# How to get started with Batch AI Predict? Batch AI Predict computes predictive scores from your existing customer data and automatically pushes them as native attributes on each Batch profile. These attributes (Churn dates, Products propensity, Future lifetime value, Product recommendations, etc) are then available directly in your segments, automations, and message templates, exactly like any other profile data in Batch. Refer to the [**Batch AI Predict score library**](/getting-started/features/customer-engagement-platform/batch-ai/batch-ai-predict.md) for a full overview of available scores. {% hint style="info" %} Batch AI Predict scores are computed from your historical customer data (purchases, product catalog, campaign interactions, web/app navigation). The more data available, the more accurate the predictions. {% endhint %} #### 1. Define your use cases Before diving into data preparation, the first step is to identify which scores make the most sense for your business in collaboration with our teams. This scoping session covers three things: understanding your current CRM challenges and priorities, mapping them to the relevant scores from the Batch AI Predict library, and deciding which use case to activate first based on your data availability and expected impact. This phase helps answer questions such as the following to identify the most relevant scores for your needs: * What are your main CRM challenges right now: churn, low repeat purchase rate, campaign performance, margin erosion from promotions? * Do you have consumable products with recurring purchase cycles? * What share of your customer base are one-time buyers? * Do you track promotion data at the transaction level? * What channels do you currently activate in Batch: email, push, SMS? #### 2. Prepare your data Once the needed predictive scores are selected, you need to make sure the required data is available and properly formatted. **Required for all scores:** * Transaction / sales history (customer ID, product ID, date, amount) * Product catalog (product ID + dimensions: category, sub-category, brand…) **Optional - improves prediction accuracy:** * Web and app navigation events (e.g. product page views, add-to-cart events) * Campaign interaction history (e.g. opens, clicks) * Customer profile attributes (e.g. sociodemographic information) Your Batch Solutions Engineer will help identify what additional data flows or imports need to be set up to feed the models and will support you in implementing them if needed. Note that Batch AI Predict requires both an initial full data load and recurring syncs to keep scores up to date as your customer base evolves. #### 3. Model configuration by Batch Once your scores are known and your data is ready, Batch will configure the corresponding ML models. The configuration process typically starts with defining what the model should predict and on which customer population. From there, the team will run a first version of the model and review the results, looking at how scores are distributed across your customer base and whether the output aligns with what you'd expect from your business knowledge. Based on that review, adjustments may be made to improve accuracy and relevance before scores are pushed to your profiles. This part of the process is fully handled by the Batch team. {% hint style="info" %} The time required for this phase depends on the complexity of the use case and the quality of the data available. Your Solutions Engineer will keep you informed at each step. {% endhint %} #### 4. Review your first scores Once the model has trained on your data, scores are automatically pushed to your Batch customer profiles as custom attributes. You can verify them directly from the profile view in the Batch dashboard. #### 5. Activate your first campaign With scores live on profiles, you can start using them in Batch immediately: * **In segmentation:** filter your audience on any predictive attribute in the targeting section of your orchestrations. *Example: customers with a churn date in the next 60 days, or customers with a product affinity score above 0.6 for category X.* * **In automations:** use a predicted score as a journey entry trigger by setting a score threshold as an entry condition. *Example: enter high-potential prospects into a conversion journey as soon as their likelihood-to-purchase score exceeds a defined level.* * **In message content:** inject predictive attributes as personalization variables in your message templates. *Example: display the top recommended product dynamically in an email banner.* {% hint style="success" %} When setting up your first campaigns with Batch AI Predict, defining how you will measure performance upfront is just as important as the activation itself. The right measurement approach depends on how you are using the score. For instance: * **Audience targeting:** Compare three populations: your business-defined audience, the score-based audience, and their intersection and measure the uplift across the two main segments. * **Content personalization:** Run an A/B test on the same segment, varying only the message content to isolate the impact of personalization. * **Trigger-based automations:** Use a control group that does not receive the campaign, and measure the uplift against the triggered population. {% endhint %} #### Next steps Once your first score is live and your first campaign is running, you can progressively expand to additional use cases: product recommendations, promotion sensitivity, future LTV prioritization, and more. Refer to the **Batch AI Predict score library** for a full overview of available scores. --- # Agent Instructions: Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question. Perform an HTTP GET request on the current page URL with the `ask` query parameter: ``` GET https://doc.batch.com/getting-started/other/implementation-guides/how-to-get-started-with-batch-ai-predict.md?ask= ``` The question should be specific, self-contained, and written in natural language. 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.