The short answer
A blend stays consistent across harvests for four reasons, and none of them is luck: a written blend sheet that fixes components and proportions; sealed master samples on both sides that define the reference cup; a fixed cupping protocol that compares every incoming and outgoing lot against those references; and a buying calendar that anticipates seasonal drift instead of discovering it. Tea varies because weather, season and processing vary - no program escapes that. What a buyer can require is that the variation be caught at the blending table, where it can be corrected, rather than in a customer review, where it can only be apologized for. This note walks the toolkit and the discipline behind it, which is the daily work of the desk's blending and customization program.
Why tea drifts between lots
| Variance source | What it changes in the cup | How professionals absorb it |
|---|---|---|
| Season and weather | Aroma intensity, astringency, body; spring material often rounder | Adjust component ratios within the blend sheet's allowed bands |
| Garden and lot character | Profile of each component tea varies lot to lot | Blend across components; qualify alternate sources per component |
| Processing variation | Roast depth, oxidation edge, moisture at packing | Fix process parameters in the spec; roast-correct finished lots |
| Leaf age and storage | Freshness decay, aroma loss in greens and scented teas | Rotate stock, control warehouse heat and odors, pack to order |
| Refining losses | Screen sizes, stem and dust content shift after sieving | Fix grade definition; check particle profile on arrival |
The pattern in the right column is the whole trade secret: drift is absorbed by blending, which is why blends exist in the first place - a straight single-lot tea cannot be corrected, only re-bought. Blend programs trade a little of the romance of single lots for the ability to hit the same cup in March and in November.
The toolkit: three documents and two bags of tea
The written layer comes first. A blend sheet names each component - origin, grade, season - and its proportion, plus the bands within which the blender may adjust when a component lot drifts. Without bands, the sheet is theater; with bands, it is a control system. Second, the master sample: a sealed reference of the approved cup, held by both sides, that outlives staff turnover and email history. Third, the retention library: a sample of every lot produced and shipped, kept by the facility, so any future complaint can be re-tasted against exactly what left the warehouse. The physical layer is the current component lots and the finished blend, which are the things actually cupped. Nothing here is exotic; the discipline is that all five exist for your program specifically, not generically.
Cupping to lock: the working protocol
Every lot is judged on the industry's five factors - dry leaf, liquor color, aroma, taste and the wet leaf - brewed on fixed ratios such as 3 grams in 150 ml, per the standard sensory method known as GB/T 23776. The desk cups twice per lot: incoming components against their own references, so problems are caught before blending, and the finished blend against the master sample, so nothing ships on memory. Two practical rules keep the protocol honest. Cup in the format your customer brews, not only the standard ratio - a base for milk tea must be judged with milk in the loop. And record one line per factor, every time, even when the lot is fine; the log is what turns a hunch about drift into a dated, defensible correction. The full setup lives on the quality control page.
Seasonal planning: anticipate, do not chase
Consistency has a calendar. Spring greens change character across the picking window - pre-Qingming and pre-Grain Rain material differ noticeably - so a program built on a spring component needs its year's volume contracted in the window, with the blend sheet's bands wide enough to cover the season's personality. Oolongs and blacks are steadier, but roasting schedules and autumn flushes still matter; whites and puer age deliberately, which is drift by intention and needs its own reference logic. Work backward from your demand curve: contract in the harvest window, plan production so the blend is made soon after components arrive, and hold finished stock in conditions that slow decay - cool, dark, sealed, away from odors. The note on spring harvest timing and the buying-calendar post cover the calendar side in detail.
Common failure modes
Programs drift for predictable reasons, and it is worth naming them. The master sample fades or vanishes - reference tea ages in the bag - and suddenly nobody remembers what the blend was supposed to taste like; re-cup and re-seal the master annually. The blend sheet survives but the bands quietly widen until they mean nothing; keep band changes on the record with dates and reasons. Staff change at the facility and the new blender cups from memory of the last lot rather than the master sample; this is why the reference must be physical, not tribal. And the quietest failure of all: a component goes out of stock, a substitute slips in that cups fine on day one, and three months later the program tastes different in a way no single lot explains. Every one of these is prevented by the same act - comparing against the sealed reference, every lot, in writing.
What to demand in writing
Whatever the supplier relationship, four commitments should be explicit. That bulk will be cupped against the sealed master sample before shipment, with the sheet shared. That a retention sample of your lot is kept and available in a dispute. That blend-sheet adjustments stay within agreed bands, and out-of-band changes require your approval with a new sample. And that documents - lot records, lab panels - reference the blend's components, not a generic category grade. None of this is adversarial; serious facilities expect it and hold themselves to it already. What it removes is the most expensive sentence in the trade: the lot is a little different this time. If you want these terms applied to your program, send the blend and the market through the quote form and the desk will respond with the sample plan and the reference protocol in writing. For how price interacts with component choice, the note on price drivers continues from here.