# Welcome to DS-1 docs

Everything you need to find, build, and activate audiences with DS-1

### Find, build, and activate audiences

{% embed url="<https://www.youtube.com/watch?v=A6QN259qBKg&feature=youtu.be>" %}

DS-1 is Dstillery's agentic advertising platform. Describe your campaign goal in plain language to find, build, and activate the right audience.

This site is your guide to getting the most out of DS-1.

### Find your way

* **Want the big picture?** Start with [Product Overview](/ds-1-foundations/product-overview) to learn how DS-1 works.
* **New to DS-1?** Start with [Getting started](/ds-1-foundations/getting-started) to log in for the first time.
* **Ready to build?** Continue to [Create your first audience](/ds-1-foundations/create-your-first-audience) for a step-by-step walkthrough.
* **Going deeper?** Head to [Core Use Cases](/ds-1-foundations/core-use-cases) for specific capabilities.

### Need help?

{% hint style="info" %}
Have a question not answered here? Reach out to your Dstillery team.
{% endhint %}


# Product overview

Learn how DS-1 turns a campaign goal into an activated audience.

### Meet DS-1

If you've ever planned a campaign, you know the drill. You start with a goal in your head, something like "reach people who are likely to buy a new SUV in the next three months." Then comes the gap between that idea and an actual audience you can target. Pulling segments, checking overlap, lining everything up for activation. It works, but it's a lot of steps between the thought and the thing.

{% hint style="success" %}
**DS-1 closes that gap.** It's an agentic advertising platform: an AI agent that handles the entire audience buying experience, from planning to discovery to activation. The way you talk to it is the same way you'd describe your campaign to a colleague: **in plain language.**
{% endhint %}

<figure><img src="/files/BCM0hlHOwikUFlVriYZF" alt="DS-1 campaign planning interface" width="100%"><figcaption></figcaption></figure>

### How it works

Think about the last time you explained a campaign to someone on your team. You didn't hand them a spec sheet. You said something like "we're launching a premium skincare line, we want women who care about clean ingredients, and we need to be live by the end of the month." They got it immediately, because that's how people actually talk about goals.

DS-1 works the same way. You describe what you're trying to do, and it figures out the audiences that fit. No segment codes to memorize, no menus to dig through. You chat, it listens, and it brings back the audiences most likely to move your campaign.

Most of the time, that conversation is all you need. But when you already know exactly what you want, DS-1 also offers workflows: guided shortcuts that take you straight to a common action, and you can save your own for the steps you run again and again. Either way, the flow below is what's happening beneath the surface.

{% stepper %}
{% step %}

### You describe the goal

Tell DS-1 who you're trying to reach and what you're trying to achieve. The more context you give, the sharper the result, but you don't need to speak its language. You just speak yours.
{% endstep %}

{% step %}

### DS-1 plans and discovers

It interprets what you're after and surfaces the audiences that match, drawing on the full range of targeting available to you.
{% endstep %}

{% step %}

### You refine

Maybe the first pass is close, but not quite. You can push back, ask for adjustments, or explore alternatives, all in the same conversation. It's a back-and-forth, not a one-shot query.
{% endstep %}

{% step %}

### You activate

When the audience is right, DS-1 takes it through to activation. The idea you started with is now live and working.
{% endstep %}
{% endstepper %}

### Why it's built this way

Most tools make you meet them halfway. You learn their interface, their naming conventions, their particular way of organizing things, and then you translate your idea into their system.

{% hint style="info" %}
**DS-1 flips that.** At its core, the interface is a conversation, which means there's almost nothing to learn. If you can describe a campaign goal out loud, you can use DS-1.

When you want to move fast, it also offers recommended workflows that take you straight to a common action, and if you run a process often, you can build your own workflow to match.

**The expertise stays with you, and the busywork moves to the agent.** That's what makes DS-1 an agentic advertising platform and not just another tool: it not only answers your questions, it takes action on your behalf and carries the work all the way through to activation.
{% endhint %}

### What you get out of it

The audience buying process usually has a lot of moving parts, and a lot of them are repetitive. DS-1 takes on that load so you can spend your time on the parts that actually need your judgment: the strategy, the creative bets, the calls only you can make.

So instead of asking "how do I build this audience," you get to ask the better question: "who do I actually want to reach, and what do I want to happen next?" DS-1 handles the rest.

{% hint style="success" %}
**Now that you know what DS-1 is**, the natural next step is seeing what an agentic advertising platform can do for you day-to-day. From audience planning and discovery to activation, it takes a lot off your plate. **Let's dig into the Getting started section in the next lesson.**
{% endhint %}


# Getting started

Log in to DS-1 for the first time.

### Log in to DS-1

Choose the sign-in method that matches your organization's email setup.

{% hint style="success" %}
**Your outcome:** Access to DS-1 with your work email address.
{% endhint %}

{% stepper %}
{% step %}
**Open DS-1**

Go to [ds1.dstillery.com](https://ds1.dstillery.com).

<figure><img src="/files/m0WDmVgLDahKrbW5sDUP" alt="The DS-1 login screen with email, password, and Continue with Google options." width="420"><figcaption><p>The DS-1 login screen.</p></figcaption></figure>
{% endstep %}

{% step %}
**Sign in**

Choose the option that matches your work email setup.

**Google-managed email**

1. Select **Continue with Google**.

   <figure><img src="/files/FJiNHdnkyriT4vt0j27x" alt="The DS-1 sign-in screen with Continue with Google highlighted." width="320"><figcaption><p>Select <strong>Continue with Google</strong> if your organization uses Google-managed email.</p></figcaption></figure>
2. Sign in with your work email address.

**Email and password sign-in**

1. Select **Reset password**.

   <figure><img src="/files/T0zEEcB28esLqeF1AZ0x" alt="The DS-1 sign-in screen with Reset password highlighted." width="320"><figcaption><p>Select <strong>Reset password</strong> to create a password.</p></figcaption></figure>
2. Enter your work email address and select **Continue**.

   <figure><img src="/files/AA9OYiihUP4z5gZjuMo4" alt="The DS-1 password reset screen with an email address field and Continue button." width="320"><figcaption><p>Enter the email address for your DS-1 access.</p></figcaption></figure>
3. Use the email instructions to create a password.
   {% endstep %}

{% step %}
**Start using DS-1**

After creating your password, return to [ds1.dstillery.com](https://ds1.dstillery.com) and log in.

{% hint style="success" %}
Use the work email address configured for DS-1. Check spam if the reset email does not arrive. Contact your Dstillery account team if you still need help.
{% endhint %}
{% endstep %}
{% endstepper %}

**Next:** [Create your first audience](/ds-1-foundations/create-your-first-audience).


# Create your first audience

Build, refine, and syndicate your first audience in DS-1.

{% hint style="info" %}
**Before you begin:** Confirm that you can access DS-1. If needed, see [Getting started](/ds-1-foundations/getting-started).
{% endhint %}

### Your first audience, from idea to activation

Turn an audience idea into an activated segment in six steps.

{% hint style="success" %}
**Your outcome:** A built audience, ready to send to your DSP or SSP.
{% endhint %}

{% stepper %}
{% step %}
**Confirm your workspace**

Use the sidebar selector to confirm your active workspace.

For example: `Audi`

<figure><img src="/files/LcYh5XcJpJL4ypnmppmj" alt="The workspace selector in the DS-1 sidebar."><figcaption><p>Select the workspace before you start.</p></figcaption></figure>
{% endstep %}

{% step %}
**Start a conversation**

Enter a plain-language request from the home screen. You can also select a recommended agent card.

Try: `Find me audiences that are coffee drinkers`.

Attach files to add working context across your sessions. Attach a project to use its isolated context.

<figure><img src="/files/Irksxcsw0RMPaM0dCDU8" alt="The DS-1 home screen with a prompt field and recommended agent cards."><figcaption><p>Start with a prompt or select a recommended agent.</p></figcaption></figure>
{% endstep %}

{% step %}
**Search for an audience**

Open the **Search Audience** card and enter a topic.

Try: `Luxury SUV buyers`.

DS-1 suggests related search terms, such as premium SUV dealerships. Ask for more terms, refine the list, then build the audience.

<figure><img src="/files/ZKc9strJPa5rCSJKwOxD" alt="The Search Audience card in DS-1 with audience search results."><figcaption><p>Refine the suggested terms before building.</p></figcaption></figure>
{% endstep %}

{% step %}
**Syndicate your audience**

After you build the audience, choose **Syndicate**.

Available destinations are configured for your agency. Confirm the segment name and destination when prompted.

Your segment appears in the selected platform based on its processing timeline. For example, The Trade Desk can take 24–48 hours.

<figure><img src="/files/BLYTNfoXdCWyddTMVY3l" alt="The syndication flow in DS-1 with destination options."><figcaption><p>Choose the platform that will receive the segment.</p></figcaption></figure>

<figure><img src="/files/eBqTaLIJDhrG5gjfaz6w" alt="The DS-1 syndication confirmation prompt."><figcaption><p>Confirm the segment name and destination.</p></figcaption></figure>
{% endstep %}

{% step %}
**Create a project**

Use a project for recurring client campaigns. It keeps related chats and files together.

Projects are isolated workspaces. Their history and files remain within that project. You can also attach a past conversation from **History**.

<figure><img src="/files/iDukH7XWuJu55jwmkxQM" alt="The project creation flow in DS-1."><figcaption><p>Create a project for a recurring client campaign.</p></figcaption></figure>
{% endstep %}

{% step %}
**Return to your work**

Open **History** to find earlier conversations. Use the search bar to locate a conversation, then assign it to a project when needed.

<figure><img src="/files/tpvSPM7B3ncw17VIPCAv" alt="The History view in DS-1 with a conversation search field."><figcaption><p>Search previous conversations from History.</p></figcaption></figure>
{% endstep %}
{% endstepper %}

**Keep exploring**

Learn how DS-1 protects and handles your data in [Privacy and Data Concerns](/ds-1-foundations/privacy-and-data).


# Core use cases

The audience buying process has a few distinct stages. You plan who you want to reach, discover the audiences that fit, activate them wherever you're buying, and see what's actually performing. Normally, each of those lives in a different tool, and you can spend as much time moving between them as you do on the work itself.

DS-1 covers that whole arc in one window. You're not stitching together a planning tool, a data provider, an activation platform, and a measurement layer. It's all in one place, and you move between stages without ever really leaving the chat.

Here are the core things teams reach for it to do.

### Find audiences from any starting point

Start with the signal you already have. Use prebuilt audiences, search terms, product signals, or a domain.

[**Explore the workflow →**](/ds-1-foundations/core-use-cases/find-audiences-from-any-starting-point)

### Build and activate to a DSP or SSP

Turn your audience research into a compound audience. Then activate it in the DSP or SSP where you buy.

[**Explore the workflow →**](/ds-1-foundations/core-use-cases/build-and-activate-to-a-dsp-or-ssp)

### Run predictive lift analysis

Identify audiences likely to perform against a topic, brand, or URL.

[**Explore the workflow →**](/ds-1-foundations/core-use-cases/analyze-predictive-lift-scores)

### Create pixels and seed custom AI segments

Use pixel data to seed custom AI segments from your first-party audience.

[**Explore the workflow →**](/ds-1-foundations/core-use-cases/create-and-seed-custom-ai-segments)

### Bringing it together

These use cases are a general tour of what DS-1 is good at, not the edges of it. The throughline is that DS-1 handles the entire audience buying process in one place, from finding and building audiences to activating them wherever you buy and measuring what works. And it adapts to the way you work: as you settle into the steps you run most, those become workflows of your own that DS-1 executes exactly the way you do. The more you use it, the more it shapes itself around how your team actually works.

Now that you've seen what DS-1 can do, the best way to understand it is to try it. Let's walk through your first audience step by step in the Getting Started guide.

<br>


# Find audiences from any starting point

Find relevant audiences using prebuilt segments, search terms, product signals, or domains.

Most audience tools force you to start in one specific way. DS-1 lets you start from wherever you already have a signal. Pull from ready-to-activate prebuilt audiences when you want something proven, and have it live in minutes. Use plain search terms to describe a topic "find me search terms related to the World Cup," and DS-1 will propose them for you as a starting point.

***

### Use the signals you already have

Lean on retail product signals to reach people by what they actually buy. Or hand DS-1 a domain and let it build an audience around the kind of people who visit it. Whatever you've got to work with, that's a valid place to begin.

### Explore audience sources

<div><figure><img src="/files/YvI4JBQ0JZThWFEOuGI8" alt=""><figcaption></figcaption></figure> <figure><img src="/files/O7ITqM9bVgRvnIZh3C5Z" alt=""><figcaption></figcaption></figure> <figure><img src="/files/LCxX1ZF1tuMClFIV0OC1" alt=""><figcaption></figcaption></figure> <figure><img src="/files/1t2XeWYm4YOwHtYef8Of" alt=""><figcaption></figcaption></figure></div>


# Build and activate to a DSP or SSP

Build a compound audience and activate it in your buying platform.

Finding an audience idea is easy. Turning it into a campaign-ready audience is harder.

DS-1 carries your research through to activation. Layer signals into one compound audience. Then send it directly to the platform where you buy.

{% columns %}
{% column %}

#### <i class="fa-lightbulb">:lightbulb:</i> Start with a signal

Begin with the research you already trust. Use a segment, search terms, product signals, or a domain.
{% endcolumn %}

{% column %}

#### <i class="fa-layer-group">:layer-group:</i> Build the audience

Combine your inputs into a compound audience. Each layer adds relevant context and precision.
{% endcolumn %}

{% column %}

#### <i class="fa-arrow-up-right-dots">:arrow-up-right-dots:</i> Activate it

Send the finished audience to any DSP or SSP. It arrives ready to target.
{% endcolumn %}
{% endcolumns %}

### From research to reach

{% stepper %}
{% step %}

#### Bring in your starting point

Use any valid audience signal. You do not need a perfect brief.
{% endstep %}

{% step %}

#### Add the signals that matter

Layer the attributes that define your campaign. DS-1 turns them into a richer audience than any single segment.
{% endstep %}

{% step %}

#### Push it live where you buy

Activate directly to your DSP or SSP. There are no exports, handoffs, or manual pushes.
{% endstep %}
{% endstepper %}

### The workflow in DS-1

{% columns %}
{% column %}
![DS-1 audience workflow](/files/qoEEAYlL56fFQUZ9dnJT)
{% endcolumn %}

{% column %}
![DS-1 audience workflow](/files/rdvyInpe52Ye7o65TiAZ)
{% endcolumn %}

{% column %}
![DS-1 audience workflow](/files/Uy8OVbfeLGwxVNrxDToM)
{% endcolumn %}
{% endcolumns %}

{% hint style="success" %}
**The result:** your audience moves from idea to activated campaign without leaving DS-1.
{% endhint %}


# Analyze predictive lift scores

Identify audiences likely to perform against a topic, brand, or URL.

Sometimes you don't want to guess which audience will work. DS-1 can run predictive lift analysis and surface the audiences it expects to perform well against, or in relation to, any topic, brand, or URL you give it. Instead of building an audience and hoping, you start from the ones the model already thinks are likely to move the needle.

### Start with the signal

Use the analysis input to focus your research on the audience opportunities that matter.

<figure><img src="/files/mooj0toJqDP9ifRHvPlN" alt="Predictive lift analysis input"><figcaption><p>Enter the topic, brand, or URL you want to evaluate.</p></figcaption></figure>

### Turn insight into action

Review the scored audiences, then use the strongest opportunities to guide your next build.

{% columns %}
{% column %}

#### Review the results

<figure><img src="/files/nWYvSCPkVG9Si1hN5KYN" alt="Predictive lift analysis results"><figcaption><p>Compare audiences against your chosen signal.</p></figcaption></figure>
{% endcolumn %}

{% column %}

#### Choose your next audience

<figure><img src="/files/uGXZMOYJ38Q7Hwcf4z7d" alt="Predictive lift audience selection"><figcaption><p>Use the findings to make a confident audience decision.</p></figcaption></figure>
{% endcolumn %}
{% endcolumns %}

{% hint style="success" %}
**The result:** a more informed audience strategy before you activate.
{% endhint %}


# Create and seed custom AI segments

Use pixel data to seed custom AI segments from your first-party audience.

Your first-party data is often your strongest signal. DS-1 helps you turn that signal into scalable audience intelligence.

### How it works

{% stepper %}
{% step %}

#### Create and place a pixel

Create and place a pixel, then use its collected data to seed custom AI segments.
{% endstep %}

{% step %}

#### Seed a custom AI segment

DS-1 starts with the people you already know.
{% endstep %}

{% step %}

#### Expand your audience

DS-1 identifies people with similar characteristics. Your existing audience becomes the foundation for a larger, more relevant segment.
{% endstep %}
{% endstepper %}

### Workflow

{% columns %}
{% column %}

#### Create the pixel

<figure><img src="/files/wI169Hy36xGBJkhj2UFu" alt=""><figcaption></figcaption></figure>
{% endcolumn %}

{% column %}

#### Seed the segment

<figure><img src="/files/5Z74iJF7GMhEWZmWN9Ve" alt=""><figcaption></figcaption></figure>
{% endcolumn %}
{% endcolumns %}


# Privacy and data

How DS-1 handles your data, keeps it secure, and protects user privacy.

### What it is

{% hint style="info" %}
An overview of how DS-1 and Dstillery handles data, model architecture, retention, security, and compliance, along with guardrails around sensitive category targeting.
{% endhint %}

***

### Data handling

* Every DS-1 interaction is stored, used both for QA and core app functionality
* Data within a chat session is retained so DS-1 can understand conversation context
* Chat sessions are also saved for product analytics (e.g. did the user get a proper response, did an error occur)
* Client first party data is strictly sequestered and used exclusively for that client's custom models, never shared or repurposed
* For DS-1 activation workflows, no external partner data is stored or persisted beyond the active session; it's read in-memory and discarded after

### Retention

* Platform generated data is stored in its original form for up to 30 days, then deleted
* A de-identified version of that data is retained for up to two years; pseudonymous data may be retained up to one year for research
* Non-PII stored in cookies expires no later than 60 days from last device encounter
* DS-1 conversation data itself is currently retained indefinitely

### Access controls

* Access is governed by Role Based Access Control (RBAC), following least privilege
* Client facing teams (Sales, Account Management) access campaign data and performance metrics only
* Data Science and Engineering access raw data for model building and troubleshooting
* All access requires multi-factor authentication (MFA) through a zero trust network
* Access control configurations are managed as code, peer reviewed, and regularly audited

### Client data isolation

* Client data is logically separated and independently secured, with unique identifiers per client to prevent commingling
* Isolation is maintained through the full data lifecycle, from ingestion through processing and storage
* Reviewed annually as part of Dstillery's SOC 2 audit
* Client specific models operate on a zero trust, logically separated basis within a secure network; all data is encrypted in transit and at rest (AES-256 at rest)

### AI models and architecture

* DS-1 runs a pre-trained Google or Anthropic foundation model, with no fine-tuning or custom weights
* Capabilities come from prompt engineering and tool definition, not model training; the model is used for inference only
* A 0-day retention agreement is in place with the model provider
* No client or partner data is used to train or update Dstillery's global models or parameters; any client specific model training (e.g. joining client data with Dstillery data) is scoped exclusively to that client
* Prompts are version controlled in git and go through PR review; model upgrades are evaluated against internal benchmarks before shipping
* Generative AI is not used in model outputs. It may be used narrowly to help select or refine seed URLs during pre-campaign modeling, based on their position in text embedding space
* Underlying models are predictive (supervised machine learning), not generative, limiting unintended output behavior. Outputs are limited to a score or probability tied to conversion propensity for specific ad inventory

### Human oversight

* Every session is user directed; before any activation action executes, DS-1 presents the proposed action and waits for explicit confirmation
* All interactions are logged for monitoring
* An AI Governance Committee reviews models for fairness, bias, and performance before deployment
* Models are trained exclusively on URL level visit and conversion data. No demographic, behavioral, or sensitive attributes are collected, inferred, or provided to any model

### Security and monitoring

* LangSmith is used for tracing and observability across agent interactions, alongside manual log review
* MCP integrations are scoped to explicit tool schemas; agents cannot execute arbitrary operations
* Internal and external networks are separated by firewalls, with regular internal and external vulnerability scanning
* Dstillery maintains a formal Information Security Program (WISP) with regular employee security training
* Dstillery's Information Security Committee conducts annual formal risk assessments, with findings reported to the CEO and quarterly updates to the Board

***

### Data subject rights

* Individuals can submit Data Subject Access Requests (DSARs) for access, correction, or deletion through Dstillery's privacy page, managed via OneTrust
* Opt-out requests are honored by removing the individual's data from Dstillery systems and excluding them from downstream partner data sharing

### Incident response

* Real-time monitoring detects incidents; any incident involving personal data escalates immediately to the Data Protection Officer and Privacy Team
* Clients are notified of a Reportable Incident within 48 hours
* Clients retain sole authority over whether to notify individuals or regulators; Dstillery will not notify third parties without written client consent, unless legally required

### Sensitive category guardrails

{% hint style="warning" %}
Dstillery maintains a Sensitive Categories Policy governing how sensitive characteristics can and cannot be used in targeting:
{% endhint %}

1. Dstillery does not offer ID-based targeting products for sensitive categories. Custom AI and Custom Built user-based audiences will not be built around sensitive characteristics, and the standard prebuilt taxonomy does not include segments related to sensitive categories. Prebuilt segments may still be applied to any campaign where they're deemed relevant.
2. Dstillery has a review and approval process for targeting sensitive categories using products built with ID-free® Technology and activated via PMP or Custom Bidding Algorithm. This includes ID-free® Custom Built (seeded with domains) and Custom Search Lookalikes (seeded with search terms). Dstillery will not build ID-free Custom AI® targeting models seeded with first party data.
3. Sales or customer success representatives must submit a written request for approval on behalf of clients wishing to target sensitive categories.
4. Dstillery is committed to ongoing review of these policies beyond legal requirements. Approval standards are inherently subjective and will evolve over time, incorporating ethical judgment and Dstillery's comfort level based on the nature of each use case.

***

### Further reading

* Dstillery Trust Page: <https://dstillery.com/client-trust/>
* Dstillery Privacy Policy: <https://dstillery.com/privacy-policy/>


# Help

Find answers, reference DS-1 terms, or contact the Dstillery team.

### Browse the glossary

Look up DS-1 and audience terminology.

[**Open the glossary →**](/ds-1-foundations/help/glossary)

### Read the FAQs

Get quick answers to common questions.

[**Read the FAQs →**](/ds-1-foundations/help/frequently-asked-questions-faqs)

### Contact us

Reach the Dstillery team for additional support.

[**Contact the team →**](/ds-1-foundations/help/contact-us)

{% hint style="info" %}
New to DS-1? Start with [Getting started](/ds-1-foundations/create-your-first-audience). For product context, see the [Product overview](/ds-1-foundations/product-overview).
{% endhint %}


# Glossary

Your reference guide to common DS-1 terms.

Use this reference to understand common DS-1 terms and capabilities.

### Agent

A specialized workflow for a specific audience task. Agents are faster than open-ended chat for defined requests. Access them from the home screen or the **Agents** page.

### Custom AI Audience

An audience built from your first-party data. DS-1 models your customers and finds similar people across the open web.

### Custom Built Audience

An audience built from proxy signals, such as domains, keywords, or retail products. Use it when you lack first-party data but understand your target audience's behaviors.

### Multimodal AI

Dstillery's AI learns from different data types. It applies those learnings to create targetable audience segments.

### Project

A workspace for related conversations, files, and context. Projects keep history and files together for recurring campaigns or client work.

### Segrank

A predictive lift score that measures expected audience performance against random targeting. DS-1 uses these scores to surface high-performing audience categories.

### Workflow

A saved sequence of steps. DS-1 provides recommended workflows for common actions. You can also build and save workflows for repeated tasks.


# Frequently asked questions (FAQs)

Quick answers to common questions about DS-1.

### What's the difference between an agent and a workflow?

An **agent** is a specialized tool for one audience task. A **workflow** is a saved sequence of steps for recurring processes. Workflows can combine several actions or agents.

### I syndicated an audience, but it is not showing in my DSP. What now?

After DS-1 confirms syndication, allow time for your DSP's activation process. If segments still do not appear after that window, contact the Dstillery team.

### How do I know Dstillery audiences stay current?

Dstillery refreshes and scores segments every 24 hours. Targeting reflects the most current behavior.


# Contact us

We are here to help.

Have a question not answered here? Reach out to your Dstillery team or contact us directly for support, partnership questions, or campaign help.

<a href="https://dstillery.com/contact/" class="button secondary">Contact us</a>


# Get started

***


# Welcome


# DS-1 overview

***


# Projects


# History


# Activation


# Overview


# Programmatic (DSPs / SSPs)


# Social


# Quickstart


# Agents

***


# Overview


# Find prebuilt audiences

Pull from ready-to-activate audiences when you want something proven, and have it live in minutes.

## What prebuilt audiences are

Prebuilt audiences are Dstillery's off-the-shelf targeting library: **10,000+ audiences** spanning behavioral, demographic, search, and exclusive partner data. There's no build step and no modeling wait: you search the catalog, pick what fits, and syndicate to your DSP in the same session.

## Why they perform

These are not static third-party lists. Four things separate them from standard prebuilt segments:

* **Same multimodal AI as custom.** Built on the same multimodal AI engine and 400M+ device graph as our custom solutions: modeled from real behavior, not declared demographics.
* **Rescored every 24 hours.** Every audience refreshes daily, so the users you reach today were active yesterday. No spend wasted on stale data.
* **Platform-validated.** The library is a top-recommended data source in The Trade Desk's Audience Predictor, independent validation of its performance.
* **Activate any way you buy.** ID-based user segments, cookieless ID-free segments, curated deals, contextual, or custom bidding inputs, across all major DSPs.

## Walk through the agent

Five steps, one session, from a plain-English request to a live audience in your seat.

### Step 1. Describe the audience you're after

Type a specific objective or a general topic straight into the DS-1 chat box: a category, behavior, vertical, or brand. No taxonomy syntax required. Here: **"find me audiences related to luxury auto buyers."**

<figure><img src="/files/sZONy1iPSBXXmb9MerH3" alt="The DS-1 chat box with a request for luxury auto buyer audiences"><figcaption></figcaption></figure>

{% hint style="info" %}
Attach a brief or RFP with **Files**, or pull from a saved **Project**, and DS-1 will search the catalog against it.
{% endhint %}

### Step 2. Or open the agent directly

Rather start from the agent? On the DS-1 home screen, under **Recommended agents**, click **Find prebuilt audiences**, circled below. Either route lands you in the same Audience Explorer.

<figure><img src="/files/2tlJ1DWQbr5Nq2garWT0" alt="Recommended agents with the Find prebuilt audiences card circled"><figcaption></figcaption></figure>

Use *Build or refine an audience brief* instead when you're still shaping the campaign, or *See all* to browse every agent.

### Step 3. Review the grouped results

Matches land on the canvas at right, with chat and suggested next steps on the left. This run returned 189 audiences, which DS-1 sorted into four themed categories. The **bolded headers** circled below are those categories: DS-1 reads your search and clusters the catalog into the groupings that fit it best, so the most relevant angles surface first instead of one flat list.

<figure><img src="/files/eXYHzL3MydcozKSoADOD" alt="Matching audiences panel with the two bolded category headers circled"><figcaption></figcaption></figure>

* Read the group header and its rationale before the rows; it explains why those audiences were pulled together and what angle the category covers.
* The count badge next to each header is how many audiences sit in that category.
* Reach figures on the right are based on The Trade Desk's reported sizes and are estimates, so treat them as directional rather than exact deliverable scale.
* Confidence tiers trade scale for accuracy: **Extreme Confidence** is tightest, then **Precision** and **Active**.
* Not right yet? Use the suggestion chips to find similar audiences, swap a category, or hand off to the brief agent.

### Step 4. Select rows, then group them

Tick individual audiences, or a whole group header, and a bar appears at the bottom with deduplicated combined reach. **Group** merges the selection into one compound audience with OR logic; **Syndicate** sends each one out individually.

<figure><img src="/files/3pvysCwmSPzaitdN0Ws0" alt="Three selected rows circled, with the Group button circled in the selection bar"><figcaption></figcaption></figure>

Grouping returns a named audience and a Segment ID in chat: here, *Luxury Brands\_Custom Built*, Segment ID 1240838, at 49.6M combined reach.

### Step 5. Syndicate to your DSP seat

Confirm the marketer, advertiser account, and destination platform in the top bar, then syndicate. DS-1 confirms the handoff in chat and the audience is ready to target in the DSP. The session stays open, so keep searching, build another group, or start over.

<figure><img src="/files/tuW8Scv84dTyY092qBpx" alt="Chat confirming the audience was syndicated to The Trade Desk"><figcaption></figcaption></figure>

{% hint style="warning" %}
If nothing in the catalog fits the campaign, hand the request to the **audience brief** or **domain seeded audience** agent to have DS-1 model something custom instead.
{% endhint %}


# Build or refine an audience brief


# Build a domain seeded audience


# Build a search term seeded audience


# Build a purchase intent audience


# Analyze predictive lift scores


# Create a pixel


# Partnership Audience Search


# LiveRamp Data Marketplace


# Campaign Optimization (Coming Soon)


# Insights


# Data overview

Understand the data and AI that power Dstillery audiences and DS-1.

Dstillery’s data and modeling power everything we do.

This section explains what that data is and how it becomes audiences. Learn how our multimodal AI uses it, whether answering client questions or comparing solutions.

### From intelligence to action

DS-1 makes Dstillery intelligence actionable. This section stands on its own. It also provides the foundation for every audience built in DS-1.

Start with the data story. Then explore how the model learns and the sources behind it.

### In this section

1. [Dstillery data & DS-1: how they fit together](/data-and-audiences/data-overview/dstillery-data-and-ds-1-how-they-fit-together)\
   See where Dstillery data ends, DS-1 begins, and how they connect.
2. [How our multimodal AI works](/data-and-audiences/data-overview/how-our-multimodal-ai-works)\
   Explore the four core signal types and why learning them together matters.
3. [Our data sources](/data-and-audiences/data-overview/our-data-sources)\
   Learn about the opted-in panel and website visitation datasets.

{% hint style="info" %}
**Where to start:** Begin with **Dstillery data & DS-1: how they fit together**. Then dive into the model and data sources.
{% endhint %}


# Dstillery data & DS-1: how they fit together

See how consented panel data and website signals become targetable audiences.

## Signals in. Audiences out.

Dstillery turns observed behavior into audiences built for activation.

It starts with two datasets working together. One teaches intent. The other identifies its scale.

### 1. Opted-in panel data teaches

About 2 million fully consented users, sourced through panel partners, reveal how decisions take shape.

Their journeys show the research, comparison, and consideration that precede an outcome.

This data trains the model. It never targets an individual directly.

### 2. Website visitation data reaches

Website visitation data spans web and mobile activity across about 400 million devices.

It is sourced from bidstream data via SSPs and exchanges. It also comes from web publishing tools, including ad, commenting, and sharing widgets.

Once the model learns a high-intent pattern from panel data, website visitation data finds devices showing that same pattern.

Audiences refresh every 24 hours and become available for targeting.

**Opted-in panel data teaches. Website visitation data reaches.**

For a travel campaign, panel data might show future bookers researching destinations.

They compare reviews and revisit flight options before converting.

Website visitation data finds devices showing that same pattern. It makes them targetable in the client's DSP.

### From behavior to activation

{% stepper %}
{% step %}

#### Learn the journey

The model observes sequences, not isolated page visits.
{% endstep %}

{% step %}

#### Recognize similar intent

The model identifies devices showing comparable patterns.
{% endstep %}

{% step %}

#### Deliver a fresh audience

Audiences refresh every 24 hours. They are ready for activation in a DSP.
{% endstep %}
{% endstepper %}

### Where DS-1 comes in

These datasets are two of the four core signal types the multimodal AI model learns from.

The other signal types are LLM-derived insights and partner data.

See [How our multimodal AI works](/data-and-audiences/data-overview/how-our-multimodal-ai-works) for how the model connects all four.

For the full technical picture, including data partnerships and Dstillery's approach to data quality, see [Our data sources](/data-and-audiences/data-overview/our-data-sources).

DS-1 turns this intelligence into an audience on demand. Describe an objective, provide an input, or work from existing data.

DS-1 builds an audience with the control and visibility teams need.

**The result:** faster audience creation, clearer inputs, and targeting grounded in observed intent.


# How our multimodal AI works

See how Dstillery connects behavioral, language, and partner signals to understand real audience intent.

{% embed url="<https://www.youtube.com/watch?v=3NKepKcsQKw&t>" %}

People do not move through the internet in channels. They search, watch, read, compare, and return. Dstillery's multimodal AI turns those connected moments into one understanding of intent.

{% hint style="info" %}
**One model. More complete intent.** DS-1 reasons from the signals behind connected actions.
{% endhint %}

### See the journey, not the channel

Picture researching a new coffee maker. You Google "best espresso machines," browse a retailer's site, watch a review on YouTube, then see a CTV ad before buying.

Each moment adds context. Together, they reveal intent. Most targeting evaluates those moments separately. Dstillery's model learns how they relate over time.

### What multimodal means

Multimodal doesn't just mean more data sources. It means learning across different forms of information.

Words carry meaning. Behaviors form sequences. The model learns from both together, along with everything in between.

### Four signals. One intelligence layer.

{% stepper %}
{% step %}

#### Website visitation data

Billions of daily behavioral events across more than 400 million devices reveal patterns at scale.
{% endstep %}

{% step %}

#### Opted-in panel data

An opted-in panel of about 2 million users shows the behavior sequences that lead to real decisions, without relying on cookies or IDs.
{% endstep %}

{% step %}

#### LLM-derived insights

Large language models connect concepts across signals. They can link CTV viewership to web browsing, or turn a written audience description into targetable patterns.
{% endstep %}

{% step %}

#### Partner data

Specialized signals add depth for CPG, healthcare, B2B, and retail use cases.
{% endstep %}
{% endstepper %}

The first two signal types here are the same two datasets covered in [Dstillery data & DS-1: how they fit together](/data-and-audiences/data-overview/dstillery-data-and-ds-1-how-they-fit-together). For the full list of data partnerships behind signal four, see [Our data sources](/data-and-audiences/data-overview/our-data-sources).

### The shared language: embeddings

Embeddings let different signals work together. They map an advertising opportunity into a shared space of meaning, the same way a language model learns embeddings that capture what words mean.

That shared space connects a search term to a CTV pattern. It can turn a plain-language brief into behavioral signal. The model isn't comparing disconnected systems, it's working in one shared language for intent.

### Start anywhere. Activate everywhere.

Multimodal learning lets the model begin with almost any useful input:

* First-party data or a CRM list
* A search keyword or URL
* A plain-language audience description

It activates that same understanding through user segments, contextual targeting, private marketplace deals, and custom bidding.

### Built for measurable outcomes

This isn't a marginal improvement. It changes the quality of the signal behind each decision.

{% hint style="success" %}
One automotive brand's curated CTV deal hit a **98% video completion rate** against traditional ID-based targeting.
{% endhint %}

An auto insurance brand testing direct response found contextual targeting beat 12 other tactics, including ID-based lookalikes.

### The foundation for agentic advertising

Agentic advertising needs more than contextual inference. It needs a complete view of intent.

An AI agent reasoning from contextual signals alone would be missing behavioral signals, which are a much better predictor of intent. Multimodal understanding is what lets an agent like DS-1 reason across every available signal type at once.


# Our data sources

Explore the consented, scaled, and specialized signals behind Dstillery audiences.

Dstillery combines consented behavioral data with website activity at scale.

### How the data works together

#### Learn from consented journeys

**Opted-in panel data**

About **2 million fully consented users** reveal the journeys behind decisions.

Panel partners provide this data. It trains the model. It never targets people directly.

#### Recognize patterns at scale

**Website visitation data**

Web and mobile activity across about **400 million devices** makes learned patterns actionable.

This data comes from bidstream partners and web publishing tools.

For the full story, see [Dstillery data & DS-1: how they fit together](/data-and-audiences/data-overview/dstillery-data-and-ds-1-how-they-fit-together).

### Data sources

#### Opted-in panel data

Panel partners provide fully consented data from participants. Their browsing journeys show the research and comparison patterns that lead to an outcome.

The model learns sequences, not isolated clicks. It recognizes the difference between casual interest and meaningful consideration.

#### Website visitation data

Website visitation data provides the reach layer. It is sourced through SSPs, exchanges, and web publishing tools, including ad, commenting, and sharing widgets.

This dataset captures billions of daily behavioral events. It shows where comparable intent appears across the open web.

#### Specialized partner data

Specialized partners add depth for distinct campaign needs:

* **NielsenIQ/GfK** supports CPG and auto use cases.
* **PurpleLab** supports healthcare use cases.
* **ScreenEngine** supports media and entertainment use cases.
* **Resonate** supports consumer insights.
* **Emporia** supports B2B use cases.

Partner signals complement the core datasets. They help the model understand specialized contexts.

This is the detail behind the “partner data” signal type in [How our multimodal AI works](/data-and-audiences/data-overview/how-our-multimodal-ai-works).

### Custom audience seed signals

DS-1 can begin with the signal you already have:

* First-party data
* A domain
* Search terms or keywords
* Behavioral signals
* Retail purchase intent

These inputs help seed a custom audience. The model identifies related patterns for scalable activation.

### How audiences refresh

1. The model identifies behavior patterns associated with an outcome.
2. Website visitation data identifies devices showing similar behavior.
3. Audiences refresh every 24 hours and are ready for delivery.

### How Dstillery thinks about data

**Just for your brand.** Every campaign runs on a custom model tuned specifically to that brand, not a shared, generic model.

**Fresher data is better data.** Custom audiences refresh every 24 hours, so targeting reflects current behavior rather than a stale snapshot.

**Balance performance and scale.** Every audience is scored and ranked against Dstillery's entire universe of devices, tuned to the right balance for each campaign's goals.

### Responsible data use

Dstillery uses data to model intent, not to target a known individual.

Learn more about Dstillery's [Privacy and data](/ds-1-foundations/privacy-and-data).

### Learn more

* [Dstillery data & DS-1: how they fit together](/data-and-audiences/data-overview/dstillery-data-and-ds-1-how-they-fit-together)
* [How our multimodal AI works](/data-and-audiences/data-overview/how-our-multimodal-ai-works)
* [Privacy and data](/ds-1-foundations/privacy-and-data)


# Audiences

Choose, build, and activate Dstillery audiences.

{% hint style="info" %}
Use this page when a client asks how an audience product works.
{% endhint %}

### Find the right audience

Start with what you know. Then choose an audience for the campaign channel.

Build it in [Agents](/ds-1-guide/agents).

{% hint style="success" %}
**Quick path:** Start with your available signal. Then match it to the campaign channel.
{% endhint %}

### Start with your seed

A seed starts Dstillery's model. It can be CRM data, a pixel, conversion logs, a domain, keyword, product, purchase intent, medical claims, or patient search data. The model finds similar patterns across its device universe.

{% hint style="success" %}
Most products need a seed. Pre-built audiences activate without one.
{% endhint %}

Custom search lookalikes and retail purchase intent are specialized Custom Built audiences. Use Custom Built directly for other domain, keyword, or product seeds.

### Choose your path

Choose the product for the campaign. Then provide its required seed. Pre-built audiences are the exception.

{% columns %}
{% column %}

#### <i class="fa-bullseye">:bullseye:</i> Custom audiences from a seed

**First-party data**\
Choose [Custom AI audiences](/data-and-audiences/audiences/custom-ai-audiences) for CRM, pixel, or conversion data.

**No first-party data available**\
Choose [Custom built audiences](/data-and-audiences/audiences/custom-built-audiences) for a domain, keyword, or product.

**Search intent**\
Choose [Custom search lookalikes](/data-and-audiences/audiences/custom-search-lookalikes) to extend intent into programmatic.

**Retail and CPG shoppers**\
Choose [Retail purchase intent audiences](/data-and-audiences/audiences/retail-purchase-intent-audiences) for in-market shoppers.
{% endcolumn %}

{% column %}

#### <i class="fa-flag-checkered">:flag-checkered:</i> Specialized campaign audiences

**CTV campaigns**\
Choose [CTV-optimized audiences](/data-and-audiences/audiences/ctv-optimized-audiences) for CTV campaigns. Provide a seed to build the model.

**Healthcare and pharma**\
Choose [Custom patient targeting](/data-and-audiences/audiences/custom-patient-targeting) for healthcare and pharma campaigns. Provide medical claims or patient search data as the seed.

#### <i class="fa-bolt">:bolt:</i> No seed or build time

**Ready to activate**\
Choose [Pre-built audiences](/data-and-audiences/audiences/pre-built-audiences) to activate immediately, with no custom model or build time.
{% endcolumn %}
{% endcolumns %}

### Audience products

Explore the product guidance and pricing for every audience type.

<table data-view="cards"><thead><tr><th>Audience</th><th data-card-target data-type="content-ref">Product guidance</th></tr></thead><tbody><tr><td><strong>Custom AI</strong><br>Model next-best customers from CRM, pixel, or conversion data.<br><br><strong>Pricing:</strong> 30% of media, $2.50 cap, or $0.90 CPM.</td><td><a href="/spaces/ELPQcRUNChhv6VKohGYM/pages/5c19dNMuTfChzaklZ2ne">/spaces/ELPQcRUNChhv6VKohGYM/pages/5c19dNMuTfChzaklZ2ne</a></td></tr><tr><td><strong>Custom Built</strong><br>Build from a domain, keyword, or product signal.<br><br><strong>Pricing:</strong> 30% of media, $2.50 cap, or $0.90 CPM.</td><td><a href="/spaces/ELPQcRUNChhv6VKohGYM/pages/pWk3lVhpGgibQF7VgOWp">/spaces/ELPQcRUNChhv6VKohGYM/pages/pWk3lVhpGgibQF7VgOWp</a></td></tr><tr><td><strong>Custom search lookalikes</strong><br>Extend search intent into programmatic with a Custom Built audience.<br><br><strong>Pricing:</strong> TBD.</td><td><a href="/spaces/ELPQcRUNChhv6VKohGYM/pages/WTXaPa9EM1qQu3hWXjyi">/spaces/ELPQcRUNChhv6VKohGYM/pages/WTXaPa9EM1qQu3hWXjyi</a></td></tr><tr><td><strong>Retail purchase intent</strong><br>Reach in-market retail and CPG shoppers with a Custom Built audience.<br><br><strong>Pricing:</strong> 30% of media, $2.50 cap, or $0.90 CPM.</td><td><a href="/spaces/ELPQcRUNChhv6VKohGYM/pages/iEo71EeHk3DDyiRpVKkv">/spaces/ELPQcRUNChhv6VKohGYM/pages/iEo71EeHk3DDyiRpVKkv</a></td></tr><tr><td><strong>Pre-built</strong><br>Activate from 10,000+ ready-made audiences without a custom build.<br><br><strong>Pricing:</strong> 25% of media, $2.50 cap, or $0.90 CPM.</td><td><a href="/spaces/ELPQcRUNChhv6VKohGYM/pages/L6UnEyhUFqzhPeBDykAn">/spaces/ELPQcRUNChhv6VKohGYM/pages/L6UnEyhUFqzhPeBDykAn</a></td></tr><tr><td><strong>CTV-optimized</strong><br>Use audiences built for CTV campaigns.<br><br><strong>Pricing:</strong> TBD.</td><td><a href="/spaces/ELPQcRUNChhv6VKohGYM/pages/V7rmToikiqPTEfnUf6Jr">/spaces/ELPQcRUNChhv6VKohGYM/pages/V7rmToikiqPTEfnUf6Jr</a></td></tr><tr><td><strong>Custom Patient Targeting</strong><br>Build from medical claims or patient search seeds for healthcare and pharma campaigns.<br><br><strong>Pricing:</strong> 35% of media or $2.50 CPM.</td><td><a href="/spaces/ELPQcRUNChhv6VKohGYM/pages/t7pFqkoS6HY5oMHj771Y">/spaces/ELPQcRUNChhv6VKohGYM/pages/t7pFqkoS6HY5oMHj771Y</a></td></tr></tbody></table>

### How audience modeling works

This flow applies to seeded audiences. Pre-built audiences skip modeling and activate directly.

{% stepper %}
{% step %}

#### Select a seed

Choose the seed that best represents the audience.
{% endstep %}

{% step %}

#### Build the model

Dstillery's multimodal AI scores signals across 400M+ devices. It delivers the model to the client's DSP within 24 hours.
{% endstep %}

{% step %}

#### Keep it current

The model refreshes every 24 hours. Targeting stays current over time.
{% endstep %}

{% step %}

#### Deliver the model

Deliver the completed model to the client's buying platform.
{% endstep %}
{% endstepper %}

### Activate the audience

Once built, an audience can activate across these options.

{% columns %}
{% column %}

#### <i class="fa-crosshairs">:crosshairs:</i> Targeting options

**User segments**\
Daily refreshed, targetable identity-based audiences.

**Contextual**\
Real-time inventory scoring beyond page keywords.

**Curated deals**\
Private marketplace packages through SSP partners.
{% endcolumn %}

{% column %}

#### <i class="fa-sliders">:sliders:</i> Buying controls

**Custom bidding algorithms**\
Model intelligence applied to bid decisions.

**Social**\
Audience extension into social platforms.
{% endcolumn %}
{% endcolumns %}


# Custom audience types overview

***


# Custom AI audiences

***


# Custom built audiences

***


# Custom search lookalikes

***


# Retail purchase intent audiences

***


# Pre-built audiences

***


# CTV-optimized audiences

***


# Custom patient targeting

***


# Resources

***


# FAQs


# Glossary


# Privacy and data


# Contact us