CTV Household Graphs and Healthcare Audience Targeting

When a data provider says it can deliver a condition audience on CTV, there is a household graph doing the work in the middle. The audience starts as people (from consumer records, surveys, purchase data, or CRM lists), and the ad lands on a TV. The graph bridges the two. Most of the risk in healthcare CTV targeting sits in that bridge.

How CTV household graphs are built

Graph providers assemble households from several signals:

• IP addresses. Devices seen on the same residential IP at the same times are grouped. This is the backbone of most CTV graphs.
• Device identifiers. TV and streaming device IDs, mobile advertising IDs, and browser identifiers seen on those IPs.
• Logins. Authenticated users on streaming services, retailers, or publishers who connect a hashed email to devices.
• Postal addresses. Matched to IPs through data partnerships, often to support offline data and direct mail targeting.
• ACR and platform data. Smart TV makers and device platforms see the TV directly and can link it to the home network.

Links built on logins and verified address matches are deterministic. Links inferred from IP co-occurrence are probabilistic. Most graphs blend both, and the blend changes by region and device type. The general evaluation questions are in how to evaluate an identity graph for healthcare media.

IP and device linkage: where accuracy breaks

Residential IPs are not permanent. Internet providers reassign them, sometimes every few days, sometimes every few months. A graph built on last quarter's IP observations can point a household's record at a different family today. Other common errors:

• Shared networks. Apartment buildings, college housing, assisted living facilities, and offices put many households behind one IP.
• Mobile carrier IPs. Phones on cellular networks share large IP pools that do not map to a home.
• VPNs and privacy relays. They mask the home IP.
• Stale addresses. People move. Address data can lag by months.
• Multiple TVs. A household with three TVs and a streaming stick looks like several devices that may not all be linked.

For healthcare, two populations are hit hardest. Older adults in senior housing sit behind shared networks. Younger adults move often. Both are core audiences for many conditions.

How errors show up in a healthcare campaign

Measurement errors from a graph do not average out. If exposed households are linked to the wrong outcome records, measured lift is pulled toward zero. If holdout households are accidentally served through an unlinked device, the test is contaminated. See household identity in DTC pharma for how spillover affects the read.

Privacy limits for health segments on CTV

Health segments on a household graph combine two sensitive things: an inference about a health condition and a link to a specific home. Several state consumer health data laws, Washington's My Health My Data Act being the best known, require consent for collecting and sharing consumer health data and define that term broadly enough to cover many inferences. Other states have similar laws or sensitive data provisions. The FTC has also brought actions over health data sharing, and its Health Breach Notification Rule applies to some health apps. HIPAA usually does not apply to an advertiser's audience data, which is not the same as saying the data is free to use.

Practical questions to settle before activation:

1. How was the health signal created, and did the consumer consent to that use? See health data provenance questions.
2. Does the segment include inferences that state law treats as consumer health data?
3. Is the segment large enough that a household cannot be singled out?
4. Will the graph provider or CTV platform retain the segment association after the campaign?
5. Does the segment follow the NAI's guidance on sensitive health data, if your partners are members?

None of this is legal advice. It is a list of what your privacy and legal teams will want to see before they approve a condition segment on CTV. State consumer health data laws and pharma media covers the laws in more detail.

How to test a household graph before buying

1. Build a truth set. Use a set of households where you know the address and devices, such as opted-in panelists, employee volunteers with consent, or a measurement partner's panel.
2. Ask the graph provider to match it without knowing which records are the truth set.
3. Score the result: share matched, share matched correctly, and share matched to the wrong household.
4. Check by subgroup: older adults, renters, rural households, multi-TV homes.
5. Check freshness: when was each IP link last observed?
6. Run a small flight to the truth set and confirm delivery through ad logs.

A hypothetical scoring example: a provider matches 7,000 of 10,000 truth set households, of which 5,600 are correct and 1,400 point to the wrong household. Match rate is 70 percent, but precision is 80 percent, and correct coverage is 56 percent. The headline number is the 70. The useful number is the 56. The same logic applies to physician households in CTV for HCPs.

Practical takeaway

Before using a health segment on CTV, ask the provider for two numbers on a truth set you supply: the share of households matched and the share matched correctly. Get your privacy team to sign off on the segment's source and consent basis in the same review. If either is missing, use contextual or broad demographic targeting until it is resolved. The series guide on CTV and OTT healthcare advertising lists the alternatives.

Editorial note. Analysis and frameworks are the author's own and do not represent Acxiom or any current or former employer, client, or named platform. Examples labeled hypothetical or illustrative are not results from real campaigns. Nothing here is legal, regulatory, or medical advice.