The Recommendation Engine, An Introduction To Predict

Lindsay Tjepkema
Lindsay Tjepkema
Marketing Director, Americas
Emarsys
Newsletter Signup
Yes, I would like to receive Emarsys email updates on products, services, news, and events. I can unsubscribe by using the 'unsubscribe' link at the bottom of each email, or by emailing unsubscribe@emarsys.com.

Predict is a recommendation engine, which analyzes behavioral data collected from your website and uses it to deliver personalized recommendations to all your customers across all channels: email, mobile, and web.

As a self-learning product recommendation engine, Predict intelligently creates real-time, personalized content based on the online behavior of your entire customer base. It is flexible, measurable, and reliable, and sits on top of an exceptionally lightweight client-side integration.

Anonymous behavior contributes to general affinity models covering your entire product catalog, while behavior from known customers (who have logged in or accessed the shop via a tracked link in an Emarsys campaign) is built into a unique, personal profile for that customer.

For a brief overview of the solution, why not watch a short video on the basics of Predict?

 

Why Should You be Using a Recommendation Engine Like Predict?

Engaging with your customers individually across all of your channels is a great way to improve their experience and significantly increase ROI. With just a few hours for implementation, Predict can begin to make a difference to your bottom line in a matter of days.

Here are a few more reasons why Predict should be part of every company’s marketing infrastructure:

  • Unified profiles: Predict takes advantage of our unique monitoring capabilities to identify each customer across all marketing channels, an industry-leading innovation that takes personalization to a new level.
  • Scientific algorithms: Our technology captures subtle, deeper relationships that develop as customers interact with your website. This results in highly accurate modeling of behavioral patterns and affinities.
  • Cleaner data: Background filters strip out irregular online behavior, e.g. from Internet bots or buyers who purchase in unusual patterns, making sure we capture only genuine crowd behavior.
  • Continuous improvement: Test different strategies using our built-in, automated experimentation infrastructure. Find the right campaign for every stage of your customer lifecycle.
  • Easy integration: By completing your Emarsys data onboarding, you have already done almost all of the integration work.

What Can This Recommendation Engine do for Your Business?

Due to its flexibility, Predict can make a difference in every aspect of your marketing activities, by presenting the right content, to the right people, at the right time, and in the right place.

And thanks to our benchmark control groups, we can show you this difference in black and white, at any time. Here are just a few ideas of how to use Predict:

  • Newsletter Personalization: Any email campaign can be improved by adding more personalized product recommendations. We’ve seen up to five times more clicks than those received by generic content.
  • Repurchase campaigns: Target existing customers with relevant products based on their most recent shopping behavior.
  • Abandoned cart campaigns: Predict can place proven related products alongside abandoned items for better cross- and up-sell opportunities.
  • Enhanced, real-time browsing experience: Predict constantly updates its website recommendations while the visitor is browsing, resulting in up to four times more conversions.

This diagram gives an easy overview of what Predict does and how this benefits your business model:

The Recommendation Engine

How Do We Know That Predict Works?

We provide a dashboard that gives you up-to-date revenue figures from your website, and specifically that which was directly attributable to Predict, in both % and absolute terms. We measure this attribution as follows:

  • Website recommendations: When a customer clicks a recommendation in the website and purchases that product in the same session, we attribute this revenue to Predict.
  • Email recommendations: When a contact arrives at the website after clicking a recommendation in an email campaign, all purchases made by that contact in the next seven days are attributed to Predict.

Additionally, our system offers a built-in A/B testing feature, which lets you compare the performance of your existing product recommendations with the performance of our personalization solutions. We think you’ll be impressed!

What Makes Predict Better Than Other Recommendation Engines?

Here are just a few of the reasons why Predict really is one step ahead of the rest:

  • Advanced Machine Learning: Our software processes diverse behavioral data, including page views, checkouts, add-to-cart events, and search queries. Millions of customer interactions across thousands of products are processed in real time, giving up-to-the-minute, individual recommendations with every page refresh. Our ongoing research provides a constant feed of new science that’s tested and built right into the core product, giving you a living, continually evolving platform.
  • Understanding the Buying Process: Each Predict recommendation model is designed specifically for its respective stage of the buying process (research & discovery, cart, purchase, post-purchase) and is tailored to take into account behavior differences according to the channel used (e.g. email or web).
  • Data Cleansing: Predict has a built-in mechanism that filters out traffic noise and irregular online behavior from the learning algorithms, making sure we capture genuine crowd behavior. In other words, non-human visitors (internet robots) and institutional buyers don’t affect us.
  • Feature Attractor Model: We have an advanced algorithm for dealing with long-tail items with little or no behavioral traffic. Thanks to this model, clients with large or frequently changing catalogs don’t have to suffer from the ‘cold start’ problem, and new visitors can get meaningful recommendations right from day one.
  • Unified Customer Profile: Predict uses cookies to identify anonymous website visitors and personalize their experience across desktop, mobile, and tablet devices. Through our unique identifier, Emarsys technology is able to recognize users across various channels (website, email, and mobile) and combine their browsing behavior into one Unified Profile.

By merging multi-channel behavior into a single profile, we are able to ensure the most relevant and personalized experience possible for each individual user. With the tight integration between our recommendation engine and the Emarsys email infrastructure, anonymous profiles are easily matched with identified contacts to ensure maximum touch point coverage.

Learn how our smart personalization engine helps you individualize your content for each recipient.

Related Resources

Lindsay Tjepkema
Lindsay Tjepkema
Marketing Director, Americas
Emarsys