4.24 glitch

This is indeed tricky – because the title does not actually start with “viatmin a”, and the keyword “a” is simply too broad for a match.
Let’s try something different here.

On the index table settings make sure the following options are set:
– For the index product categories are selected: https://i.imgur.com/7WHS2hL.png
– The keyword length is set to 1 and stop words are disabled: https://i.imgur.com/vE3CRR4.png
Now, even if you don’t change anything, create a new index anyways: https://i.imgur.com/GasZcMn.png

For the search settings, turn off taxonomy term results for now: https://i.imgur.com/2LqhksS.png

Now, the product categories with the exact keyword match are in the system now. Let’s try to prioritize the term matches, by significantly changing the weights for the index matches like this: https://i.imgur.com/TuQGq2x.png
For starters use the value 99999, then you can lower it afterwards.

If all goes right, this should put the matches based on the product category to the top. That means typing “vitamin a” should find products categorized “vitamin a” and give that match a super high relevance value, (hopefully) putting it to the top.

Okay, so it is closer now. That happens very likely because the keyword occurence for the first result is denser, as it has the keyword “vitamin” more times in the term fields, which has a very high weight now.
You can still probably balance this somehow with a bit of experimenting. Try bringing the terms weight and the titles weight closer together, something like this: https://i.imgur.com/zKyC0P0.png
It may take some experimenting, but try playing with the terms weight value only first. Eventually you should see changes in the order, and there might be a sweet spot, where the terms weight is just about right and flips the first two results to the desired location, without actually affecting the other ones too much.
I know it is super tricky to get it right, but for a human eye some items seem logically more relevant, but when a program calculates the actual relevance based on keyword occurences and density it may show something completely different.

Yes – in the past customers tend to set all kinds of values for no reason and those are used in both addition and multiplication operations. It is only a guidance, for most cases it is sufficient. Having lower numbers makes the calculations more straightforward and does not override certain relevance calculation values. Your case is an exception, so you can ignore that suggestion.

Neat!

In case of WooCommerce the popularity is determined by the “total_sales” custom field. If you use that, it should work. However using it as the primary ordering field is probably not the best idea, as other items matching the query with higher sales numbers will be displayed first – which are not neccessarily the best matches.
The secondary ordering could work though, but that is obviously applied only after the primary and only for matches having the same exact relevance. It is very unlikely to have matches with the same relevance when the keywords are very short (like “a” in “vitamin a”), so you may not see much difference in the final results order.

You might be seeing items not having a brand assigned to them. Try turning off this option: https://i.imgur.com/ayiNSOC.png

Okay, let me know if you have any specifics and I will check on that too.