FAQ

The numbered questions below focus on interpretation, evaluation, and edge cases. They are not meant to restate what is already in the docs.

How are FollowGraph scores aggregated around an address?

FollowGraph geographic scores are delivered at the census block group level by default. When multiple geographies are combined into a trade area, radius, or custom region, the recommended method is a weighted average of the underlying index scores, not a simple average. The weight is usually population or household count, though visit or visitor counts may be appropriate in some mobility-based analyses.

So, if the 1-mile-radius output is built from block groups or subsegments around an address, the general expectation is:

sum(weight * FollowGraph score) / sum(weight)

What does it mean if a local area scores below average for an interest that seems locally strong?

A FollowGraph score is not a direct measure of local participation, spending, or category ownership. It is a modeled estimate of likelihood to follow a social account or group of accounts, expressed as an index relative to the national average.

It is also important to distinguish between different comparison contexts. An area can be strong relative to its local market or DMA while still scoring below the national average. For example, a neighborhood may have some of the highest golf affinity in the Boston area, but if golf-following behavior is even stronger in other parts of the country, that neighborhood could still index below 100 nationally.

So for golf, a wealthy Boston suburb can be genuinely high-income and anecdotally golf-oriented while still scoring below the national average for the specific social-following pattern used to define the golf variable. Possible reasons include:

• The score reflects likelihood to follow golf-related social accounts, not actual golf participation.
• The selected golf accounts may capture a particular type of golf affinity, such as PGA/media/fan behavior, not private-club participation.
• The radius may include surrounding areas that dilute the immediate neighborhood signal.
• Golf interest nationally may be especially strong in other demographic or regional contexts.
• Some high-affluence behaviors are less visible through public social-following signals.

The right interpretation is: “This area is modeled as below/above average for this social-following-based golf affinity compared with the selected benchmark,” not necessarily “people here do/do not play golf.”

What does a FollowGraph score represent?

FollowGraph estimates the likelihood that people in an area follow a given account or interest on social media. Scores are then converted into 100-based index values, where 100 equals the national average, 200 means twice the national average, and 50 means half the national average. For Interests, the score is based on a group of related accounts rather than a single account.

At what level are predictions made?

The standard delivery is at the census block group level. Predictions are generated for each block group using modeled relationships learned from social-following behavior, demographics, mobile behavior embeddings, and regionality/distance features where relevant.

7. How should customers interpret demographic skews in a single area score?

A single geographic score represents the modeled average for the population in that area. If an interest is strongly skewed by gender, age, income, or another demographic factor, the aggregate score can mask subgroup differences.

For example, an area may have a strong male golf signal and a weaker female golf signal, but the standard area-level score blends the full population into one number.