Your Agent Is Only as Good as Its Audience: When to Use Syndicated vs. Custom
A framework for choosing audiences in agentic media buying
by Kayla Jordan, VP, Programmatic Sales, InMarket
Every major DSP shipped an agentic mode this year, so for most programmatic buyers the question has moved past whether to use it. What gets less attention is what the agent is optimizing against, and who’s checking that it’s right. I’d start that conversation with the audience.
Audiences are a proxy, too
Most of us already know ROAS has limits. It counts motion, a click or a view-through somewhere along the path, and whether that motion turned into incremental sales or visits is a separate question. None of this is news; marketers have been arguing over these ideas for years.
The same issue sits one layer up, in the audience data feeding the buy. An audience is a bet that a certain group of people will convert. Under manual buying, someone on the team looked at segment performance every so often and caught drift eventually. When an agent makes the buying decision, that check tends to go away, and the agent optimizes confidently against whatever signal it’s given. So how well the audience is initially built matters more now than it did a year ago.
Two audience types built for two different jobs
Syndicated audiences are pre-built segments you can activate right away. They’re ready off the shelf and sized for reach, which makes them the fastest way to get a campaign live. Plenty of syndicated data across the industry is modeled on resemblance, meaning it finds people who look like your buyer. InMarket builds its syndicated segments from observed purchase and location behavior and refreshes them monthly, so an off-the-shelf InMarket audience is still grounded in what people bought and where they went, and refreshed regularly so they’re relevant to your current campaigns.
Custom audiences are built to your exact criteria or from your own outcome data. They take a few days to stand up. They’re intentionally narrower, designed to get closer to conversion, and at InMarket they can refresh as often as weekly.
Both are the right tool in the right situation, and how well an audience is built matters more than which label it carries. Trouble shows up when a campaign is planned around one kind of question and then graded on a different one.
When syndicated is the smart choice
Syndicated audiences carry a lot of campaigns well. They’re usually the best fit when:
- You need to launch fast. A campaign that has to go live this week can’t wait on a custom build, and a syndicated segment gets you in market with real behavioral data behind it.
- Reach is the goal. Awareness and prospecting programs benefit from the scale syndicated segments are designed for.
- Your category moves at a steady pace. Someone making a seasonal trip to a sporting goods store doesn’t change behavior week to week, so a monthly refresh keeps the segment current.
- You’re testing before you commit. Running a few syndicated segments side by side shows you which buyer types respond, which gives you a sharper brief if you decide to build custom later.
For InMarket buyers, one point is worth repeating. If your KPI is a purchase or an in-store visit, a syndicated InMarket audience is already built from people who bought and/or came in. You’re starting from real-world outcome either way.
When a custom build earns its place
Custom becomes the better fit when the campaign asks a narrower question than an off-the-shelf segment was built to answer. You’ll usually see it in situations like these:
- The KPI hinges on a specific real-world outcome. When success is a visit or a sale tied to precise criteria, a custom audience gives you a fuller, more detailed view of that consumer’s behavior.
- The category moves faster than the segment refreshes. Grocery shoppers picking up shelf-stable staples like macaroni and cheese change their habits week to week. A weekly refresh keeps the audience inside the window where the data still describes how people behave.
- Two buyers look alike on paper. Custom criteria like party size or time of day can separate buyers that a broad segment would group together.
- You want to build from your own outcome data. Brands with first-party purchase or visit data can use it as the foundation for the audience.
Three questions to ask before you choose
Run your plan through these before you pick an audience. Once you’ve answered them, the choice turns into a sizing and data criteria exercise.
| Question | Leans syndicated | Leans custom |
|---|---|---|
| Does my KPI depend on a purchase or a visit? | Broad purchase or reach goals, where a segment built on observed behavior covers the need | A specific action with narrow criteria, or goals tied to your own outcome data |
| How fast does my category move? | Steady or seasonal categories where a monthly refresh keeps up | Categories or campaign windows where behavior shifts week to week |
| Is a person or an agent making the final call? | Either works, as long as the audience and measurement come from the same source | Either works, and a tighter custom build gives the agent a cleaner signal to optimize against |
Whichever you choose, close the loop
The agentic piece is where splitting your stack gets expensive. If your audience comes from one vendor and your measurement comes from another, the gap between them is exactly where nobody is watching, and the agent keeps optimizing through it.
InMarket’s Campaign Incrementality ties visits and sales back to the media that drove them at the user level, using the same data source as the audience. It optimizes in flight, so the campaign can correct itself while it’s still running. Reporting goes all the way down to individual audiences, showing incremental sales lift and incremental visits for each segment. That works the same way whether you bought syndicated or built custom, and it lets you see which audiences are driving real-world results.
Maestrø, InMarket’s agentic optimization layer, puts those results where an agent can act on them. Through the InMarket MCP, it brings causal incrementality data into the AI tools your team already works in, so the agent can adjust a live campaign based on the visits and sales each audience is driving.
When AI is making the buying decision, the data behind it has to be real, and that holds for syndicated and custom alike. Start with what your campaign needs answered, then check whether the data behind your audience can answer it.
If you’d like to run your own plan through the framework, the InMarket team is happy to walk through it with you. Reach out to us at audiences@inmarket.com for more information.