> For the complete documentation index, see [llms.txt](https://docs.mediarithmics.io/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.mediarithmics.io/setup-and-monitor/plan-and-integrate/choose-your-deployment-mode.md).

# Choose your deployment mode

Decide whether your datamart stores its own data, reads from your cloud data warehouse, or does both.

## Your goal

Decide where the data of record for your project lives, and where the computation on it runs: inside mediarithmics, inside your own cloud data warehouse, or split between the two.

## Expected outcome

One deployment mode agreed for each datamart in scope, and a shared understanding of what that choice implies for collection, for identity resolution, for segmentation, and for which platform modules you can use.

## When this decision arises

* At the start of a new integration, before any collection is configured.
* When your organisation moves its reference data into a cloud data warehouse and wants the CDP to read from it rather than hold a second copy.
* When a new datamart is created for a business unit whose data architecture differs from the rest of the group.

{% hint style="info" %}
The modes that rely on a cloud data warehouse are in Alpha. Contact your account manager to learn more.
{% endhint %}

## Composable, and what it is not

The two terms are used interchangeably in the market, and they do not mean the same thing.

* **Composable architecture** is the platform's ability to interface natively with your cloud data warehouse - [Google BigQuery](https://app.gitbook.com/s/XSPACE_CONNECTORS/sources/google-bigquery) or [Snowflake](https://app.gitbook.com/s/XSPACE_CONNECTORS/sources/snowflake) - and to treat it as the source of truth. It is a capability of the platform, not a mode you run in.
* **Zero-copy** is one way of operating composably: no data is ingested into the platform, and queries and calculations are executed directly on the warehouse instead.

{% hint style="success" %}
A CDP can be composable without being zero-copy. Being composable is the technical prerequisite for operating in zero-copy.
{% endhint %}

## The three modes

The deployment mode is a property of a **datamart**, not of your organisation. Two datamarts in the same organisation can run in different modes.

| Mode          | Where the data of record lives                                                                                                        | Where computation runs                                                              | Suits                                                                                                       |
| ------------- | ------------------------------------------------------------------------------------------------------------------------------------- | ----------------------------------------------------------------------------------- | ----------------------------------------------------------------------------------------------------------- |
| **Hosted**    | In the [datamart](/setup-and-monitor/plan-and-integrate/set-up-the-environment/prerequisites-organisation-and-datamart.md#datamarts). | In the platform.                                                                    | Organisations with no cloud data warehouse, or no intention of making one the reference for marketing data. |
| **Hybrid**    | In your warehouse.                                                                                                                    | In your warehouse, with the data needed for personalisation cached in the platform. | Organisations that want warehouse governance without giving up real-time use cases.                         |
| **Zero-copy** | In your warehouse.                                                                                                                    | In your warehouse only.                                                             | Organisations whose priority is that marketing data is never duplicated outside their own environment.      |

In **Hybrid** and **Zero-copy** alike, the warehouse holds the source of truth. That single fact drives every consequence below.

## What zero-copy gives you, and what it costs

{% columns %}
{% column %}

### What it gives you

* **Freshness.** Data is accessed where it is produced, so there is no synchronisation delay between the reference copy and the copy you query.
* **Rationalisation.** No duplicated storage, and no pipelines to build and maintain between your warehouse and the platform.
* **Governance.** Your data stays inside your own secure environment.
  {% endcolumn %}

{% column %}

### What it costs you

* **Functional limits.** Real-time personalisation generally requires some data to be held in the platform, which a strict zero-copy deployment does not allow.
* **Compute cost.** Moving segmentation and identity resolution onto the warehouse moves their cost there too, and warehouse compute is billed by use.
  {% endcolumn %}
  {% endcolumns %}

**Hybrid** exists for exactly this trade-off. It keeps the warehouse as the source of truth and caches only the data your individual activation use cases need, which is what limits replication to the minimum those use cases require.

## The consequences to plan for

### Live tracking has to reach your warehouse

In Hosted mode, tracked activity is written to the datamart and stays there. When the warehouse is the source of truth, the raw activity has to be there instead, because that is the data identity resolution is computed from. Tracking therefore transits through the platform and is stored in your warehouse.

The tag on your sites and apps does not change. What changes is where the data comes to rest, and what happens to activity that does not fit your schema.

[Replicate live tracking to your warehouse](/setup-and-monitor/collect-and-manage-your-data/sync-with-your-data-warehouse/replicate-live-tracking-to-your-warehouse.md)

### Your schema decides what the platform can see

The warehouse can hold far more than your marketing use cases need. Your schema is where you declare the subset that becomes queryable and segmentable, by mapping warehouse tables onto the objects of your data model.

[Map your warehouse tables to a schema](/setup-and-monitor/collect-and-manage-your-data/design-your-data-model-and-schema/map-your-warehouse-tables-to-a-schema.md)

### Segmentation and identity resolution move, but your work does not

You keep building segments in [OTQL](/setup-and-monitor/operate-and-monitor/query-your-datamart/otql.md), in the same builders. The query is translated to SQL and executed on your warehouse.

[Run your computation in the warehouse](/setup-and-monitor/collect-and-manage-your-data/sync-with-your-data-warehouse/run-your-computation-in-the-warehouse.md)

### Retention becomes a shared responsibility

Data the platform writes into your warehouse is subject to the retention the platform manages. Data you hold in your own warehouse and expose to the platform is retained according to your own rules, not the platform's [cleaning rules](/setup-and-monitor/collect-and-manage-your-data/clean-and-delete-data/cleaning-rules.md).

Decide, per table, which side owns the deletion obligation before you connect anything.

### Some modules require data in the platform

A module that has to answer in a few milliseconds about a single user cannot wait for a warehouse query. Those modules require a **Hosted** or **Hybrid** deployment, because both keep the data they need inside the platform.

{% hint style="warning" %}
Feature availability by mode is the part of this decision that is easiest to discover too late. Before you commit to zero-copy, list the use cases that depend on real-time personalisation and ask your account manager to confirm, module by module, which of them your intended mode supports.
{% endhint %}

## How to decide

{% stepper %}
{% step %}

### Establish whether you have a warehouse that is a source of truth

Not simply a warehouse: one your organisation treats as the reference for the data your marketing use cases need. Without it, Hosted is the answer and the rest of this page is context.
{% endstep %}

{% step %}

### List your use cases, and mark the ones that answer in real time

On-site personalisation, individual activation on the next page view, and anything that reacts within a session belong in this list. They are what separates Hybrid from Zero-copy.
{% endstep %}

{% step %}

### Decide who owns warehouse compute cost

Segmentation and identity resolution will consume warehouse credits on a schedule. The team that owns that budget has to be part of this decision, not informed of it afterwards.
{% endstep %}

{% step %}

### Confirm the mode per datamart

Record the answer against each datamart in scope, and take it into [Prerequisites - Organisation and Datamart](/setup-and-monitor/plan-and-integrate/set-up-the-environment/prerequisites-organisation-and-datamart.md) as an input to environment setup.
{% endstep %}
{% endstepper %}

## How you know the decision is made

* Every datamart in scope has one named mode, and one named owner for the choice.
* You can say, for each real-time use case, whether the chosen mode supports it.
* You know which side owns retention for each table you intend to expose.
* The team owning warehouse compute has agreed to the recurring cost.

## Related pages

<table data-view="cards" data-search="false"><thead><tr><th></th><th></th><th data-hidden data-card-target data-type="content-ref"></th></tr></thead><tbody><tr><td><strong>Frame your CDP project</strong></td><td>The other decisions this one sits alongside.</td><td><a href="/setup-and-monitor/plan-and-integrate/frame-your-cdp-project.md">Frame your CDP project</a></td></tr><tr><td><strong>Sync with your data warehouse</strong></td><td>Connecting the warehouse, and what the platform reads from it.</td><td><a href="/setup-and-monitor/collect-and-manage-your-data/sync-with-your-data-warehouse.md">Sync with your data warehouse</a></td></tr><tr><td><strong>How mediarithmics works</strong></td><td>The datamart, the UserPoint, and how data flows in and out.</td><td><a href="/setup-and-monitor/understand-mediarithmics/how-mediarithmics-works.md">How mediarithmics works</a></td></tr></tbody></table>


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation by asking a question.

Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters:

```
GET https://docs.mediarithmics.io/setup-and-monitor/plan-and-integrate/choose-your-deployment-mode.md?ask=<question>&goal=<user_goal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case.

The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.