Cupping protocol — ISO vs gongfu-led

Last month in the tasting room at Yiyang, six of us sat down with four unfamiliar oolongs — two from Wuyi, one Phoenix single-bush, and a Taiwanese high-mountain whose origin had been deliberately stripped from the tin. The question that opened the session was procedural, not sensory: do we cup these the ISO way, three grams to one hundred fifty millilitres, six minutes off the boil, into white porcelain bowls, or do we cupping-by-gaiwan, multiple short infusions, smelling the lid between each? We ended up doing both, in parallel, with two people on each protocol and a third rotation halfway through. The notes diverged. Not on whether a tea was good — we mostly agreed on that — but on what each tea actually was. The ISO bowl flattened the Phoenix into a single legible profile: Mí Lán Xiāng (蜜兰香) honey-orchid, medium roast, slight astringency on the back palate. The gaiwan revealed the same tea as a sequence: first steep all floral high notes, third steep showing the roast, fifth steep going mineral and slightly cooling. Same leaf. Two different teas, depending on how you asked the question. This thread is for that argument. It is also, in my reading, the central methodological question facing anyone who teaches tea to a mixed room — judges, drinkers, sourcing buyers, beginners. The ISO protocol exists for a reason. So does gongfu. They answer different questions, and pretending otherwise has consequences for how we buy, how we grade, and how we describe tea to people who weren’t in the room. I’ve been running blind oolong panels in Hunan since 2019, and my own position has shifted twice. I’ll lay out the cases below and then hand it to the floor. For broader context on tasting pedagogy, the curriculum threads on tea.school cover the ISO methodology in proper detail, and the regional sourcing notes on puerh.app document gongfu-led evaluation as it’s actually practised in Yunnan markets.

What ISO 3103 was actually designed for

The standard cup-set — 150 ml white porcelain, lidded, three grams of leaf, six minutes brewed with freshly drawn boiling water — was published in 1980 by the International Organization for Standardization. It was not designed to help you understand a tea. It was designed to let two buyers in different countries grade the same lot and arrive at comparable conclusions. That is a very specific use case. The protocol deliberately over-brews. The six-minute steep on three grams produces a liquor that is harsh, astringent, and reveals defects — stewed character, smoke contamination, off-notes from poor storage, uneven oxidation. These are the things a buyer needs to see before committing to a container. I learned this in Anhua during a 2021 trip when an older grader watched me sniff a Wò Duī (渥堆) sample with what he clearly considered excessive politeness, then dumped it into the ISO bowl and pointed at the foam pattern on the surface. The foam told him about the pile fermentation in a way the gaiwan never would. ISO is a defect-finding protocol. It is honest about that. The problem is that it has been borrowed wholesale by competition judging and home tasting circles where the question is not ‘is this lot acceptable’ but ‘is this tea any good, and what is it doing’. Those are different questions, and the ISO bowl answers the first one well and the second one badly. Anyone running blind panels should at least know which question they are actually asking.

What the gaiwan answers that the bowl cannot

A gongfu-led tasting — call it five to seven grams in a hundred-millilitre gaiwan, water just off boil for most oolongs, infusions of five, eight, twelve, twenty, forty seconds — gives you a tea as a developmental sequence rather than a single image. This matters more for some teas than others. Wuyi yancha and Phoenix dancong both reward the gaiwan disproportionately because their character is sequential: a Phoenix Mí Lán Xiāng opens floral, develops roast in the middle steeps, and goes mineral and huí gān (回甘) sweet-returning in the late infusions. The ISO bowl gives you the average of all of that, which corresponds to nothing anyone would ever actually drink. Master Lin in Chaozhou, who I’ve sat with three times now, simply refuses ISO for dancong. He argues — and I think correctly — that you cannot judge a single-bush oolong on a protocol designed to detect defects in commercial blends. The gaiwan also lets you read the yè dǐ (叶底) wet-leaf properly across infusions, watching how the leaf opens, where the roast sits, whether the oxidation is even from edge to centre. That is impossible in a six-minute ISO steep that essentially ruins the leaf in one go. For green and yellow teas the trade-off is narrower — those teas don’t have the same multi-infusion architecture — but for oolong and Shēng Pǔ’ěr (生普洱), the gaiwan is doing analytical work the bowl simply cannot do.

The blind-tasting problem

Here is where the argument gets harder. Blindness is easier in ISO. The bowls are identical, the dose is identical, the timing is identical, and a panel can taste twelve samples in an hour with reasonable consistency between tasters. Gongfu-led blind tasting is structurally harder — the operator pouring the gaiwan can influence the result by a second here, a degree there, and reproducibility between sessions drops. I ran an internal study with our Yiyang panel in 2023, twelve tasters across three sessions on the same six oolongs, and the inter-rater agreement on descriptors was noticeably tighter with ISO. So if your goal is statistical — competition scoring, large-panel descriptive analysis, comparing seven hundred lots for a buying decision — ISO wins on operational grounds even if it loses on resolution. If your goal is to understand a single tea in depth, the gaiwan wins. Most people running blind panels are not honest about which of these two goals they have, and end up with neither rigour nor depth. The compromise I now teach is dual-track: ISO bowl for the first pass to screen for defects and rough quality tier, gaiwan for the second pass on anything that survived. It doubles the session time but produces notes that are both reproducible and analytically useful. The equipment notes on tea.equipment have a decent piece on dual-protocol setups if you’re outfitting a room for this.

What changes for unfamiliar oolongs specifically

The thread title specifies unfamiliar oolongs, and that constraint shifts the answer. With a tea you already know — say a Dà Hóng Páo (大红袍) from a producer you’ve bought from for years — ISO is sufficient because you’re checking the lot against a remembered standard. With an unfamiliar oolong, you do not yet have that standard, and the ISO bowl can mislead in a particular direction: it tends to over-emphasise roast and astringency and under-represent the late-steep development that is often where the tea actually lives. I’ve seen good Phoenix dancong scored below mediocre Taiwanese high-mountain in ISO panels because the dancong’s roast read as harsh in a single long steep when in the gaiwan it would have integrated by the fourth infusion. The unfamiliarity also matters for descriptor vocabulary. If you don’t yet have a mental model for what yán yùn (岩韵) — the rock-rhyme of Wuyi terroir — actually tastes like, the ISO bowl will not teach you. The gaiwan, taken across multiple steeps with the lid sniffed between each, will. So for unfamiliar territory I’d argue gongfu-led is not just preferable but pedagogically necessary, even at the cost of reproducibility. You learn the tea first, then you can grade it.

A working compromise

The protocol I’ve settled on for our Hunan panels, after two years of revision, runs like this. First pass: ISO 3103 bowls, three grams, six minutes, blind. Tasters mark a pass/fail on defects and a rough quality tier. Anything that passes moves to the second pass. Second pass: gaiwan, five grams in one hundred millilitres, six infusions of timed lengths (5s, 8s, 12s, 20s, 40s, 90s), still blind. Tasters write a developmental profile across the infusions, plus a final wet-leaf read. Third pass — only for teas being purchased or used in teaching — open identity reveal and a free-form session with the host present where origin and processing can be discussed. This three-stage method is slower but produces notes that survive being read six months later, which is the real test of any tasting protocol. The community calendar on tea.community has dates for the next round of these in Yiyang and Saint Petersburg, and event listings live here on tea.events under the cupping-methodology series. The argument I want to provoke in this thread is not which protocol is correct — neither is — but which question you are asking when you sit down with a tea you don’t know.

Open questions for the thread

For those running regular blind panels: have you found a reliable way to make gongfu-led tasting reproducible across tasters, or do you accept the variability as a cost of resolution? Second, do competition judges in your local circuit use ISO exclusively, and if so, do you think it disadvantages certain styles — dancong, aged sheng, lightly oxidised oolongs? Third, when teaching beginners, which protocol do you start them on, and why?