Productivity systems framework reviewed for 2027

A framework is a set of questions asked in a fixed order. The six below ask different ones, which is why picking between them matters more than picking the newest.

What to take away

The six frameworks, and what each is for

Theory of constraints (Eliyahu Goldratt, The Goal, 1984). Find the one step that limits throughput, protect it, then re-find it once you have moved it. Suits job shops, machine-limited plants and any operation with an obvious bottleneck. No licence fee; the cost is the discipline of not optimizing non-bottlenecks.

Value-stream mapping (originated at Toyota, documented by Mike Rother and John Shook in Learning to See, 1999). Draw every step, mark which add value, and count the waiting between them. Suits mixed office and production flows where nobody agrees on where the delay lives. Free method, needs a facilitator and a wall.

Lean (Toyota Production System; Womack, Jones and Roos, The Machine That Changed the World, 1990). Remove waste in seven named categories, with pull and flow as the goal. Suits repetitive work with stable demand. Training is widely available; certification bodies charge for credentials.

Six Sigma (Motorola, 1986; popularized by General Electric). Reduce variation with DMAIC and statistical process control. Suits high-volume processes where defects carry real cost. Belt training and certification are commercial, priced by the provider.

OKRs (Intel under Andy Grove, spread by John Doerr). Set a few measurable objectives and check them quarterly. Suits knowledge work where output is hard to count. Free to adopt; the cost is the quarterly review time.

Eisenhower matrix (attributed to Dwight Eisenhower, popularized by Stephen Covey). Sort work by urgency and importance, then delegate or drop. Suits individual workload triage, not process design. Free.

The distinction between time when something is being worked on and time when it is merely in the system is the central idea behind value-stream mapping, which came out of manufacturing and transfers to office work with almost nothing lost.

How they compare

Tracing five items before you choose

Pick five items that recently finished, chosen by a rule rather than memory, such as the five most recent. For each, reconstruct arrival and completion, which is lead time, and the minutes anyone spent working on it, which you get by asking the people involved. Rough is fine; you want a ratio, not a figure.

Illustrative shape, not a benchmark: an item with a lead time of eleven working days and a touch time of about four hours. That ratio, roughly one to twenty, is ordinary in office work, and it is why asking people to work faster changes almost nothing.

Count only working time, not elapsed time inside a person's day, or the ratio flatters you. Do not choose the five after seeing the data, because you will pick the ones that prove your point.

The four kinds of waiting

Every gap between arrival and completion falls into one of four. Mixing them is why generic advice does not help.

Mark each gap in your five traces with one of the four. The distribution is the useful output, and it is rarely what people predicted.

What to do about each

Read the right column twice. Each entry is the intuitive response, each is easier to get approved than the move beside it, and each makes the waiting it addresses worse.

Where each framework stops

Theory of constraints hides the next bottleneck once you fix the first. Value-stream maps go stale within a quarter if nobody owns them. Lean cuts capacity that variable demand still needs.

Six Sigma produces statistics nobody acts on when no process owner is named. OKRs decay into a long list with no single owner per number.

The matrix is a personal tool and collapses as a team ritual.

None of them measures whether the work was worth doing. A fast pipeline for reports nobody reads is worse than a slow one, and no flow measure will say so.

Where work varies enormously in size, trace within a band, or the ratio averages two different operations into one meaningless number.

Any countable measure can be gamed by splitting items so each finishes faster, which is a reason to keep it off individual scorecards, not to skip measuring.

The specific ways an improvement of this kind backfires are set out in productivity systems mistakes. The wider view of personal and team systems is in productivity systems.

The general rules for building an operational measure that survives contact with the people it measures are in operating processes metrics.

Where the fix turns out to be structural rather than local, it becomes a change to how the work is done, and that is treated in change management.

Common questions

How long does the five-item trace take?

About two hours, most of it spent asking people how long things took. It is the cheapest diagnostic in this subject and the one least often run before a framework is bought.

Our lead times vary wildly. Is the ratio meaningless?

The average is; the individual traces are not. Read the five separately and look at where the gaps sit. A wide spread usually means some items hit a queue and others do not, which points at a routing decision.

Should we publish lead time as a target?

Publish it as a shared number for the work, not as a target attached to people. As a personal target it produces item splitting and cherry-picking within a quarter, and you lose the diagnostic value.

What if all four kinds of waiting show up equally?

Start with decisions, because that fix is usually free. Naming who decides and by when takes a conversation rather than a project, and in most traces it is the largest single block anyway.