one failure mode: a measure that becomes a target stops measuring. That is the same observation as the Hawthorne effect, and every one of the six has a story about it.

Where they disagree

The disagreement is about the unit of improvement.

Toyota assumes the process is the thing to fix and that operators know it best. Goldratt assumes the process is fine and one step is starving the rest. Those are opposite starting points, and choosing wrong wastes the whole effort.

Deming assumes the problem is in the system, not the person, and that most variation is common cause. The Balanced Scorecard assumes the problem is that nobody agreed what winning means. Lean Six Sigma assumes defects are countable and worth counting. OKRs assume the problem is too many priorities.

An example: the same trace, read six ways

The example below is a representative two-week trace for one operations role, typical of an office job with mixed work. It covers ten working days at 16 half-hour blocks a day, 80 blocks a week, across seven categories written down before the trace started.

Rework is 18 of the 160 blocks, or 11 percent. Blocked time spent waiting on an input is 16 blocks, or 10 percent. Internal meetings are 24 blocks, or 15 percent.

Core process work arrives in single half-hour blocks 22 times in the ten days. Its longest unbroken run is three blocks, or 90 minutes. Daily rework ranged from 0 to 3 blocks and daily blocked time from 0 to 4 blocks.

Toyota reads the rework line first, as a defect signal, and asks what standard the work departed from. Goldratt reads the blocked line and the single-block runs, and asks which step is starving the rest. Deming reads the spread across days and asks whether the variation is common or special cause.

The Balanced Scorecard reads the whole table against the stated strategy, and asks whether these categories are even the right ones. Lean Six Sigma reads the 11 percent rework as defects per opportunity and asks what the denominator is. OKRs read the meeting line against the quarter's key results and ask which meetings serve them.

Six readings, one table. The table did not change. The question did.

Which one to pick

If flow is visible and one step is clearly slower, Goldratt. If the same defect keeps appearing and nobody knows why, Deming. If rework is measurable and expensive, Lean Six Sigma. If the strategy is unclear but the work is fine, the Balanced Scorecard. If teams are pulling in different directions, OKRs. If the work itself is unstable, Toyota.

Running two at once is common and usually fails, because they assign the improvement to different owners. A constraint fix belongs to the constraint's owner. A variation fix belongs to whoever controls the system. Those are rarely the same person, and the change management question is who carries the change once it is agreed.

What the trace gives all six

Every system above needs a baseline, and the cheapest baseline in an office setting is a time trace. Two weeks, half-hour blocks, six or seven categories defined in writing before anyone starts.

The categories must be phrased as what the person is doing, not what it is for. "Answering questions from other teams" is a category. "Value-adding work" is not, because two people will file the same activity differently.

Where quality problems are cheapest to catch, and why the answer is almost never at the end, is in quality management. The rework line is where Deming and Lean Six Sigma overlap most closely.

Doing it for a team

Trace three or four people in the same role, same fortnight, same seven categories. The categories are core process work, rework, blocked waiting on an input, internal meetings, answering questions from other teams, admin and email, and breaks and gaps between tasks. The output worth reading is the variation between them, not the average.

Where two people in the same role have very different shapes, work is being routed unevenly, and that is a process finding. Where they all show the same fragmentation, it is structural and no individual change will move it.

Switching between tasks carries a real cost in resumption time, which is why fragmentation matters more than the total, and the evidence about human multitasking is consistent on that point.

Say before you start that individual sheets stay out of any assessment. How personal habits and team throughput interact is in productivity systems.

The flow measure that pairs with the trace is in productivity systems framework. Improvements that look right after a trace and make things worse are in productivity systems mistakes. For operational questions where no system holds the data, the counting technique itself is in operating processes metrics.

Common questions

Do these six systems actually conflict, or is that a presentation trick?

They conflict on the unit of improvement, which is not cosmetic. Toyota assigns the fix to the process owner and Goldratt assigns it to the constraint owner. When those are different people, running both at once produces two improvement plans competing for the same hours.

Can a small operation use more than one?

Yes, sequenced rather than simultaneous. Run one system until its answer is stable, then move. Two systems at once usually means two sets of measures and no decision.

What if none of the six fits?

Then the problem is probably not a productivity problem. If the constraint is a licence, a supplier or a customer, no improvement system will move it, and the work belongs in commercial or regulatory planning instead.

How long before a system shows a result?

Long enough that the baseline matters. Deming and Toyota both assume months, not weeks. If you need a result inside a quarter, OKRs or a constraint fix will show movement sooner, and both will show it less reliably.