1. Normal drift, real change
A viscosity reading of 18,000 cP in one lot and 21,000 in the next is ordinary for a thickened emulsion. A reading of 18,000 that becomes 9,000 is a formula or process change wearing a batch number. Buyers who treat both as noise, or both as faults, end up arguing about the wrong thing. The first job is to know which band is normal for that product.
2. Method and lab as variables
Two labs testing the same sample can differ by 10 percent on viscosity and by 0.3 on pH simply through spindle choice, sample temperature and equilibration time. When a COA comes from a third-party lab, ask for the method reference and the conditions. If the method is not stated, the number cannot be compared with anything, including the spec it is supposed to meet.
3. Sampling position in the drum
Pigment and thickener settle. A sample drawn from the top of a 200 litre drum can be several points off the sample drawn after the drum is rolled. Goods-in sampling should specify a rolled or stirred drum and a draw from the middle, and should note it. Much of the drift reported by brand owners is a sampling artefact that disappears when the drum is properly prepared.
4. Limits built from history
Set specification limits from at least twenty production lots, using the mean plus or minus three standard deviations, then narrow the band if the customer’s process needs it. Limits copied from a textbook or from a competitor’s data sheet are usually too wide to be useful and occasionally too narrow to be achievable, which produces deviations that teach nobody anything.
5. Trend charts, not single values
A single COA value tells you a lot passed. A chart of the same parameter across ten lots tells you whether the process is centred and whether it is moving. One point outside the band is an event; six points drifting in one direction is a process change, and it will eventually produce the failure that gets noticed. Review the chart monthly, not at audit time.
6. When to open a deviation
Open a deviation when a result is out of spec, when a parameter moves more than two standard deviations from its recent mean, or when a test method or supplier changes. The third case is the one most often missed. A new emulsifier supplier with an identical data sheet can shift viscosity by 15 percent and still meet every incoming specification.
7. Time between production and testing
Emulsions continue to structure for days after filling. A viscosity test run at hour two and a retest at day seven are measuring different materials. Agree a conditioning period, usually 24 to 72 hours, and apply it to every lot. Where the customer tests on arrival, state the conditioning assumption in the specification so the comparison is fair.
8. Causes on the supplier side
Raw material lot variation, a changed milling screen, a different water source or a replacement homogeniser valve all move finished values without any visible change in the process sheet. Ask suppliers to notify formulation or source changes rather than only failures. Notification is cheap; discovering the change through a customer complaint is not.
9. Documenting the decision
When a lot is released inside the band but at the edge, write a short note explaining why it is acceptable and what would have triggered a hold. That note is what a brand owner needs when their own customer asks. It also builds the institutional memory that makes the next limit revision defensible.
10. Where it fits
Consistency between lots is the same discipline that governs shade, fill weight and stability review. Our cream range shows the formats this applies to, and encyclopedia covers the underlying test methods.