Shopping Muse leverages AI technology to offer users a unique product discovery experience through interactive conversations and personalized product guidance. Powered by advanced deep learning, the system engages your users with natural language conversations. Whether they're searching for specific products, fashion styles, or gifts for special occasions or holidays, Shopping Muse provides tailored suggestions to meet their needs.
Onboard Shopping Muse
- Open the Shopping Muse app, and then click Create Muse.
- A wizard appears to guide you as you customize various aspects of Shopping Muse behavior, appearance, and functionality. This includes tone of voice, the Shopping Muse icon, your brand name and colors, and the app's title.
Upon successful completion of this process, the Shopping Muse dashboard panel appears. The next steps are to customize Shopping Muse, preview it on your site, and then implement it so your users can begin their Muse experience.
Customize Shopping Muse
Before going live, customize Shopping Muse to ensure it's completely aligned with your brand guidelines. Click Customize to set advanced customizations to the look and feel, and to edit assistant behaviors you defined during the onboarding process.
Use the following tabs to customize Shopping Muse:
-
General: Customize the overall design, including:
- The assistant icon
- The colors and font
-
The welcome screen layout and strategy:
- Layout: Grid or basic
- Strategy: The strategy that powers the welcome screen grid. This strategy has no effect on the Shopping Muse recommendations logic, including the filters that are applied to it.
- Search by image: Enable/disable uploading an image and searching for similar items. You can also set the icon for the image upload.
-
Product tile: Customize how each product tile looks in the recommendation widgets, including:
- Main and secondary texts
- Price
- Previous price: Use when you offer a sale price and want to compare it to the "old" (higher) price. You must have a column in your feed with the old price. Shopping Muse always displays the price from the "Price" column as the current product price (for a sale, this is the new price. For a regularly priced item, it's the only price displayed).
- Product Preview: New Tab opens the preview in a new browser tab, and Quick View generates a light version overlay of the product page. You can define elements to hide from the quick view.
- Currency localization: Customize how currencies are displayed, including formatting and symbols.
Quick view
Quick view uses your website's native product page. It displays the product page in an overlay and conceals elements that are not necessary for the quick-view experience. For example, including "header" and "footer" in the product customization settings hides the top and bottom portions of the product page.
It's important to note that every website is unique, and assistance from a developer or our customer support might be needed to achieve the desired experience.
- Set your default language
-
General:
- Assistant name
- Input field placeholder
- Similar items response: The text that appears when a user asks for additional products similar to a given product (for example, "Here are some similar items").
- Add to cart response: The text that appears when the user adds an item to the cart.
- Inspirational suggestions: Messages in the Assistant grid suggesting searches for the user.
- Disclaimer: Disclaimer text and links.
-
General:
- Assistant at: Your brand that the assistant represents. We recommend you use the full URL of your brand for clarity (www.mybrand.com). Avoid using placeholders or fictitious names.
- Tone of voice: Define the assistant's conversational style.
- Customer care: If your user attempts to ask your Assistant customer care questions, this text and added links direct them how to contact customer support.
Notes:
- Shopping Muse training is unsupervised. This means that if the column names are not human-readable (for example, they use "g" for gender or internal abbreviations), the training is likely to be ineffective and might even harm the results.
- There can be up to 5 columns, each with up to 10 values. So it's important to use columns efficiently. For example, it isn't necessary to train on every category. Rather, focus on the more complex ones where the Muse needs reinforcement, such as "Swimwear" vs. "Underwear".
Preview your Muse
Depending on the size of your product catalog, Shopping Muse can take up to 24 hours to complete the initial training on your products. During this process, Shopping Muse learns about your brand and products and how to recommend them. When training is complete, the Preview button becomes available in the Shopping Muse main dashboard.
Click Preview to open your website in a new tab, with the Shopping Muse experience available for your testing and quality approval.
Implement Shopping Muse and go live
For a streamlined onboarding and implementation experience, Shopping Muse is designed to activate automatically every time a click is tracked on a designated HTML tag (<dy-chat>). There are two ways to implement this tag: Directly in the website codebase by your developers or using a Dynamic Yield campaign.
Using a Dynamic Yield campaign
With this method, you implement the Shopping Muse <dy-chat> tag through a Dynamic Yield experience. To simplify the process for you and reduce costly development time, we've designed out-of-the-box templates that expedite the implementation effortlessly.
Using manually embedded code
Using this method, you manually embed the HTML tag <dy-chat> to wrap your native entry point and open the Shopping Muse interface upon click.
Notes:
- As soon as you set up an entry point and activate it (via Dynamic Yield or an HTML tag), Shopping Muse is accessible to users who click the corresponding experience.
- We highly recommend using Dynamic Yield experiences to implement Shopping Muse, as it streamlines the process of implementation, managing, and monitoring.
- To optimize Shopping Muse's use as a conversational discovery tool, we highly recommend integrating Shopping Muse entry points across various areas aimed at enhancing seamless product discovery. These areas include the search bar, floating notifications, menu options, slim banners on the homepage, and category pages.
Reporting
The Shopping Muse report enables you to track and monitor the performance of important e-commerce KPIs (key performance indicators). These include sessions, revenue, purchases, add-to-cart events, and average order value (AOV). In addition, you can keep track of product clicks and conversions performed by users, helping you gain valuable insights into user behavior and campaign effectiveness.
FAQ
Shopping Muse suggestions are influenced by the type of images it's trained on:
-
Wild Images: Product images featuring human models in atmospheric or inspirational backgrounds, often including additional products. For example, a picture of a model wearing a pair of jeans that also clearly shows sunglasses, a shirt, and shoes. In this case, the algorithm focuses on recommending items that match the overall style and look represented in the image.
If you use wild images, we suggest applying recommendation filters to align with the specific category of the item in context. - Studio Images: Clear images of the product against a clean background. The algorithm prioritizes recommending similar products based on attributes like color, texture, material, and style. Studio images typically yield better results.
Note: By default, the model processes the image_url column in your catalog. If you want to use different images, contact your Customer Success Manager or Support for help with custom configuration.
- The model requires access to your image CDN to download and process your image. Have your technical team add the following IP addresses to your allowed list:
-
- 8.197.234.165
- 3.127.189.219
- 18.158.108.19
-
- 18.192.189.87
- 18.195.163.252
- 3.92.38.28
-
- 3.227.221.249
- 54.145.37.112
- 75.101.154.167
-
- If the CDN does not respond within 5 seconds for a specific image, that image is skipped.
- Maximum image size is 10MB
- The image path must be explicit. For example:
Correct: https://mydomain.com/myimage.jpg
Incorrect: //mydomain.com.myimage.jpg or //myimage.jpg
The model we use for the Shopping Muse response is proficient in many languages. However, its performance tends to be best in the following languages:
-
- English
- Spanish
- French
- German
-
- Dutch
- Italian
- Portuguese
-
- Russian
- Chinese (Simplified)
- Japanese
Shopping Muse can handle most other languages with varying degrees of proficiency, including Arabic, Polish, Hebrew Korean, Turkish, and many more. The model’s capabilities are extensive but can vary depending on the complexity and nuance required in a specific language.
Product recommendation quality does not vary between languages.
Shopping Muse incorporates multiple safeguards to ensure its responses are always brand-safe and aligned with your values. The model is trained specifically for each brand, guaranteeing consistency in voice and tone. Moreover, Shopping Muse only recommends products that are pre-approved by your team.
In addition, Shopping Muse employs a micro-service architecture. The product recommendation task, responsible for suggesting relevant products to the user, and the assistant task, which responds in natural language, are entirely separate. The LLM does not have access to product meta data. Therefore, Shopping Muse can't, for example, incorrectly claim that a specific product is available for free due to any penetration tactics.
All personalization calculations are done on Dynamic Yield servers and comply with regional limitations such as GDPR. Users' personal data is not shared with the LLM or third parties.
Dynamic Yield has a zero retention policy. This means that no data is retained.
No. No specific brand or user data is used for training Shopping Muse.
JPEG, GIF, HIEC, and PNG. The maximum image size allowed is 5MB.
No, Dynamic Yield does not save any of the images uploaded by users. We encode the images into a base64 representation and search for similar-looking products. When the process is complete, we don’t save either the image or its base64 representation.