Production Reporting Metrics for a PCB Shop

Production reporting is the practice of putting a small number of numbers in front of the people who can change them, often enough that the numbers influence what they do next. Most shops report too much and review too rarely, which produces a monthly document that nobody uses. The alternative is a short list of metrics at the area level, updated daily, plus a short review that ends with actions, and it usually improves performance before any process change has been made.

Why Most Reporting Fails

Reporting fails when the numbers are aggregated above the level at which they can be acted on. A monthly plant wide scrap figure describes the outcome of thousands of decisions and gives no one person a reason to change anything. The same scrap data shown by area, and by shift, gives the people running those areas something they can influence before the end of the week, and that is the whole point of reporting.

It also fails when the number is used for judgement rather than improvement. Once a metric is understood to be a performance assessment of individuals, the data collection starts to bend: borderline rejects are reclassified, delays are reported as something else and the metric gradually becomes a work of interpretation. Shops that keep the reporting at the level of the process rather than the person avoid most of this.

Choosing the Metrics

A useful set of key performance indicators is small enough to be read in a minute. Safety, delivery against the original promise, first pass yield, scrap cost, on time completion of the schedule and work in progress are enough for most board shops, and each one connects to a decision that someone can make. Adding more metrics dilutes attention rather than improving control.

Each metric should also have an owner and a defined reaction. A figure without an owner is a fact, and a fact that nobody must respond to will be tolerated at any value. When the delivery metric slips, the reaction should be a specific action by a specific person, and if no such action exists, the metric is not yet ready to be reported.

Data Collection That Survives

Data collection fails when it is slower or more complicated than the work it describes. A shop that requires a terminal entry at every step usually ends up with partial and delayed data, while a shop that records a few figures at the point where a decision is made gets data that is timely and reasonably accurate. The design question is always what decision the data supports.

Automatic collection is preferable where it exists. Machine counters, timestamps from a test system and the inspection results from an AOI platform all provide data without any additional operator effort, and they are usually more reliable than manual records because they cannot be forgotten. Manual entry should be reserved for the information that only a person knows, such as the reason a lot was held.

Reporting Frequency and Format

Frequency should match the speed at which the metric can change. Production figures are useful daily, quality trends weekly, and cost figures monthly, because that is the interval over which each one carries a signal rather than noise. Reporting a noisy number daily produces a series of false alarms and eventually indifference.

The format matters more than most shops expect. A single sheet with the current value, the target and a simple trend is far more effective than a detailed table, because it can be read at a glance and compared with yesterday. The same sheet should show the open actions from the previous review, which is what keeps the reporting connected to what actually changed.

The Management Review

A management review is useful when it is short, regular and focused on exceptions. An hour a week with the area figures, the missed targets and the actions from last week will produce more improvement than a quarterly meeting with a long presentation. The purpose is not to admire the data but to decide what to do about the parts that are not where they should be.

The review should include the people who run the processes, at least for the items that concern them. A review conducted only by managers produces actions that are imposed rather than owned, and imposed actions are the ones that quietly stop happening once the attention moves elsewhere. Including the operators also improves the data, because people report more accurately when they can see how the figure is used.

Avoiding the Traps

The common traps are well known and easy to fall into. Reporting a metric that the shop cannot influence, changing the definition of a metric part way through a year, and using a single figure to represent several different things are the three that cause the most damage. Each of them destroys comparability, and comparability is what makes a trend meaningful.

Another trap is optimising one metric at the expense of another. Delivering everything on time by shipping work that should have been held will improve the delivery figure and damage the quality figure a month later. Reporting a small balanced set together, rather than one number at a time, makes that kind of trade visible while it is still a choice rather than a discovery.

Practical Rules

Choose a handful of metrics with an owner and a defined reaction for each, collect the data where the decision is made, and report at the frequency at which each number carries a signal. Keep the definitions stable and record them with the production records.

Review weekly, briefly and with the people who do the work, and close the loop on the actions from the previous review before adding new ones. Production reporting works when it changes what happens on the floor, and every metric that does not change behaviour should be removed from the report rather than carried indefinitely out of habit.

Process Control and Verification

Where a value sits close to a process limit, the drawing should say so, since the shop can then open the process window rather than working to a nominal figure that carries no tolerance.

FAQ

How many metrics should a shop report? Five or six at the area level is usually enough. Each should have an owner and a defined reaction when it moves.

How often should they be reviewed? Production figures daily, quality weekly and cost monthly, with a short weekly review that closes the previous actions.

Why do metrics get manipulated? Usually because they are used to judge people rather than to understand processes. Keeping the reporting at process level avoids most of it.