Targeting Strategies

Use this page to decide what to build. The mechanics are covered in Building an audience; this page is about which audiences are worth the effort and which patterns quietly waste money.

Important: Advertising platforms are not HIPAA business associates and do not sign BAAs. Hashing an email does not make it anonymous. Read Compliance for audiences before sending an audience to any platform.

---

Start with exclusion, not acquisition

The instinct is to build audiences of people to reach. The higher-return first audience is usually a list of people to stop reaching.

Every advertising platform will happily keep showing an acquisition ad to somebody who converted last week. The platform has no idea they converted, because conversion happened on your site or in your clinic rather than in the ad account. So you pay for impressions against people who are already customers, and the people who see them are the ones most likely to click, which makes the campaign look like it is working.

An exclusion audience fixes it in one move. Build an audience of people who completed the action, then attach it as a negative or excluded audience in the ad platform. Ours Privacy sends the membership; the platform decides what to do with it, and every major platform supports using an audience as an exclusion.

The pattern. One event condition: the conversion event, count is greater than or equal to 1, over a window long enough to cover how long somebody stays a customer. For a booked appointment that might be 90 days. For a repeat purchase it might be 30.

Why a daily sync matters here. Yesterday's converters need to be out of today's targeting. A CSV you uploaded last month excludes last month's customers and nobody since. See Syncing to destinations.

Worth knowing: not every platform gives you a built-in exclusion control, so on some of them the equivalent is a "has not yet converted" audience instead of an exclusion. In Ours Privacy the zero-count pattern is a first-class condition, so you can express either shape. See The zero-count pattern.

---

Exclude within a funnel, not only at the end

The same logic applies at every step, and this is where most of the upside is.

If you are running a campaign to get people to book, exclude people who already booked. If you are running a campaign to get people to complete an intake form, exclude people who completed it. Somebody who is one step further down the funnel than your ad is asking for is somebody the ad is wasted on, and the ad may be actively unhelpful: an ad asking a current patient to become a patient is a bad experience, not just a bad impression.

A useful audit: list your live campaigns, and for each one write down the action it is asking for. Every one of those actions is an exclusion audience you probably do not have yet.

---

Retargeting the right window

A retargeting audience is people who showed intent and did not finish. The failure mode is not building it; it is the window.

Too short and you lose people with a considered decision. Healthcare decisions in particular are rarely same-session, and a seven-day window on a procedure page throws away most of the people who were genuinely thinking about it.

Too long and relevance collapses. Somebody who looked at a page five months ago and never came back has moved on, and you are paying to remind them of a decision they made.

Sensible starting points, to be adjusted against your own sales cycle:

Always pair a retargeting audience with the exclusion. The correct shape is not "viewed the page" but "viewed the page and did not convert", which means two event conditions: the intent event with a count at or above 1, and the conversion event with a count of exactly 0. Without the second condition you are retargeting your own customers.

---

Layering intent with something you know

An event condition tells you what somebody did. A visitor condition tells you who they are. Layering the two is how you turn a broad audience into one worth targeting.

Two conditions is usually the sweet spot. Viewed the pricing page and is in a state you actually serve. Started an application and arrived from paid search rather than from your own email. Visited three times and has an email address on file.

Past two or three conditions the audience tends to get small enough that platform minimums bite and the estimate stops being reliable. See Previewing an audience.

---

Lookalike seeds

Platforms can build a lookalike or similar-audience from a list you provide. The quality of the result is almost entirely determined by the seed.

Seed on your best outcome, not your biggest list. A seed of everyone who ever converted models an average customer. A seed of people who converted and stayed, or who bought the higher-value thing, models a customer you actually want more of. Smaller and sharper beats larger and diffuse.

Respect the size floor. A seed needs enough people for the platform to find a pattern. Meta's guidance for a customer list is at least 1,000 people; Google recommends at least 5,000 for a Customer Match list to serve reliably. Below that, tighten less rather than more.

Exclude the seed from the campaign. This is the step people skip. A lookalike campaign that does not exclude its own seed spends part of its budget re-reaching the customers you built it from, and those customers convert at a higher rate, which makes the lookalike look better than it is. Attach the seed audience as an exclusion on the campaign.

Refresh it. A seed frozen in a CSV models the customers you had when you uploaded it. Syncing keeps the seed current as your customer base changes.

---

Suppression beyond advertising

Not every audience exists to be targeted. Some exist so that somebody is left alone.

People who asked not to be contacted, people in an active support escalation, people who already have an appointment this week: each of those is an audience whose only purpose is to be excluded from something. Building them explicitly, and syncing them so they stay current, is cheaper than remembering.

---

Practices worth adopting

Name for the team, obscure for the platform. The strategy above is visible in a descriptive audience name, and ad accounts are shared with agencies and contractors. Put the description in Name and generate the Export Name. See Export names.

Preview before you sync. An audience of 40 people will not serve anywhere. Check the estimate and the matchable count first.

Change one thing at a time. If you widen the window and add a condition in the same edit, a change in performance tells you nothing about which one did it.

Measure what the audience did. Audience Performance reports on conversion for audiences, which is the difference between believing an audience works and knowing it.

Retire audiences. A synced audience runs every night until somebody stops it. Put a recurring review on the calendar.

---

Next Steps

• Building an audience for the mechanics of each pattern above.
• Event conditions for windows, counts, and the zero-count pattern.
• Recipes for these strategies written out as configurations.
• Audience Performance for measuring the result.

Need help? Contact support@oursprivacy.com.