Traits

Traits overview

Traits are interesting features about individuals and/or their households, sourced from both your data and Faraday's. The Faraday Identity Graph (FIG) includes 1,500+ rich attributes of data on 240 million adult consumers, with built-in identity resolution. Faraday's data is licensed, permissioned, and responsibly-sourced from reputable data brokers, and includes data from public record, surveys, self-reporting, and non-cash transactions.

Traits are primarily used in cohorts and persona sets. In cohorts, they can be used to help define and add nuance to groups of people that your business is interested in, and in persona sets, they're how your persona predictions are made–by grouping the individuals in your cohorts by the highest coverage traits among them. You can also add traits to pipelines to add more context to people you're making predictions for. Lastly, you'll find traits when using score explainability in your outcomes.

Traits are broken down into categories, such as property and demography, and tiers, such as standard and premium. Traits are purchased in combinations of these two terms, resulting in packages like standard demographic and premium property.

A full list of traits in Faraday is available for viewing and CSV download via the Dashboard UI and our website, and traits can be listed, created, deleted, and more via our API. This includes both traits from Faraday, and those that you have created within your account.

Property traits

Property traits in FIG are sourced from data brokers that specialize in the collection of property data, such as tax assessor offices. Due to this, coverage for any given trait tends to follow a geographic pattern, based on the data collection requirements for a given county or municipal tax assessor office. For example, some tax assessor offices might collect heating system type, while others do not–these are represented as null values, and should be interpreted as "unknown." In Faraday, an example of this would be, if the trait heating system type is not populated, it means Faraday does not know the heating system type as opposed to the property not having any heating system at all.

Consumer traits

Consumer traits in FIG are similarly sourced from data brokers that specialize in collection of demographic, lifestyle, and financial information. The source of these traits is wide-ranging, but includes data from self-reporting, surveys, public record, and inferred traits from non-cash transaction data. Null values should be interpreted as unknown–for example, an individual with the dog owner trait not populated would mean that Faraday does not know if they have a dog as opposed to them not having a dog.

Trait accuracy

With a dataset of the size of FIG, trait inaccuracies are possible. Sources of error include–but are not limited to– data currency, misreporting by individuals, and the application of household values to individuals. As a means to combat inaccuracies, Faraday sources data from a variety of vendors to improve data quality through multi-vendor corroboration.

Trait currency

Faraday receives updated data from our data partners on varying schedules, but FIG is generally updated every 6-8 weeks. It should be noted that this does not mean that every record within FIG is updated with each installment.

Deceased individuals are identified and updated through reports from funeral homes and family members to the Social Security Administration (SSA), which our data providers collect and then we feed into FIG alongside any other data updates. While reports to the SSA for deceased individuals are highly accurate, we can’t guarantee total removal of deceased individuals from FIG for several reasons, including a family member using a joint credit card with the deceased’s name, a telephone number listed with the deceased’s name, or a change of address marking the deceased at a previous household.

Geonormalization

Geonormalization, or 'geonorm' for short, is the process of normalizing a trait or score attached to an individual so that it's relative to their location. Geonormalization is applied during modeling to help prevent geographic bias that is often present in acquisition outcomes, and geonormalized traits can be recognized by the (geonormalized) tag following the trait name.

For example, Net Worth (geonormalized by county) represents the trait Net Worth geonormalized at a county level.

To see this in action, say Jane Doe has a trait attached to her called net_worth_geonorm_county, whose percentile value is 80. This means that Jane's relative net worth is in the top 20% of individuals in the country. Geonormalization means that we use Jane’s net worth relative to individuals in her county instead of her absolute net worth because someone that lives in a location that is less like your customer data shouldn't necessarily be punished in scoring for that.

Read more about geonormalization, including examples for when it's used, on our blog.

Specific trait guides

Below, you'll find additional descriptions or tables, where applicable, for some Faraday Identity Graph traits.

Housing density definitions

The housing_density trait corresponds to “Units per sq/mi” in the table below.

Housing density definitions

Rank	Name	Units per sq/mi
0	Unpopulated	0
1	Sparsely populated	1-72
2	Lightly rural	72-337
3	Moderately rural	337-640
4	Low density suburban	640-1,200
5	Medium density suburban	1,200-2,400
6	High density suburban	2,400-3,456
7	Low density urban	3,456-4,672
8	Medium density urban	4,762-6,400
9	High density urban	6,400-10,000
10	Moderate urban core	10,000-16,000
11	Dense urban core	16,000+

Architectural style guidance

The architectural style trait (style) describes the building style of each structure.

The entry 'MULTI-FAMILY' is often used to filter for multi family housing, but it is not the most effective way to do this. The field Units in building (units) has much higher coverage, and the same filter can be built using units >= 2 and units <= 4. If you are trying to find multi family homes for any reason, Units in building should be used instead of Architechtural style.

Target valuescore

Target valuescore values translate to a rough credit score seen in the table below.

Valuescore definitions

Rank	Valuescore	Credit score
A1	Best profit margin	760+
A2	Best profit margin	740
B1	Above average profit margin	725
B2	Above average profit margin	710
C1	Average profit margin	690
C2	Average profit margin	675
D1	Below average profit margin	660
D2	Below average profit margin	650
D3	Below average profit margin	640
E1	Poor profit margin	625
E2	Poor profit margin	610

Shopping styles

The shopping styles trait segments individuals based on their shopping behavior through transactions, media habits, demographics, financials, self-reported surveys, and other datapoints to give you an indication of how they like to shop.

Shopping styles

Shopping style	Definition
Amazon-centric	Prefers shopping on Amazon for nearly everything, including groceries, for convenience, good prices, and fast shipping. Will only shop elsewhere if similar prices and two-day shipping offered.
Prefers online	Prefers shopping for most things online (also via smart device apps) for convenience, including groceries, but without specific brand/outlet loyalty. Not as price-concerned.
Bargain hunter	Shops heavily, both online and in-store, to hunt for bargain-priced items through catalogs.
Luxury offline	Big-spenders with disposable income, preferring in-store shopping at high-end name brands.
Retail therapy	Enjoys socially shopping with friends, may make impulse purchases, but has a tight budget and therefore often uses coupons/discounts. Often influenced by free samples and giveaways. Tends to leave reviews.
Brick and mortar	Prefers in-store shopping for nearly everything, using physical circulars & newspapers for inspiration. Unlikely to use credit to purchase.

Market Trend (MT) traits

Some traits in Faraday are labeled as Market Trend (MT) traits. MT traits use behavioral and survey data to determine how likely an individual is to take a certain action. For example, MT - Baseball Enthusiasts ranks individuals on a scale of 1-100 for how likely they are to plan to watch baseball for entertainment. These traits will often appear as features of importance for your predictions in Faraday, as they are strongly predictive.