Real-time Segment and Amplitude Connectors
This release includes new real-time analytics connectors to your favorite customer data platforms (CDP) and analytics applications. Our Segment and Amplitude connectors have taken initial model creation down from a few days to 10 minutes. The great thing about these connectors is that you can stream the events in real-time, which powers our session-reranking models and gives your product a more dynamic, reactive feeling and that will engage your users. We’ll have more information about these connectors coming out in the next few days, but take a look at our real-time connectors docs to see how to get setup today!
Shaped Evaluation Dashboard
Understanding what your end users are seeing and the macro analytics of your product, platform or marketplace is critical when using personalized AI algorithms like Shaped. Before integrating, it’s important to simulate how your users’ experience will change and ensure that it aligns with your product’s vision or discovery experience expectations. After integrating into production, A/B testing and monitoring online performance uplift is necessary so that you can understand the value personalization brings and correctly communicate this value to key stake-holders.
All of this is why we built the Shaped Evaluation Dashboard. This application provides all the monitoring tools you need to evaluate recommendation, search and discovery algorithms both in the offline stage (before integrating Shaped), and online stages (after integrating Shaped). The dashboard will be released next week, when we’ll share more information -- keep a look out!
Performance, V1 Docs, and more
There was a huge amount of bug fixes, error logging, and performance improvements in this release. Shaped is now more stable than ever and has been tested in production at new magnitudes of scale. We’ve continued to on-board new use-cases, and have seen up to a 25% increase in engagement and conversion uplift for some scenarios — keep an eye out for several case study posts we’ll share soon.
Finally, we did a complete revamp of our docs, to make it easier to get started, view our key value props and provide typical use-case setup and evaluation guides. Our v0 ReadMe docs had served us well for the last year but I think you’ll agree that these new docs blow it out of the water!
Today, we’re also launching a new command line interface (CLI) to quickly create recommendation and search models. This CLI is long overdue for release as it makes it significantly easier to write and verify your model creation SQL transforms directly from the command line. It also helps with managing your models, and understanding their results.
As an example from our docs, here’s how simple it is to create a collaborative video recommendation model from your events using the new CLI.
1. Write your create model config
This includes specifying your connector details and the transforms needed to retrieve your user, item and events data.
2. Create your model
The CLI’s create model command is where the magic happens. When you give it the model config it’ll tell Shaped: how to provision your infrastructure, setup your data ingestion pipelines and continuously train and deploy your model.
3. Fetch your personalized rankings
You can fetch your recommendation and ranking results directly from our CLI to evaluate results before integrating to production with our REST API. Take a look at the guides or tutorials on our docs to see more about the response of this command and how to understand your evaluation results.
We hope you find it easier than ever to get started with Shaped. Please reach out to our team if you want to start integrating real-time personalized AI into your product, platform or marketplace.
If you’re a developer that just wants to try things out we’ve been giving out API keys if you join our waitlist. If you’re a company, we offer a 30 day free trial to give you the chance to evaluate uplift when using Shaped. For more information about working with us check out this page or reach out through our contact us.