Decided Work And Untested Work Belong In Different Columns

The argument is never about whether the change works

A plant leader can spend a year saying there is no opportunity in his building, read the improvement plan, and tell you every item on it was already planned. Both of those statements can be defended. Equipment really was on order. Positions really were slated to move when it landed. What the plan failed to record is which lines were already decided, on what date, and what triggered the decision.

That is not a politics problem. It is a plan structure problem, and it is the one you can fix before the plan leaves your desk.

The fix is a single rule applied to every line: the status of a line is the event that will change it.

Two states, one column called open

Open an improvement funnel on a mid-market food or beverage site and you will usually find two different things stacked in the same list.

The first is decided and funded. An ergonomic lift table is in fabrication. Capital was approved. Two positions are slated to come out of that area once the table is installed. There is nothing to test here. The only thing standing between this line and the result is a delivery date.

The second is a hypothesis. An area runs with a given crew size today, the plant data and the floor sampling together say it can run with fewer, and nobody has yet run it that way. The only thing standing between this line and the result is a test with a pass condition.

Both lines sit under the heading open. Both carry a number. So the plan reports one total built from two incompatible kinds of certainty, and that error propagates in two directions.

Downward, nobody can sequence the work. A decided line needs a milestone watched and a handoff planned for the week the equipment arrives. A hypothesis needs a trial designed, a step size chosen, and a supervisor told in advance what a good result reads like. Those are different jobs with different owners, and a flag buried in a spreadsheet column does not separate them.

Upward, the decided lines are the easiest in the whole plan to reclassify as work that was always coming. That reclassification is not dishonest. It is true. The decision was made. The plan simply never recorded whether the decision came before the finding or after it, so the question has no answer in the document, and it gets settled by whoever is more insistent in the room.

Write the state, then write what changes it

The practical sequence is short, and all of it happens before anyone presents a number.

Split the funnel into two visible groups, decided and under test. Two groups, not a flag in a column. A reader should be able to see the split without filtering anything.

For each decided line, record the date the decision was made and the artifact that proves it: capital approved, purchase order issued, equipment in fabrication, position slated to move. The status field then reads as a delivery milestone. A line whose status is a percentage has no status.

For each line under test, write the pass condition and the observation window before the trial starts. Stage the change one step at a time rather than moving to the target in a single jump. A staged test gets scored by the plant on its own floor, in its own numbers, which removes the argument about whose measurement to believe. It also gives you a stopping point that is a result rather than an opinion.

Take the plant's own stated current state as the baseline, not the one your stopwatch produced. A baseline you measured invites a conversation about your methodology. A baseline the plant publishes out of its own system of record invites a conversation about the gap, which is the conversation you want. Your measurement still matters; it sets the working target, not the starting line.

Then keep the states separate all the way through, because they are separate things. Equipment ordered is not equipment installed. Equipment installed is not a position that stopped being filled. A position that stopped being filled for one week is not a position that stopped being filled. Collapse those states into one word and you will either count a benefit twice or lose one you earned.

What a well-run improvement plan reads like

Every line has one named owner and one state drawn from a fixed set: decided, installed, verified, sustained. Every decided line shows a delivery milestone and the date its decision was made. Every line under test shows a written pass condition and the observation window that will settle it. Every baseline cites the plant's own system of record by name. Each state change carries the date it changed. No line reports a percentage as a status, and no benefit appears under two states at once. A reader who has never seen the site can tell in one pass which lines are waiting on a truck and which are waiting on a result.

The test to run this week

Pull your current improvement or savings plan and take the three largest lines. For each one, name the single event that will move it from open to captured.

If that event is a date, the line is decided, and the only open work is delivery. Stop reviewing it as though it might fail, and record now, while it is uncontested, that it was decided and when.

If that event is a result, the line is a test, and it needs a pass condition in writing before the trial runs. Choose the step size and the observation window this week.

If you cannot name the event at all, the line is not an opportunity yet. It is a topic, and it is inflating a number that someone will have to defend in front of a CFO.