G2

Bilal M.

Research and Development Engineer

Medical Devices

Mid-Market (51-1000 emp.)

8/18/2026

"Lightning-Fast, Smart Real-Time Recommendations with Hands-Off Scaling"

4.5/5

What do you like best about Google Cloud Recommendations AI?

What I like most about Google Cloud Recommendations AI is that its intelligence is genuinely smart at spotting unusual user patterns that we definitely would have missed manually. We regularly use the "Others You May Like" and "Frequently Bought Together" models, and the way it works with real-time data is impressive; it adapts recommendations almost instantly as users browse. Performance-wise, the API calls are lightning-fast, which is critical for our site speed. On top of that, the fact that it automatically handles model training and autoscaling means my engineering team doesn’t have to spend hours babysitting infrastructure, and that has massively improved our deployment workflow.

Getting the initial integrations set up was also very smooth because it plays nicely with BigQuery and Google Analytics 360. That meant we could pipe in our existing historical data without having to build an entirely new ETL pipeline. One unexpected benefit I found was how well it handled cold-start scenarios for new products with barely any traffic; it uses product metadata in a way that’s clever enough to start serving solid recommendations right away. In terms of pricing and ROI, the lift in average order value and conversion rate we saw almost immediately delivered a strong return on investment. Onboarding was straightforward thanks to the documentation and the console UI, although configuring IAM roles for different team members was a bit finicky at first. Review collected by and hosted on G2.com.

What do you dislike about Google Cloud Recommendations AI?

What I dislike about Google Cloud Recommendations AI is that its pricing structure is extremely complex and can get expensive very quickly, especially since you pay separate fees for data ingestion, model training, and then per-thousand prediction requests. On the UI/UX side, the main dashboards look clean, but fine-tuning serving configs and troubleshooting why certain products are being recommended is tedious because there isn’t much visibility into things like model weights or feature importance. Performance and overall AI “intelligence” also vary a lot by product category. If your catalog includes many low-traffic items or short-lived seasonal products, the models can struggle with accuracy and often fall back to generic recommendations unless you add extra manual filtering logic, which defeats the purpose of using an automated solution in the first place. Review collected by and hosted on G2.com.

What problems is Google Cloud Recommendations AI solving and how is that benefiting you?

Before we started using Google Cloud Recommendations AI, our product suggestion setup was a complete mess: static, hardcoded rules and basic “top sellers” widgets that showed the exact same items to everyone. With a small engineering team, we were constantly bogged down manually updating product carousels every week. It was tedious work, and engagement stayed poor because the suggestions weren’t relevant to what people were actually browsing in the moment.

We used to rely on generic recommendations that took hours of manual curation each week, but now we can automatically serve personalized, real-time product suggestions powered by Google’s ML models. That shift has led to an 18% boost in our average order value (AOV) and a big jump in click-through rates. Implementation also saved our developers about 10–12 hours every week on manual merchandising tasks. And since the models retrain automatically and handle cold-start items on their own, our conversion rate on checkout-page upsells increased by nearly 22% within the first three months. Review collected by and hosted on G2.com.