--- title: "Predictive delivery times" canonical: "https://ingrid-support.refined.site/space/KB/444497922/Predictive%20delivery%20times" format: markdown --- **Delivery times shoppers trust at checkout** --- ## What is it? **Predictive Delivery Times** is a feature within Ingrid Checkout that replaces static, manually configured delivery windows with AI-predicted delivery times calculated for each individual order. Instead of showing a fixed "2–5 business days" for all orders, the system predicts a delivery window for the specific order being placed — accounting for the destination, origin warehouse, shipping method, and current carrier and fulfilment performance. The result is a delivery promise that reflects what will actually happen, not a conservative estimate padded for safety. Predictions adapt automatically as conditions change. When carrier performance improves, windows tighten. When delays emerge — during peak season, for example — windows adjust accordingly. No manual updates required. --- ## Who is it for? Predictive Delivery Times is designed for: - **E-commerce and delivery experience managers** responsible for how delivery options are presented at checkout - **Customer service and operations teams** dealing with delivery-related support volume - **Retailers who want to improve checkout conversion** without changing their carrier setup It delivers the most value for retailers with: - Wide, conservative delivery windows used to avoid missed promises - Multiple carriers, shipping methods, or destinations with different performance profiles - High delivery-related support volume (WISMO contacts) --- ## How it works When a shopper reaches the delivery selection step, Ingrid calculates a predicted delivery window for that specific order — based on the destination, origin warehouse, shipping method, and real historical performance data from your carriers and fulfilment operations. The predicted window replaces your static delivery times in checkout. Shoppers see "Arrives Friday" instead of "3–5 business days". Predictions update automatically over time. As carrier performance changes — whether improving or deteriorating — the model adjusts without any manual input. **Key outcomes:** - Shoppers see specific, realistic delivery times instead of broad ranges - Delivery windows tighten safely where performance data supports it - Promise accuracy improves, reducing delivery-related support contacts - No manual maintenance of delivery rules as conditions change --- ## Requirements | Requirement | Details | | --- | --- | | Ingrid Checkout | Version 2 required (V1 is evaluated case-by-case) | | Tracking events | Tracking events must feed into Ingrid — this is required for predictions to work. | | Order volume | Best results with more than 1,000 orders per shipping method in the past 13 months | | Markets | Currently available for shipments from and to Nordic countries (SE, NO, DK, FI). Expanding to EU in Q2 2026. | --- ## How to Set Up Predictive Delivery Times is self-serve and can be enabled directly in IMP (Ingrid Management Platform) without involving engineering. **Step 1: Enable the feature** 1. Go to IMP → Delivery Checkout → Features. 2. Select the relevant country or configure as the site-wide default. 3. Under Checkout Configuration, toggle on **"Use predictive delivery times"**. ![Screenshot 1 - Global feature setting.png](media://bce73183-271d-40e9-84cc-cf0dd2c06bda) **Step 2: Exclude specific delivery categories (optional)** If there are specific delivery categories where you prefer to keep static delivery times, you can disable the feature per category without affecting the global setting. 1. Open the relevant delivery category in IMP. 2. Toggle on **"Disable predictive delivery times"** for that category. ![Screenshot 2 - Disable on ctaegory level.png](media://ad4c741e-4ca3-46a8-b362-bf52a944e96d) **Step 3: Test on production** We recommend running an A/B test on production to validate the impact before full rollout. Predictive Delivery Times is not available in stage environments. --- ## Frequently Asked Questions - **What data does Ingrid use to make predictions?** Predictions are based on historical delivery performance data — actual carrier tracking events for past shipments. The model accounts for destination, origin warehouse, shipping method, and observed carrier and fulfilment performance over time. - **What happens if a prediction can't be made?** If Ingrid cannot generate a prediction for a specific order — for example, due to insufficient data for a route — the feature falls back gracefully to your existing static delivery settings. Shoppers always see a delivery time. - **Do I need to update the model manually as conditions change?** No. Predictions adapt automatically as new tracking data comes in. There is no manual maintenance required. - **Can I keep static delivery times for some methods?** Yes. You can disable Predictive Delivery Times per delivery category in IMP, without affecting the global setting. This lets you roll out gradually or keep specific methods on static times. - **Is this available in stage for testing?** No. Predictive Delivery Times is only available in production environments. We recommend running an A/B test to measure impact before full rollout. - **How do I measure impact?** Key signals to track: - Conversion rate at the delivery selection step (compare PDT vs. static windows via A/B test) - Cart abandonment rate at delivery selection - Delivery promise accuracy — % of orders delivered within the predicted window - Delivery-related support volume (WISMO contacts) before and after rollout - **Is there a cost to use Predictive Delivery Times?** Predictive Delivery Times is part of Ingrid AI, which is available free of charge until 31 May 2026. Pricing takes effect from 1 June 2026. --- ## Additional Notes **Part of Ingrid AI:** Predictive Delivery Times is part of Ingrid's AI intelligence layer — a set of features that use real delivery data to make smarter delivery decisions at checkout. The other feature in Ingrid AI is [Profit-Optimized Delivery](https://ingrid-ab.atlassian.net/wiki/spaces/KB/pages/454918145), which ranks delivery methods at checkout based on estimated shipment cost to improve margin per order. ---