Several thoughts on blending

Introduction

I'm recovering from an illness which ment no tea. Today I could feel tastes at last and I do not feel too sick - which means tea time! Now I'm meeting Banna Gu Yun for the first time and it could have hardly left a better impression. But as I do not have any tea notes ready for publishing, this post will not be about a particular tea.

I wanted to write about my experiments with blending things for some time so why not now? I admit I'm an amateur blender, but some observations might be interesting to someone.

What were my blending experiments like? I took ten or so different cakes of different areas and tastes, tried to write a "taste vector" for each of them, started blending them in various ratios and made notes about the resulting things. 

Experiments

Initial failures
My initial attempts were generally bad as I tried to combine teas which were too different. E.g., I combined sweet, full tea with little taste with a tea that was rather thin, but had a strong fruity taste (Mengku), expecting fuller sweeter, pleasant fruity tea. Nope. I got a rather boring thing with hints of good things happening, but the tea did not work overall.

Although it is a heretical statement, these thick, uninspiring teas with hints of interesting things going on reminded me strongly of the 2009-10 Xizihaos I have tasted (and the 2007 Dragon and Phoenix too, to an extent) by its character.

Anyway, there were positives about this phase too. I learned that too much of something in a tea may hinder good mouthfeel (well, Man'E style bitterness does for sure). For example, the Pa Sha 2006 when alone, has a really nice mouthfeel, but it is not really top notch because something blocks it. But add it to something with a bit of cooling mouthfeel of its own and it makes that bit a huge chunk of cooling feeling.

Some successes
I must admit that out of my many initial trials, as little as 10% came out better than sum of its parts in my opinion (and in many cases, the result was worse than the minimum of its parts). But with time, the hits were much more frequent than they used to be when I started. Why is it? I started combining teas that have something in common. And that is the most important thing I learned - only when I combined teas that were close in some aspect, the result worked.

Hypothesis

I offer the following simple idea - too simple to be true, I know, but it is a reasonable approximation I hope. At least, it is consistent with my observations. It basically is - take taste vectors of teas and sum them up.

Our idealized tea, represented by a histogram, will have two important aspects to watch for - the aspects that are more intense than 100 (arbitrary units) form the main, intense taste. The aspects that are more intense than 50 help in complexity. The numbers 50 and 100 mean nothing, they were chosen to illustrate the general idea only.

a) very different teas blended
Let us have three teas:

What does the histogram mean? Y axis is the intensity, X axis means tastes - you may think of 5 as of honey, 7 as dark fruit, 9 sugary sweetness, 11 light fruitiness, 13 floweriness, 15 grasiness... it really does not matter what it is exactly. We may see that all three teas have reasonable intensity (4-5 bars), i.e., have components stronger than 100. They have 7 bars of complexity (i.e., a bar more intense than 50) and little in common.

Now we take a third of each and put it together to get a blend of size of the original teas... we get:

As you can see, there is no component that would provide any sort of "main taste", i.e., there is nothing of sufficient intensity. You can, I think, sort of meet this in the XZH Banzhang/Yiwu blend - areas which have very little in common - and the result is hardly impressive, nothing stands out. The feeling of complexity may be there, but the tea has no theme, no character.

b) similar teas blended
Now, let us consider another three teas:

These teas do have something in common (but they are not the same, of course).
Their blend:

This is obviously very different from the previous blend. The tea still has 5 bars of intense taste (and they are quite balanced, nothing stands out too much), but it has 9 bars of complexity, i.e., we just created an about 30% more complex tea than any of its single components! 

Self-criticism and mini-conclusion

You could object that teas are not as one-dimensional as I displayed them. I agree - the features do not have to be "together" as in my images (it was simple to generate this kind of data, that's all). But whether they sum up to something interesting or not does not depend on actual positions of a feature in a histogram.

Also, you could say that some components will not simply sum up, but that the interactions will be more complex. I definitely agree here. But I hope that the importance of these nontrivial influences is not large enough to make my hypothesis completely wrong. And, above that, as I have no way of describing these nontrivial influences, I can not work with them really.

Also, I have not covered the aspect of aging. On MarshalN's blog, I believe, there has recently been raised the question whether blended tea ages similarly to pure components aging separately and then being blended. I do not know anything about that and so I could not incorporate into my model either.

As a proof-of-concept, I can safely say that even with simple "adding histograms", I could get much better results than by random blending. I.e., the model has at least some predictive capabilities which is good. Thus, I think that it might be helpful to blend your own teas, considering their features, trying to combine what they have in common rather than to patch one tea by another completely different, which "fills" the desired hole in a taste histogram.

Also, what the model says and what is my experience too, it is not really worth it to combine too many teas. Ok, some big factories may have made multi-component blends, I offer an explanation in the next section.

Why blending?

Indeed, blended teas often have more complexity. Why are single mountain cakes made then? The nice explanation is that smaller, "boutique" vendors like western connoisseurs and want to give them "single-origin", pure, clean experience. 

I tend to be more pessimistic trying to explain the single-mountain phenomenon and I believe that it is simply money-driven (or it started as such). Let's go back in time and ask ourselves - why were teas blended? With the few factories and huge productions, it was pretty much impossible to make a huge heap of tea from a single mountain as the single mountain just did not have enough material to make the desired amount of cakes. The blending had  to be employed then. That is the explanation I mentioned above - if you need to have a huge amount of cakes, then you have to use a lot of areas to get the necessary amount of leaves. It may not make the result much more interesting than if you used 3 areas only, but the 3 areas would not give you enough material.

Smaller producers, e.g., western eshop owners doing their own productions, were in completely different situations than the past big factories. Not mentioning the lack of knowledge of blending, if you do not have enough money to buy a lot of materials, let's buy from several single mountains, do not blend it, and make the bug (not blended tea) a feature (single-origin tea). In some areas (whiskey), e.g., single malts are rated higher than blends, so the process of making it a feature is made easier. People get used to the single-origin tea and want more and more of it - happy seller.

I do not want to sound like I'm against single-origin teas, not at all - it's fun to talk about Guafengzhai taste, how it is very different from Gaoshanzhai, you have that Banzhang transforming and Man'e non-transforming bitternesses and all that - it is fun. But with blends, you may really get those 9 bars of complexity, while you have only 7 bars with the single-origin cakes...