Elasticsearch mapping with dynamic templates for customer attributes

template.json

{
  "template": "account-*",
  "settings": {
    "analysis": {
      "analyzer": {
        "default": {
          "tokenizer": "keyword",
          "filter": "lowercase"
        }
      }
    }
  },
  "mappings": {
    "contact": {
      "properties": {
      },
      "dynamic_templates": [{
        "long_to_integers": {
          "match_mapping_type": "long",
          "mapping": {
            "type": "integer"
          }
        }
      }, {
        "double_to_integers": {
          "match_mapping_type": "double",
          "mapping": {
            "type": "integer"
          }
        }
      }, {
        "float_to_integers": {
          "match_mapping_type": "float",
          "mapping": {
            "type": "integer"
          }
        }
      }, {
        "strings_to_texts": {
          "match_mapping_type": "string",
          "mapping": {
            "type": "text"
          }
        }
      }, {
        "fields_ending_in_at_as_dates": {
          "match_mapping_type": "string",
          "match": "*_at",
          "mapping": {
            "type": "date"
          }
        }
      }, {
        "fields_ending_in_count_as_integers": {
          "match_mapping_type": "long",
          "match": "*_count",
          "mapping": {
            "type": "integer"
          }
        }
      }]
    }
  }
}

I use this template for a multi-tenanted elasticsearch instance with one index per account.
Each account pulls data from multiple sources and can have any manner of attributes.

Things to note:

I don't care too much about exact replicas of the data - it is stored in Postgres and this is just used for search. To that end, I very aggressively coerce floats, doubles and longs into integers.

All matches are done on the lowercased term, with none of the ES magic.