Analysis answers "is this product what it claims to be", but it gets the answer from a comparison. What is compared against is the reference, and the quality of the reference directly sets the quality of the result.
A pool that works from day one
When an organisation begins measuring it has no history of its own. Comparison is therefore made against a shared pool: known, verified profiles for the same product group. That pool gives a meaningful result from the first day — it shows gross deviation, obvious adulteration and unexpected components.
Its limit is equally clear: it does not know what is particular about your product. Geography, harvest period and production method shift the profile, and a shared pool can read that shift as a deviation.
The organisation-specific layer
Every verified sample added to the library brings the comparison closer to your product. After a few months the system no longer says just "is this honey"; it says "does this match the profile of the honey you buy from that region".
A reference library is not a setup task but an asset that accumulates.
What has to be recorded
- Sample origin: producer, region, period.
- Route to verification: a laboratory result, or a trusted supplier declaration.
- Measurement conditions: samples not taken under the same conditions cannot be compared.
The third is frequently skipped and does the most damage to results. A difference between two measurements taken under different conditions comes from the measurement, not the product; that is why every sample entering the library carries its conditions.
Where the library starts
A reference library is the known state of your own product. So the first entries are not taken from outside: they come from your own production, from batches whose identity is beyond dispute. A library built on a contested batch places every later measurement on top of that dispute.
A few batches are enough to begin; what matters is not the count but that they genuinely represent. Where season, supplier or production line differ, each has to go in separately.
Separating natural variation
With natural products no two batches are ever identical. The library's job is not to erase that difference but to draw its boundary: which range is normal, and which deviation is worth looking into. Draw the range too narrow and every batch raises an alarm; too wide and a real deviation disappears inside it.
That boundary is not set once and forgotten. A new season, a new supplier and a new line all call for re-measuring it. As the library grows, the boundary sharpens.
Who ends up holding the records
Reference data is commercially valuable information: it describes how your product behaves. Unless who may access it, how long it is kept and what happens when the relationship ends are written down at the start, the library one day leaves its owner's hands.
If you do not know your own product, you cannot recognise an imitation of it.