In this guide 6 sections
Results drift away from batches at handoffs such as sample bottles, notebooks, instrument exports and spreadsheets, so trace released lots backwards through each one to find the break. Fix it at the source. Give every sample its batch number the moment it is drawn and store each test against that number instead of a date.
Trace five released lots to find the break
Pick a batch that shipped last week and gather the assay printout, the viscometer reading, the pH log and the certificate of analysis for it. Then check whether each document carries the same batch number, the same sample time, the same analyst initials and the same tank or reactor reference.
Mismatches follow patterns. Viscosity may carry its own number. The pH may belong to the tank, not the batch, and the pattern points at the handoff that fails, whether a bottle, a notebook or an instrument export.
Do this for five lots, not one. A single case can be a typo, but five cases show a habit, and a habit can be fixed with a label, a list or a rule.
- Batch number on every printout
- Sample time against reactor time
- Analyst and instrument identity
- The number quoted on the certificate of analysis
- The label on the retained sample
Look first at where samples are labelled
Most breaks start at the sampling point. An operator draws a bottle, writes the number on masking tape and walks to the lab, where peeling tape and smeared ink make consecutive bottles look identical.
Is the label printed or handwritten? Does it name the stage, in-process, pre-filling or final, since a pH taken before a correction and one taken after share a batch number but mean different things?
Composite samples add a trap. If a storage tank holds three blended batches, whose number does the sample receive, and who writes down the blend ratio so the lab can trace it later?
Instruments and spreadsheets lose the link quietly
Titrators, viscometers and pH meters often save a result with a timestamp and a sample name typed at the bench, so B-1042, B1042 and 1042A become three entries. Typed names always drift.
Spreadsheets add a second break. Someone copies a value into the monthly sheet and rounds it, so the certificate quotes the rounded figure while the notebook holds the original reading. Nobody notices until a customer complains.
Retyping is the last weak point. One transposed digit in a batch number or a viscosity figure survives all the way to the buyer, who then holds a certificate that describes a different lot.
- Sample names typed freehand at the instrument
- Instrument clocks that disagree with the office PC
- Results copied by hand into monthly sheets
- Retests that overwrite the first reading
- Viscosity units mixed between centipoise and millipascal seconds
Tie each test to the lot it released
The fix is one record per batch with the tests hanging beneath it, each test row holding the method, limits, result, unit, instrument, analyst and time. A reviewer signs the release there.
Set the specification limits on the product itself, so the system flags an out-of-limit assay or viscosity at the moment of entry and places the batch on hold. Nobody has to remember.
Retests need rules. Keep the first result, record why the repeat was run and let a reviewer decide which value counts, with both readings staying visible on the batch record.
- Batch number picked from a list, never typed
- Specification limits stored per product and method
- Automatic hold on an out-of-limit result
- First result kept beside any retest
Run one product for one week before widening
Do not fix every product at once. Choose the product that draws most complaints, then run a week of habits, printing sample labels with batch and stage and picking instrument sample names from a list. Use one entry sheet.
Repeat the five-lot trace when the week ends, and if every printout matches on batch number, sample time, analyst and certificate value, move to the next product.
Keep the trace sheet. An auditor will like it, because it shows that you test your own records and fix the handoffs that fail, not only the batches that get rejected.
Batch-to-test trace sheet
Five released lots, test by test. Columns are batch number, stage, test, method, specification, result, unit, instrument, analyst, sample time, value on the certificate and a match mark. A closing box records which handoff failed, who will fix it, which product is traced next and the date that second trace is due.
Open a blank worksheet to printQuestions owners ask
Why do assay and viscosity carry different batch numbers?
Usually the two tests were sampled at different times or by different people, and each analyst wrote the batch number in their own way. Compare the sample times, the bottle labels and the instrument clocks to see which test sits on the wrong lot before anyone edits a certificate.
Can we fix old results?
Correct them only where the original sample label or notebook proves the right batch, and record the correction, the reason and the approver in the batch file. Never bend a result to fit.
Do we need a LIMS for this?
Not always. A small plant can start with sample labels and an entry form tied to the batch record, and a LIMS helps once instruments, sample volumes and analysts multiply.