Inaccurate Record Filling, Don't Blame the Employees — Five Steps for Designing Quality Record Forms

1. Three Records from the Same Process, Three Different Filling Methods

A metal stamping company was audited by a customer last year. The auditor randomly selected three "First Article and Patrol Inspection Record Forms" from the stamping workshop: August 12, September 3, and September 18, all for the same part on a 300T stamping machine. The first form had "OK" in the dimension column; the second had "3.98"; the third had "合格，见图" (qualified, see diagram).

The auditor asked the workshop director, who was accompanying the audit: What should be filled in this column? The director was taken aback and said, "Numbers, I guess." The auditor then asked: If numbers are to be filled, are the units in millimeters or inches? What is the judgment criterion? The director couldn't answer and called over the IPQC team leader. The team leader explained: We've always filled it this way; the experienced workers understand it.

The audit conclusion included this sentence: This record cannot prove that the process was monitored and measured according to the control plan, and it is judged as lacking in record validity.

After the audit, the quality manager reviewed the situation and realized that the issue wasn't with the employees' attitude. The form hadn't been revised since the factory was established in 2016: the dimension column lacked units, specification limits, a judgment column, and filling instructions. Out of 25 fields, 6 had never been filled since the factory opened. The "备注" (notes) column was commonly used as a catch-all, with the most frequent phrase being "正常" (normal). Three inspectors filled the form in three different ways, essentially because the form had never been designed to specify a "correct answer."

Most complaints about "inaccurate record filling" ultimately boil down to the same issue: the form itself was never designed; it was merely drawn up.

Rectifying the issue wasn't easy. The customer required supplementary process capability data for the past three months. The quality department retrieved the records to prepare for the statistics but stopped halfway through: the dimension column contained "OK," "3.98," and "见图" (see diagram), and less than 60% of the entries could be directly used for statistical analysis. The remaining entries had to be re-measured. A two-hour audit resulted in three weeks of rework—during which the production line continued to use the same form to produce new "unusable data."

2. Forms Are Data Interfaces, Not Administrative Stationery

In the context of ISO 9001:2015, records are considered documented information. Clause 7.5.2 requires "appropriate identification and description, format and medium" when creating and updating documented information, and clause 7.5.3 requires control of documented information to ensure it is "available and suitable for use" when needed. These requirements, though abstract, boil down to one simple statement on the shop floor: forms are the data interfaces between the system and the site, determining whether data can be collected and whether the collected data can be used.

Understanding forms as data interfaces provides a unified criterion for many debates. A qualified interface should have four elements:

Field Definitions. Field names, data types, units, and allowed value ranges. When designing a form, ask yourself: Should this cell contain a number or text? What is the unit—mm or ℃? What is the range—0 to 100 or only "合格/不合格/让步" (conforming/nonconforming/conditional acceptance)? Without definitions, there is no data, only text.

For the "尺寸" (dimension) column, for example, the term has been used for a decade without anyone considering what it lacks. In reality, it hides at least five undefined items: which dimension on the drawing it refers to, the unit, the number of decimal places corresponding to the measuring tool's resolution, the judgment limits, and how to mark deviations. Any missing item makes the data incomparable—records from the same day cannot be directly summarized between two different teams, let alone compared with data from three months ago to identify trends.

Judgment Criteria. The basis for determining "合格/不合格" (conforming/nonconforming) must be in the same cell or clearly and uniquely referenced on the drawing or control plan. If the judgment criteria are out of sight, the inspector relies on memory and habit, which inevitably differ from person to person.

Responsibility and Timing. Who fills in which field and when, and who reviews it. The confirmation of the first article and the recording of patrol inspections have entirely different responsible persons and timing. Mixing them on one form leads to overlap and confusion.

Controlled Status. Forms are a type of controlled document, with a number, version, effective date, usage period, and回收要求 (recycling requirements). If the form version is out of control, the data口径 (caliber) is also out of control: A/2 version records dimensions, A/4 version records torque, and the two sets of data cannot be combined for trend analysis.

From a data flow perspective, forms also determine how far two things can go. The first is traceability: when a nonconforming product occurs, can the relevant process parameters and inspection data be traced back within 30 minutes? The second is analysis: whether the fields are standardized directly affects whether these records can be summarized into trends and whether they can be entered into SPC or a database. Many companies get stuck in their first year of digitalization, not because the system is inadequate, but because they realize that the same column contains both numbers and text, as well as "正常" (normal), "OK," and "一般" (average)—the system can neither calculate nor compare these entries.

3. Five-Step Method: Designing Forms as Interfaces

Step One: Clearly State Who the Form Serves and What Questions It Answers

Before designing, list the "data consumers" of the form: operators use it to confirm startup conditions, IPQC uses it to release products, engineers use it to check parameter drift, and customers and auditors use it to prove process control. If trend analysis is required, quality engineers and the system are also included. Different consumers have different key fields, and mixing them on one form leads to infinite expansion.

A simple operational method: write a one-sentence purpose statement for the form and place it in the top-left corner of the header. For example, "This form is used for the first article confirmation and every two-hour patrol inspection process judgment and traceability on the 300T stamping machine." Once the purpose is clear, the next step of cutting fields has a basis—any field that does not serve this purpose should be moved to an appendix or the system, not the main form.

There are two old rules worth adhering to during design. One is one row per record: each row should correspond to one measured object (one part, one device, one time point). Do not cram data from three time points into one row, as this will require manual decomposition during summarization. The other is one-time entry: the same data should not be written on a paper form and then copied into another ledger or Excel. Once discrepancies appear between the two records, the cost of tracing them back far exceeds the initial cost of designing a standardized interface.

Step Two: List and Define Fields

Do not draw the table directly. First, list the fields and define them one by one, then format the table. The list should include at least five items: field name, data type, unit, value range, and whether it is mandatory. Here is a comparison example.

The value of listing fields is to expose problems. Many forms have been used for a decade without anyone ever asking, "What exactly should be filled in this cell?" Going through the list usually results in cutting 30% of the fields and adding 20% of the definitions.

Step Three: Place Judgment Criteria in the Form

There are two ways to position judgment criteria. One is to print the specification limits and tolerances in small text next to the measurement cell, so the inspector doesn't need to flip through files. The other is to reference a uniquely numbered drawing or control plan item and specify in the header, "Judgment criteria see XX control plan version X page X." Both methods are acceptable, but mixing them is not.

For projects judged by sensory methods (appearance, feel, assembly interference), the judgment criteria are not numbers but limit samples. The form should include a column for "limit sample number and validity period," anchoring the judgment to a physical sample rather than the experienced worker's eyes. After adding this column to the connector appearance inspection form at an electronics company, the consistency of judgments by three inspectors on the same batch of materials significantly improved because they were all comparing against the same sample, not their individual impressions.

Another technical detail often overlooked is that the number of decimal places should match the resolution of the measuring tool. If a caliper with a resolution of 0.01 mm is used, the record column should not contain values like "12.0034"—the extra two decimal places are not precision but noise, and they can mislead downstream personnel into thinking the measurement system can identify micrometer-level differences. Printing an example format directly in the form (e.g., "_. mm") not only standardizes the number of decimal places but also reminds the person filling the form which measuring tool to use.

Step Four: Write Filling Instructions and Responsibility Matrix

Filling instructions are usually brief, no more than five lines, and printed in the header or footer of the form. The key points are three: when to fill (timing), who fills (role), and who reviews (signature). Additionally, add a "responsibility row" in the form, using symbols to mark O (fill), R (review), and — (not applicable).

This step, though seemingly trivial, is crucial for eliminating "three different filling methods." After a stamping company revised its first article confirmation form, it added three lines of instructions in the header: "Fill the actual measurement in the dimension column, unit mm, retain two decimal places; mark '×' for deviation and note the disposition number in the remarks; blank cells are considered uninspected." In a three-month record check, the proportion of blank mandatory fields dropped from 18% to 2%, and the number of non-numeric entries in the dimension column (such as "OK" and "正常") decreased from over 40 per month to single digits.

Step Five: Control,回收 (Recycling), and Iterative Review

Forms should have independent numbers and versions, with clear usage periods and old version回收 (recycling) methods. A common mistake is to update forms along with procedure documents: the procedure document has been revised three times, but the form still has the same version number, and the on-site copies are from three years ago.

Iterative reviews can be conducted quarterly, focusing on three aspects: field usage rate, fields that have not been filled for two consecutive quarters should either be deleted or automatically collected by the system; abnormal code distribution, if a particular abnormal code is consistently zero, it may not be that the issue never occurs but that the person filling the form doesn't know how to use it; record traceability test, randomly select three records and try to trace the parameters back to the equipment, batch, and nonconforming product disposition number. If the traceability is successful, it indicates that the interface is functional.

Transformation Effect Comparison

The five steps were fully implemented on a 300T stamping line at a metal stamping company. Records from three months before and after the transformation were compared as follows.

It's worth noting that the personnel configuration, measuring tools, and equipment on this line did not change; all changes were in the form itself and the filling instructions. This explains why measures such as "strengthening training and emphasizing the importance of records" often have limited effectiveness in many record issues—training can change whether people are willing to fill in the form seriously, but it cannot change the fact that they don't know how to fill it correctly.

4. Five Common Pitfalls

Pitfall One: One Form for All. To "unify management," all first articles, patrol inspections, final inspections, and equipment checks are crammed into one form. The result is that each cell is never fully filled, and each time point requires additional copying. The correct approach is to split the form according to data consumers, using the same numbering system to link multiple forms.

Pitfall Two: Only Adding, Never Reducing. When a new issue arises, a new column is added to the form. Over a few years, the form expands to over 20 columns, requiring two A4 sheets for horizontal printing. The more fields there are, the lower the filling quality—people's attention is limited, and beyond the 3 to 5 minutes it takes to fill out the form, the remaining cells start to contain "正常" (normal) and "OK."

Pitfall Three: Judgment Criteria "Elsewhere." Writing the judgment criteria in the control plan is correct, but requiring inspectors to flip through files while filling out the form means the judgment is left to memory. The criteria should either be printed in the form or clearly referenced in the header.

Pitfall Four: Using Checkmarks Instead of Data. Setting only "合格/不合格" (conforming/nonconforming) checkmarks collects no numerical data. This is a short-term convenience but a long-term abandonment of process capability analysis and trend judgment. A layered approach can be used: routine patrol inspections use checkmarks, key dimensions and special characteristics use numerical entries, and the two types of fields coexist but must be clearly distinguished.

Pitfall Five: Treating Form Design as a Clerical Task. Form design requires understanding of processes, judgment logic, and data usage, which falls under the responsibilities of process and quality engineers. Handing the design to someone unfamiliar with the site will result in a form that looks good on paper but is difficult to fill out on the shop floor.

It's important to note that these five pitfalls are mostly not issues of responsibility but of design oversight: if forms are not designed as "interfaces," they can only be patched up on the shop floor based on experience, which is always unevenly distributed.

5. Summary

The lower limit of record authenticity is not the employees' awareness but the standardization of form design—nailing down the five aspects of fields, units, criteria, responsibility, and version during the design phase ensures that the data collected is usable and traceable.

The upper limit of record quality depends on the precision of form design.

Knowledge code: 2.3.1

Version: v20260919

Author: QTank QTank is dedicated to providing systematic professional knowledge, methodologies, and practical tools for quality management practitioners to help companies continuously improve their quality capabilities.