How an Adherence Percentage Becomes an Investigation List

An adherence percentage tells you the plan is not being followed and gives you nothing to do about it. The count of operations behind that percentage is a finite list you can actually review, and pairing adherence with attainment tells you whether your problem is a floor that is not following the plan or a plan that was never right. EDGEBIC by User Solutions shows both, side by side, with the count printed beside the percentage.

This post covers one return: turning a metric people have stopped reading into a week's worth of specific work. It sits under the EDGEBIC results guide and extends late versus overdue, which makes the same distinction at the job level.

Why Percentages Go Stale

Almost every plant that measures schedule adherence eventually stops acting on it. The mechanism of that decay is predictable.

The number arrives as a percentage. Somebody asks whether it is good. Nobody can say, because there is no obvious ceiling for a shop with real disruptions. The discussion becomes an argument about whether the target is fair, which cannot be resolved, so it recurs monthly until people quietly stop bringing it. The metric stays on the dashboard as decoration.

The problem is not the metric. It is the format. A percentage is a judgment, and judgments invite debate. A count is a work list, and work lists invite work.

The Mechanism: The Delta Note Is the Point

On the planner view, adherence appears as a tile: operations that started within roughly an hour either side of their planned start, over the past fourteen days, against a target of ninety percent or better. Beside it sits a delta note stating how many operations moved off plan.

That note is not decoration. It is the investigation count, and it is the reason the tile is actionable while the percentage alone is not.

Consider the difference in how the two framings land in a meeting.

The second framing has an end. You can review ten operations in half an hour, and what comes back is a set of causes rather than a debate.

Next to adherence sits attainment: completed jobs that hit their due date, against a target of eighty-five percent or better. That pairing is the second half of the mechanism, and it is what makes the tile diagnostic rather than merely accusatory.

Reading the Pair: Four Situations, Four Responses

Adherence and attainment can disagree, and every combination means something different. This is the most useful table on the dashboard, and almost nobody draws it out.

The row that surprises people is the second. Hitting dates while adherence is poor feels like a success and is actually the most fragile state on the table, because delivery is being produced by something other than the plan: slack in the dates, supervisors making good calls in their heads, or overtime absorbing the difference. It works until the person carrying it is away or the load rises, and then it fails without warning. It is also invisible in any delivery report, which is exactly why it persists for years.

The third row is equally important and gets misdiagnosed constantly. If the floor is following the plan and dates are still missed, no amount of accountability pressure helps, because the people being pressured are doing what they were asked. That is a signal to go and audit the plan's inputs: routing hours, capacity, and the constraint.

The Biggest Bucket Is Usually Not Indiscipline

When people first work an adherence investigation list, they expect to find operations that were skipped, delayed, or resequenced by somebody ignoring the schedule. Some are. But because adherence is measured on start times, it is only as good as the capture of those start times.

An operation that ran precisely when planned and was never punched in reads as off plan. So does one where the punch happened at the end of the shift, from memory. Neither is a discipline problem in the sense people assume, and both are cheap to fix compared with changing behavior.

This is the same effect that makes an Andon board read red or idle across the board by mid-morning: the logic that fires when an operation is past its planned start with no actual start is working correctly, and what it is detecting is missing capture rather than missing work. Kiosk actuals closing the planning loop is upstream of any adherence number being meaningful.

Sort your investigation list into buckets before you conclude anything:

1. Ran on time, never captured. A capture fix, usually the largest bucket at first.
2. Ran late for a named external reason. Material, a machine, an operator elsewhere. Real, and countable.
3. Resequenced deliberately by a person. Often a good call, sometimes a call the plan should have made.
4. No explanation available. The genuinely concerning bucket, and typically the smallest.

Only bucket four is what most plants assume the entire number represents.

The Causal Chain to Something Worth Money

1. You read the count, not the percentage. The list is finite.
2. You review a sample and bucket the causes. An hour of work.
3. The buckets have different owners and different costs. Capture, materials, maintenance, planning.
4. You fix the largest bucket first. Which is usually capture, which is usually cheap.
5. The count falls, and the remaining list becomes readable. Now the real deviations are visible instead of buried.
6. The plan becomes trustworthy enough to be used. At which point adherence stops being a scorecard and starts being a control loop, and the second row of the table above stops describing you.

Step six is the actual return. A plan that people follow is a plan whose dates mean something to sales, whose capacity numbers mean something to management, and whose disruptions can be contained by a reschedule instead of absorbed by individuals.

Measuring It in Your Own Plant

1. Record adherence and attainment weekly, together. One line each. Never one without the other.
2. Record the count alongside them. The percentage moves with volume; the count is what you will work.
3. Place yourself on the four-way table. Which row describes you this month? That determines everything else you do.
4. Sample ten off-plan operations and bucket them. Use the four buckets above. Record the split.
5. Estimate the capture share. If a large fraction ran fine and was not punched, you have a data project rather than a behavior project, and it is far cheaper.
6. Watch the trend, not the tick. A number at 88 percent once is noise; 88 percent falling week over week is a process problem. Three points of movement in one week means nothing.
7. Value the second row. If you are hitting dates with poor adherence, estimate the overtime and expediting that is absorbing the difference. That figure is what the fragility currently costs you, and it is the business case for making the plan real.

For documented outcomes rather than typical ones, the User Solutions and RMDB lineage includes GE Railcar moving from 30 percent to 90 percent on-time delivery, and Homestead Furniture cutting schedule build from 40 hours to 2. Those belong to their engagements and are quoted as heritage, not as forecasts.

Where Adherence Misleads

Five limits worth stating.

It is a start-time metric, so it inherits your capture quality. Without reliable punch-in, the number measures your data collection more than your execution. Fix capture before drawing conclusions about behavior.

A loose plan produces flattering adherence. If the schedule is vague enough that almost any start satisfies it, adherence will look excellent and mean nothing. High adherence is only evidence of control when the plan is tight enough to be violable.

High adherence is not the goal. A trustworthy plan is. Sometimes deviating is correct, and a good deviation should appear in the count and be explainable in a sentence. The target is not zero deviations, it is zero unexplainable ones.

It says nothing about whether the plan was worth following. A plan that faithfully executes the wrong sequence will post excellent adherence. That is the third row of the table, and it is why the pair matters.

It does not identify causes. The count hands you a list. The reasons come from the reschedule trail, the swap audit, the kiosk pause reason codes, or a conversation at the machine. Anyone promising causes from a percentage is guessing.

The takeaway

Adherence dies as a metric when it is reported as a percentage, because a percentage produces an argument about targets and nothing else. Reported as a count of operations that moved off plan, it produces a finite review that yields buckets you can fix, and the largest bucket is usually missing capture rather than missing discipline. Then pair it with attainment, because following the plan and missing dates, and ignoring the plan and hitting them, are opposite problems that look identical from either metric alone. To see both tiles against your own schedule, book a demo of EDGEBIC, and if you are on the older platform, the move from RMDB to EDGEBIC brings the planner view with it. Read this next to getting your team to trust the schedule and how EDGEBIC reduces schedule churn.