- What is Predict?
- Why should I be using Predict?
- What can Predict do for my business?
- How do we know that Predict works?
- What makes Predict better than other recommendation solutions?
What is Predict?
Before you start, why not watch a short video on the basics of Predict?
Predict is a recommendation engine which analyzes the behavior data collected from your web shop by Web Extend and uses this to deliver personalized recommendations to all your customers across email, mobile and web.
A self-learning marketing solution, 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.
Predict is made up of separate solutions, which are explained in more detail on the next pages:
- Using Email Recommendations
- Using Web Recommendations
- Discovery interface
Why should I be using Predict?
Engaging with your customers individually across all your channels is a great way to improve their experience and significantly increase ROI. With just a few hours of investment in the 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 identification method to identify each customer over all marketing channels – an industry-leading innovation that takes personalization to a new level.
- Scientific algorithms – Predict’s technology captures subtle, deeper relationships that develop as customers interact with your website. This results in highly accurate modelling 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 – Experiment with 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 Predict do for my 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:
- Personalized recommendations in newsletters – Any email campaign can be improved by personalized content with up to five times more clicks than on regular 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 browsing experience – Predict 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:
How do we know that Predict works?
We provide you with a dashboard that gives you up-to-date revenue figures from your web shop, and next to this the revenue that can be attributed to Predict, in % as well as absolute terms. We measure this attribution as follows:
- Web 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 recommender – When a contact arrives at the web shop after clicking a recommendation in an email campaign, all purchases made by that contact in the next seven days are attributed to Predict (figures are also displayed for purchases made within seven days of the click).
Additionally, our system offers a built-in A/B testing feature available through all delivery channels, 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 solutions?
Here are just a few of the reasons why Predict really is one step ahead of the rest.
Behind the Scenes
Our state-of-the-art machine learning software processes diverse behavioral data including page views, checkouts, add-to-cart events and search queries. Millions of customer interactions with 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, evolving and continually improving 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 in which it is being served (e.g. email or web).
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.
By merging cross-channel behavior into a single record we are able to ensure the most relevant and personalized experience for each individual user. In addition, thanks to the tight integration between our recommendation engine and the Emarsys email infrastructure, anonymous profiles are easily matched with identified contacts, ensuring maximum touch-point coverage.
The result is that you can personalize significantly more email to more contacts than you could hope to achieve by integrating 3rd-party recommendation engines with an ESP.