What is Miso?
Miso is an API-first personalization platform. Our mission is to make it possible for any marketplace, eCommerce platform, or subscription media service to deliver immersive, contextual, and tailored experiences, and do it at the scale of millions of users and hundreds of thousands of catalog items. This means any product team can rapidly build search, recommendation, answers, and email marketing at the level of Facebook, Amazon, Netflix, eBay and Spotify—without needing the infrastructure, machine learning PhDs, and big data that those companies have a monopoly on.
Miso is unlike traditional/legacy search and recommendation systems:
- Miso constructs a personalization model for each and every user on your site, rather than modeling personas and audiences.
- Models work in real-time. At cold-start, Miso can instantly personalize for new users and catalog updates without needing to wait for a model refresh.
- Miso is built for high performance on fully anonymous data. User data is completely optional (and can be hashed).
- There is no minimum threshold for events needed in order to personalize. Miso works with the data you already have.
- MIso is a unified platform, so you can personalize entire user journeys with the same data and intelligence end-to-end. This is the way FANG companies build personalization (we know because we used to work there…)
What is personalization?
- Tailoring a unique experience for each and every user that ensures they can find exactly what they're looking for and discover things they never knew they needed
- Semantically understanding your products and closely observing your site's visitors in real time, then doing expert matchmaking between the two
- Having the capability to show entirely different site layouts for each user and change how content and products are presented
- The technical crown jewel 💎 of companies like Facebook, Amazon, Netflix, and Google
Personalization is not:
- Presenting things to the user that they've told you to remember (like saving their home airport on a travel app)
- "Customizing" the UI to display the user's name
- Notifications and alerts based on user-generated information (like letting them know when their order has shipped)
- A search query that returns the same results to an entire group/cohort of users
- A recommendation based on rules, like always suggesting a "buy together" offer if the user has a certain item in their cart
To do personalization, many of the big platforms have gone in a direction that is, quite frankly, creepy. At Miso we feel strongly that personalization should not and in fact need not be:
- Tracking and surveilling your users around the web
- Buying third-party data
- Horizontally mining your customers' data
How does personalization work with Miso?
Miso trains real-time personalization models at the level of each individual user or visitor, based on the product and interaction data you upload and stream directly from your site. Miso's real-time models can triangulate what a user is interested in after just a few clicks, even at cold-start. User data is completely optional—Miso supports anonymous personalization using just your session IDs.
Miso is based on the founders’ NSF and MacArthur-funded research on privacy-centric small data personalization with Professor Deborah Estrin at CornellTech. The breakthrough idea: Instead of the traditional personalization approach of mining user data to do “collaborative filtering,” Miso instead analyzes your product data first. We do so with pre-trained neural networks that can understand natural language content and context at a deep semantic level.
Then, with this core baseline of catalog comprehension, Miso also pays attention to every clickstream interaction on your site and understands how it fits into your conversion funnel (whether it leads to the user becoming more engaged, or dropping off).
With Miso Recipes, you can use APIs to build practically any kind of personalized experience, from search to product recommendations to category pages. Because Miso is a unified platform, as you add layer in more search and discovery integrations, all of these components work together. For instance, as a user searches and filters on your SERP, Miso’s recommendation engines are paying attention to prepare optimized product recommendations for the context and intent cues your user is expressing on the SERP. In other words, more integrations make Miso better overall—our customers typically see an immediate increase of 50–200% on their conversions with steady growth over time as they build out additional recipes.
Where should I start?
- If it's your first time using Miso, we'd recommend going straight to Quickstart. This guide will walk you through building an end-to-end Miso Recipe with sample data in about 20 minutes.
- Once you've completed the Quickstart, we suggest taking a look at Miso's Recipes. There you'll find a clear set of personalization use cases we support, along with step-by-step instructions for building each one.
- Next, you might want to learn about what it's like to work with Miso's team.
- If you're the type of person who likes to read the whole textbook before jumping in, you can visit our Deep Dives.