Shaped is quickly becoming a leading tool for adding personalization into any product or website. Developers benefit because of how quickly it is to connect, train and deploy state-of-the-art recommendation models for their discovery use-cases.
Shaped isn’t just any recommendation API, though — it provides semantic understanding of your unstructured user and item data and can understand your user’s session in real-time.
Algolia, a fellow YC company, has traditionally been a search focused company but has recently released Algolia Recommend for a subset of personalization use-cases. This posts dives into the trade-offs of using Shaped compared to Algolia.
The top four differences
1. Shaped connects to and transforms your raw data directly
Companies typically store data across multiple data stores and applications. For example, users and items may be stored in BigQuery and events may be logged to Amplitude.
Shaped connects directly to all these data sources, and provides the pipelines to ingest and transform the data in real-time. To get started you just need to use Shaped’s declarative SQL API to specify where your data is, who you want to personalize for and what items you want to be ranked.
Setting up Algolia, is basically like setting up a CMS, you need to manage all the data engineering work to keep your catalogues of items and users in sync. Feeding real-time events is error-prone as you’re required to manually push them through an SDK, which is disconnected from your data stack. These extra steps require more time, effort and on-going maintenance for your engineers.
2. Real-time session based ranking
If you’ve ever used TikTok, you’ll know how good it is at recommending personalized content that reacts to every like, comment, watch or click you make. If you start interacting with cat videos within a session, TikTok is going to show you more cat videos within that session. Real-time session based ranking is particularly good improving the experience for cold-start users or anonymous traffic. We’ve added the same technology to Shaped, allowing you to get the same reactive real-time recommendations that you see in leading social products. As Algolia on the other hand, requires you to push data to their cloud service, they don’t have the same level of real-time support as Shaped. For more information see our blog post.
3. Companies using Shaped get white-glove treatment
Our team of machine-learning engineers from FAANG will set up your initial models and discuss your business objectives with you. This will save you hours during setup and potentially months of experimentation time. We’ll explain how it works, what features are important and share performance insights with you regularly. While machine-learning is complicated, we’re not a black box. It’s extremely important to us that the results are interpretable and you understand how your models work.
4. Technology designed to solve the cold start problem
We use state-of-the-art machine-learning techniques and pre-train our models so you don’t need much data to get started. In addition, we support unstructured data types without you having to manually tag the metadata of your items and users. This gives our models better signals and helps deal with the cold start problem.
Shaped vs. Algolia
Ease of use and maintenance
What about pricing?
Algolia is priced at $0.60 per 1,000 requests/mo. So at 3M requests per month that would be approximately $1800. This can fluctuate wildly month-to-month depending how you use their API.
At Shaped we want to keep things simple so we do flat-fee monthly pricing based on the number of users you have and your approximate forecasted usage. We have two tiers, self-serve which is similarly priced to Algolia and white-glove support which is more expensive but allows us to take more work off your plate.
Thanks for reading, if you have any questions or want to discuss personalization for your product feel free to reach out to email@example.com