Introduction
If you've landed on this article, you're likely looking to implement one of the best products that Google Cloud offers to retailers, Discovery AI. If you've started reading through the product documentation, you will have learned how important it is to import your catalog data and user events correctly, define attribution tokens for each sale, and authenticate to the API securely.
The three most common approaches to activating Cloud Retail Search and Recommendations AI on storefronts include building custom implementations, buying a fully-managed solution, or using an integration tool like Retail Cloud Connect. This article focuses on examples of each, pros and cons, and aims to help you determine the best path forward for your organization.
DIY or Custom Implementation
A DIY or custom implementation assumes writing the underlying ingestion code and ETL pipelines required to import your catalog and user event data into Google Cloud on a schedule or in real-time. Additionally, you must write custom code that matches your architecture to deliver results on your ecommerce storefront. In addition to writing the code, you must manage the underlying infrastructure and incorporate scaling considerations (typically an elastic solution) required for fluctuations in site traffic such as Black Friday or Cyber Monday holidays.
When is this the best approach?
The ROI and benefits of building a custom solution are often tied to many other build vs. buy conversations in your organization. If you are running on a unique or home-grown ecommerce platform, have expert developers in house or through a trusted agency, or want custom functionality that is not built into Google's underlying Retail API, then a custom implementation is most likely the best option.
What are the risks to this approach?
The team implementing the solution must become and remain experts in Google Cloud’s Discovery AI tooling. A custom approach will also take much longer to implement than an off-the-shelf solution and the project timeline must account for the learning and onboarding of the new tools. This will also become a key piece of critical infrastructure for SREs or infrastructure teams to manage, ultimately sitting between your users and their search results, so the cost of an outage could be substantial depending on how much revenue your ecommerce business drives.
Key Points
You develop tooling and code unique to your organization
Implementation is completed by specialized partners or in-house developers
You manage the underlying ingestion and serving infrastructure
Longest time to implement and usually the most expensive
Fully-Managed Solution
Some companies have built fully-managed solutions that white-label or wrap Google Cloud’s Retail API. These platforms give you the opportunity to benefit from Google’s expertise in AI and ML while deploying a purpose-built solution focused on product search and discovery.
When is this the best approach?
Purchasing a fully-managed solution is the best approach when you don’t want to deal with any of the intricacies of Google Cloud or the data in the underlying platform. Some fully-managed solutions also offer additional functionality such as data enrichment and user-friendly merchandising tools aimed at business users, aiming to supplement gaps in Google's own products.
What are the risks to this approach?
Any new Google Cloud platform functionality may take weeks or months to find its way to the solution that you purchased. Similarly, you are limited to the Google Cloud functionality that your vendor chooses to integrate, and there may be catalog optimization or Recommendations AI model training opportunities for your specific catalog data that you are unable to take advantage of. Depending on the complexity of your vendor's solution, your product catalog, and ecommerce platform, the deployment timeline may not be faster than leveraging Google Cloud directly or building a custom solution.
Key Points
The fully-managed solution is responsible for importing your catalog data into a proprietary platform and serving results on your storefront
The vendors wrap their software and services around Google Cloud’s Retail API
There is a potential delay between new Google functionality and vendor platform functionality
You do not have a path to direct support from Google during an outage
There is no infrastructure for you to manage
Implementation times vary depending on catalog complexity and ecommerce platform
Retail Cloud Connect
Retail Cloud Connect is a SaaS tool that helps retailers quickly enable Google Cloud’s Discovery AI solutions on major ecommerce platforms such as Shopify.
When is this the best approach?
You are running on a major ecommerce platform like Shopify and want to implement Google Cloud’s Discovery AI solutions quickly. You want to maintain full control of the underlying Google Cloud Discovery AI platform and a direct relationship with Google Cloud. If you plan to leverage Google Cloud for additional solutions, such as performing BigQuery analyses over marketing data or building customer data platforms, then having all of your data in one place can be an attractive outcome. Total cost of ownership is often lower because pricing can be negotiated with Google if you are a high-volume customer.
What are the risks to this approach?
Retail Cloud Connect is limited to major ecommerce platforms and is only available on a handful of platforms at this time.
Key Points
Retail Cloud Connect is responsible for importing your catalog data to Google Cloud and serving results on your storefront
You maintain a direct relationship with Google Cloud, have full access to the underlying Google Cloud console, and maintain full ownership and control of your data
You are provided with immediate access to Google’s new features and functionality inside of the Google Cloud console
You can contact Google Cloud support directly if required
There is no infrastructure for you to manage
Fastest implementation time if you already use a major ecommerce platform
