Identity Graphs
An identity graph is a database that links the many identifiers belonging to one person or household "” emails, device IDs, cookies, IP addresses "” into a single, connected profile for targeting and measurement.
Key takeaways
- An identity graph connects the scattered identifiers of one person or household.
- Matching is deterministic (based on known links) or probabilistic (inferred).
- Graphs can resolve to an individual or, commonly in CTV, a household.
- They underpin cross-device targeting, frequency capping, and measurement.
Deterministic vs probabilistic
Deterministic matching links identifiers through verified connections "” for example, the same login across devices. Probabilistic matching infers that identifiers belong together from signals like shared IP, location, and behavior. Deterministic is more accurate; probabilistic extends reach where deterministic links are missing.
Household resolution and CTV
In connected TV, graphs typically resolve to the household via shared IP rather than the individual, which is why CTV frequency and targeting operate at the household level. Getting household resolution right is central to accurate CTV reach and measurement.
| Purpose | Unify one person's/household's IDs |
|---|---|
| Deterministic | Verified links (e.g. logins) |
| Probabilistic | Inferred from IP, behavior, location |
| CTV unit | Usually the household |
Frequently asked questions
What is the difference between deterministic and probabilistic identity?
Deterministic matching uses verified connections like shared logins and is highly accurate; probabilistic matching infers links from signals like IP and behavior, trading some accuracy for broader reach.
Do identity graphs identify individuals or households?
Both exist, but in CTV they usually resolve to the household because devices share an IP address. That's why CTV targeting and frequency are typically household-level.