73.6% of the database's blending suggestions cross odour families, and a quarter of them point at floral

About 4 minutes.

Last time I measured that the "blends well with" list is hard-truncated at 60 per kind, so for more than half of all materials you see only a top sixty.

This article asks the next question: which way do those sixty point?

It is not like-with-like

I matched both ends of each suggestion — the source material's odor_type against the suggested material's group in the grp field. 295,387 pairs resolve on both ends.

	Pairs	Share
Same odour type	77,877	26.4%
Different odour type	217,510	73.6%

Nearly three quarters cross families.

That is good news. A list that only pushed same-family materials at you would be no help in building contrast. It does not.

But it has a clear destination

Counting all grouped suggestions (331,821) by the suggested material's odour type:

Suggested odour type	Pairs	Share
Floral	80,697	24.3%
Balsamic	39,690	12.0%
Citrus	32,192	9.7%
Fruity	28,437	8.6%
Herbal	16,760	5.1%
Green	13,283	4.0%
Aldehydic	10,586	3.2%

A quarter of all suggestions point at floral, twice the share of the runner-up.

And the commonest cross-family routes — the top three all end at floral:

From	To	Pairs
Fruity	Floral	9,689
Herbal	Floral	7,809
Green	Floral	7,707
Floral	Citrus	6,468
Floral	Balsamic	5,881
Woody	Balsamic	4,946
Spicy	Floral	3,995

Fruity, herbal, green, spicy, woody — all of them get pointed at floral.

What this means

One: the list has a direction, and it does not tell you. When you read a material's suggestions, you are reading a list already weighted toward floral. That is not wrong, but you should know it.

Two: that weighting may reflect how the industry actually works, or it may reflect the compiler's coverage. I cannot separate them — both produce identical statistics.

Three: read it together with last time's cap. Each material has sixty slots, and a substantial share of them go to floral. Sixty slots is a scarce resource, and floral takes the largest cut.

Four: the woody→balsamic route is worth noting (4,946 pairs), because it is the only cross-family link in the top seven that does not involve floral. After floral, balsamic is the second hub.

What to do

• For floral pairings, this list works well. That is where its density is.
• For non-floral contrast, this list will disappoint. Use a formula corpus, or read the odour-type distribution directly.
• Do not read "this material's suggestions are mostly floral" as "this material suits floral". It may just be the database's overall bias.
• Only 26.4% are same-type, so this is a poor tool for finding "same family, different molecule" substitutes.
• Last time's cap and this time's direction belong together. The list is a sixty-slot, floral-weighted ranking.

What this doesn't establish

• The source uses odor_type and the destination uses grp. Their values look like the same vocabulary (floral, balsamic, citrus…), but nothing guarantees the two are defined identically. If the definitions differ, the same-versus-cross ratio is biased. That is the biggest methodological gap here.
• 4,172 pairs have no group on the suggestion, and a further 108,086 rows have a group field reading "no group found", all excluded.
• I computed only the kind='odor' half. The flavour half (337,088 rows) is untouched.
• "Following the list pulls you toward floral" is inferred from the distribution. I did not track how anyone actually uses it.
• I do not know how the sixty are ranked, so I cannot say whether floral's dominance comes from the ranking or from coverage.
• This article evaluates no suggestion's quality. As last time, it is about the shape of the list.