Which structure of nested field is more efficient?

I'm using Elasticsearch 7.10 and I have an ES index which includes about 3M documents.
I am going to add nested field (I call it items) which has fields as below:

  • store_id (long)
  • big_category_id (integer / there are only 5 categories now)
  • timestamp (date)

Each document has about 5 items. We have to update items data everyday.

I have two plans.

  1. Add a single nested field and put all items to it with big_category_id as below.
    We have to specify big_category_id in the query.
"items": [
  {
    "store_id": 1,
    "big_category_id": 1,
    "timestamp": "2021-10-01 00:00:00"
  },
  {
    "store_id": 1,
    "big_category_id": 2,
    "timestamp": "2021-10-01 00:00:00"
  },
  ...
]
  1. Add nested fields for each big_category_id as below:
    We don't need big_category_id in the query.
"items_of_category_1": [
  {
    "store_id": 1,
    "timestamp": "2021-10-01 00:00:00"
  },
  {
    "store_id": 2,
    "timestamp": "2021-10-01 00:00:00"
  },
  ...
],
"items_of_category_2": [
  {
    "store_id": 1,
    "timestamp": "2021-10-01 00:00:00"
  }
],
...

I think the first plan is better because of its simplicity, but I don't know which is more efficient (especially when querying and fetching data) in terms of the internal structure of Elasticsearch.
Which do you think is better?