Tabular Data Schema
To ingest data via file upload or snowflake sources, you must follow the established schema so that the ingestion process is seamless.
By conforming your data to the proper schema, you will be able to bring your first-party data into your node which is the first step to start collaborating with your partners.

Formatting your CSV or TSV file or table

Prior to uploading a CSV or TSV file or connecting to your Snowflake database, you must format your data using a predefined convention so your DCN can ingest it successfully. There are three column types that are processed: identifier, scope and trait. Each column type will require a specific header format to be properly processed.

Identifiers

There are 2 classes of identifiers: matchable and custom. For more information on what the values of each type could look like, head over to identifier types.
  • The identifier column must match one of the following headers in order for your node to recognize the values.
Identifier
Header
Format (Value)
Email address
id_e
Raw email or pre-normalized (lowercase + space trimmed) and SHA256 hashed
Phone number
id_p
Raw phone or pre-normalized (E.164) and SHA256 hashed
IPv4 address
id_i4
Four period (.) separated octets, each of which contains a number from 0 to 255
IPv6 address
id_i6
Eight colon (:) separated segments, each of which contains 4 hexadecimal digits, lowercase
Apple IDFA
id_a
Down-cased alphanumeric value, including dashes if any
Google GAID
id_g
Down-cased alphanumeric value, including dashes if any
Roku RIDA
id_r
Down-cased alphanumeric value, including dashes if any
Samsung TV TIFA
id_s
Down-cased alphanumeric value, including dashes if any
Amazon Fire AFAI
id_f
Down-cased alphanumeric value, including dashes if any
NetID
id_n
Base64 web-safe encoded value with 70 characters (TPID)
Custom ID
id_c, id_c1, id_c2, id_c3, id_c4, id_c5, id_c6, id_c7, id_c8, id_c9
Free form text, case sensitive. Cannot be used for matching, for any subsequent ids, +1 at the end of the header name (from 1 to 9)
Identifier (catch-all)
id
Any type-prefixed ID. If we do not successfully detect a valid prefix (for example "e:", "i4:", "a:", etc.) for the identifier, we will attempt to auto-detect raw email addresses and phone numbers and automatically encode and type-prefix them at import time
Scope
  • The scope column indicates either person or household scope. It defaults to person if unspecified.
  • The scope is applied on the primary ID and all the traits of that record.
Trait
  • All trait columns are prefixed with trait_ followed by the trait key name.
    • For example, columns named trait_age, trait_sex and trait_owns_house would map to traits age, sex and owns_house, respectively.
  • All traits are associated to the primary identifier and not the neighbouring IDs
  • Trait values are optional.

Ingestion Rules

If header formatting conventions are followed and the DCN can properly ingest all profiles, there are a set of rules that will be applied upon ingestion:
  • The first ID column is treated as the primary ID in the associated ID profile
  • The value specified for the first ID column must be valid, otherwise the entire row is ignored
  • Additional ID columns (if any) are treated as neighbouring IDs
    • Neighbouring identifiers are considered of scope "person". To change it to "household", insert a subsequent row with that identifier as primary ID and with the scope "household".
  • The values specified for additional ID columns are optional. If no value is provided, that specific neighbour ID will be ignored but the remaining data specified in the row will still be ingested.
The DCN does not allow columns to share the same name, if you have multiple identifiercolumns with the same type, you must add a suffix such as _suffix to the column name. For example:
  • Invalid: id_e,id_e
  • Valid: id_e_primary,id_e_secondary

Tabular data schema examples

In the following section, you can find examples of how your data should be structured with varying parameters:
Example 1: A simple list of emails
In the example above, 2 distinct person clusters will be created with their respective email identifiers.
Example 2: Two profiles with multiple emails and some traits.
id_e
id_e_secondary
trait_income
trait_ownscat
In the example above, 2 distinct person clusters will be created but only the first will have 2 email identifiers.
Example 3: Two profiles with a neighbouring ID and some traits.
id_e
id_p
scope
trait_subscribed
trait_household_income
555-555-5555
person
yes
household
no
100000
In the example above, the email provided is the primary identifier and the phone number is a neighbouring ID within this profile. Any trait or scope definition is associated to the email.
Example 4: Four profiles with a mixture of ID types with multiple neighbours and traits.
id
id_p
trait_income
trait_age
i4:8.8.8.8
555-555-5555
50000
54
s:e0ef86a8-6ebf-4c9d-9127-e69407fe748d
+15145551234
25000
24
f:6e853799-ef31-4a30-8706-9742be254d38
24000
i6:2001:0db8:85a3:0000:0000:8a2e:0370:7334
+21630010123
56
In the example above, the primary identifier for a profile is the id column which allows for any type of ID. The email column is the neighbouring ID, the scope column and the traits columns are associated to the primary id.
There is a limit of 64 traits and 64 neighbours per identifier. Any record which exceeds those limits will be ignored. Note that the time to ingest the file will depend on the number rows, number of identifier columns and number of trait columns contained in the within.
Once the data has been formatted to the accepted convention, you will be able to proceed with your File Upload source setup or begin ingestion from your Snowflake database. When done, you can create an audience based off your source which can be used for Matching, Activation or to Export to a destination.

Errors

Upon initiating an ingestion, the DCN may reject the file or view due to one or more of the following errors:

1. Invalid header(s)

All headers of CSV or TSV files must be valid. Any invalid header will prevent the entire file from being ingested, and the last successfully ingested file (if any) will remain active.
Error
Description
Unrecognized header(s)
The header(s) do not follow our convention and could not be recognized. All headers must be recognized. The unrecognized headers must be renamed to the accepted convention.
Could not find any valid ID header
No id or id_<type> header was found. The file must contain at least one id column, with either the id_<type> prefix or id as a catch-all column with type prefixes declared for each value.
Duplicate header(s)
There are at least two headers with the same name. Each header must be unique. If you want to use multiple columns for the same id type, you can use a suffix as represented here.
Empty header(s)
There is at least one header without any name. Headers with empty names (or only whitespace) are forbidden. You must name each column according to our convention or discard it.
Trait header must start with "trait_" and must be followed by at least 1 character.
The trait header has no name - the prefix must be followed by any character to identify your trait. You must add at least 1 character after the prefix in order for your trait column to be recognized.

2. Rejection threshold

Your data is rejected if the DCN is unable to recognize more than 10% of the records within your data - nothing is ingested if this threshold is passed. The data must be formatted to ensure the accepted convention is followed.
The file will still be ingested as long as the error threshold is under 10%. If you want a lower threshold, you must fix rejected records and initiate ingestion again.
For example, if 5 records out of 10 have no identifier, the entire file will be rejected because the rejection rate is too high (50% - above the 10% threshold). The last successfully ingested file (if any) will remain active.
The same logic applies for any types of errors - the file will be rejected if more than 10% of them are invalid.