3 Ways Artificial Intelligence Is Helping Retailers As Amazon Looms

Nearly every executive team at top retail companies are discussing it:

“How can we grow digital channels outside of Amazon?”

Today Amazon accounts for 34% of all ecommerce sales… and by 2021 that number is expected to be at 50%. That’s right: half of all online sales will soon be credited to Amazon.

We can see Amazon has confidence in this forecast. The retail giant just devoured 55% of online orders for Black Friday 2017. It’s workforce growth has resulted in Amazon’s footprint in Seattle becoming twice as large as any other company in any other big U.S. city, and they are still growing.

In lieu of all of this, Amazon isn’t Apple- they don’t define themselves tightly within a single segment of the market. They regularly enter new markets and quickly begin dominating them. While some retailers may be safe now, that could change instantly as many companies can attest to.

Screen Shot 2017-11-03 at 3.09.08 PM

Regardless of a retailer’s decision to join Amazon, it’s becoming increasingly important to innovate and thrive outside of their ecosystem. The good news is (just like Amazon) you can do this by using AI to create personalized 1:1 customer experiences at scale to drive brand loyalty and revenue to your business. In addition, using ReSci’s AI platform, clients are seeing superhuman growth in acquisition and retention metrics.

Below are three ways your brand can use AI’s predictive power to succeed in channels outside of Amazon using the powerful customer data you already have.

1. Predictive Product Recommendations




Definition: Predictive Product Recommendations

Predictive Product Recommendations are driven by algorithms. These algorithms account for an individual user’s behavioral, transactional and demographic data in order to display items a particular user is predicted to have an affinity toward. Predictive recommendations with AI can be automatically served to each individual user at scale. This can be done through both retention and acquisition channels.

This contrasts older, less sophisticated methods of recommending items. Previously a brand might have recommended to users the most frequently purchased items in a particular category. 

Example of Predictive Product Recommendations

Below are examples of Predictive Recommendations from Planet Blue.

Screen Shot 2017-11-29 at 3.19.27 PM

Imagine you have an online sportswear store and a user has viewed a hat.

Instead of sending a triggered email recommending only the top three most purchased hats based on the user’s gender, predictive recommendations with AI will be able to see important patterns marketers can’t.

The AI will look at transaction, browsing, demographic and other data. It may even know that users from a certain state, of a certain gender, browsing that particular hat, are more likely to purchase triathlon related gear. It might therefore recommend a pair of sunglasses or a helmet with a visor, which statistically are more likely to result in a purchase.




While Amazon deploys AI-driven predictive recommendations to personalize their user experience, brands can also find similar success with customers by using predictive recommendations across channels like email, on site and with social ads.

In general product recommendations can increase revenue by up to 300%, improve conversions by 150% and help boost average order value by 50%. Unfortunately, only 39% of retailers send any type of personalized product recommendations via email.

Beyond failing to leverage any product recommendations with customers, the next mistake is serving irrelevant product recommendations. 74% of online shoppers get frustrated with sites that show them content or products that have nothing to do with their preferences or past buying behavior.

Specifically, this is why AI-driven predictive product recommendations with contextual relevance are vital to serving customers more accurate and diverse recommendations, which in turn boosts trust, click through rates and ultimately conversions.

Older methods of recommending products to customers also fall victim to the “cold start” problem. This is defined as a user having a data-profile that is too shallow to provide enough signals for a predictive recommendation. It’s concerning that most ecommerce companies have more than 45% of their user base as cold start.

As opposed to simply recommending top items in cases where data is thin, data scientists at ReSci have found an effective solution that results in ~19% average increase in click to open rates and ~16% increase in conversion rates compared to the baseline.




To serve hyper-relevant recommendations to your customers, ReSci makes it simple. For email, there is only one step!

Easily add dynamic tags to any email template… and everything else Resci takes care of!
As each user receives an email, the tags will display products personalized to each individual, using the methods described above!

Similarly, with our robust Recommendations API you can also use ReSci to serve dynamic on-site predictive recommendations to users (just like Amazon).


First giphybutton- click here for demo


2. Predictive Item Alerts




Definition: Predictive Item Alerts

Item alerts are notifications you send to users (often through email) that inform users of a change in status for a particular item.

Traditionally marketers send an item alert to all users (“batch and blast”) or target a broad segment of users. This is wasteful, non-personalized and under-performing for a substantial portion of the user base who might not be interested in the particular item. This method is proven to cause high unsubscribes (app and email) and to lower customer retention.

Predictive Item Alerts differ by only sending alerts to users who have a predicted preference for an item, while other users instead receive more relevant messages (something ReSci’s platform automates). This results in the highest clickthrough and conversion rates.

Types of notifications include alerts for newly introduced items, items that have dropped in price, items that are back in stock, items that have a limited quantity remaining.

Example of Predictive Item Alerts

Below is an example of a Predictive Item Alert from Hollar:


hollar image




Item alerts exist to help customers have a better, more relevant user experience.

  • New arrival alerts allow a customer to discover a new item they may be interested in.
  • Price drop alerts notify the customer they can get a better deal on an item.
  • Back in stock alerts remind users they can now buy an item previously unavailable.
  • Limited quantity alerts give customers notice to make a purchase before their item becomes unavailable.

Predictive item alerts enhance these benefits for the customer significantly.

Upon testing the common “batch and blast” technique where a notification is sent to all users, vs utilizing ReSci’s artificial intelligence, the item alerts with artificial intelligence achieved a 1.7x open rate lift and a 3.4x conversion rate lift. Most powerfully, revenue per email was almost 7x.


rates for item alerts




Again, this becomes simple with ReSci’s platform. Item alerts are pre-built within the platform. Using artificial intelligence, ReSci coordinates the individual timing, priority and group of users to be targeted before sending an alert out.

1. Go to the admin screen and click the item alert you want to use
2. Create a new email or upload a template
3. Save the email and enable that item alert stage

That’s it!

item alerts giphybutton- click here for demo


3. Predictive Ads




Definition: Predictive Ads

Predictive ads significantly improve targeting on ad platforms. Marketers can create them by using artificial intelligence to find users who fall into a desired customer cohort.

The most common predictive cohorts for targeting on ad platforms include:
1. Highest customer future value (CFV)
2. Strongest affinity toward a particular item

A marketer then syncs a user cohort to an ad platform, which then expands the cohort by finding millions of similar users to advertise to.

This is different from retargeting, which is a common ad strategy where a customer is shown products they have already seen across various ad platforms. While this is a way for ad platforms to get quick and easy attribution, many marketers believe a significant percentage of these customers would have purchased anyway without retargeting. There is also concern that the retargeting model is in danger.


Example of Predictive Ads

Below is an example of a predictive ad from Buck Mason. These ads can also include predictive product recommendations as mentioned above. In general, predictive ads appear similar to normal ads. The difference is in which particular audience an ad platform serves one of your particular ads to.


buckmason amazon




As Amazon market share increases, acquisition via alternative channels will be essential for brands. Traditionally marketers only stick to retargeting on ad networks. Predictive ads however, utilize all of your customer data to supercharge your ads for maximum acquisition potential.

Predictive targeting takes a lot more into account than just a user’s browsing history (like retargeting), and will often uncover hidden patterns that are sometimes counter-intuitive. Brands can use AI to predict who your best customers will be in the future, and let Facebook take those predicted user segments and expand them instantly. The results quickly become superhuman.

While some marketers are trying to break even on Facebook ad spend using conventional methods, marketers who utilize predictive ads see a 15x return on Facebook ad spend. That’s right- for every $1 on Facebook ads, some marketers are getting $15 back.


In a retail environment where margins are tight, utilizing artificial intelligence with your social advertising efforts can be a game changer.




Again, ReSci makes everything marketer-friendly. All you have to do to utilize Predictive Ads within ReSci’s platform is enter your Facebook credentials and sync any predictive cohorts you select with the click of a button!

Those lists will be available as custom audiences within your Facebook ad account instantly. You can then target these custom audiences directly, or run a lookalike audience from them to experience the benefits of superhuman acquisition!

<style=”text-align: center;”>predictive ads giphybutton- click here for demo



The goal for most retailers isn’t to “defeat” Amazon. The goal is to thrive in alternative channels regardless of their decision to sell on the platform.

Marketing teams should take clues from Amazon around how they can improve the customer experience. ReSci’s platform makes implementing these data driven capabilities faster and easier than ever.

By taking advantage of easy to use AI in other channels, brands can harness 1:1 personalization for strong and diversified growth.
Curious about how marketing teams at Hollar, Dollar Shave Club, Target and Sugarfina leverage our AI?
Just fill out a demo request and our team will reach out to schedule a time that’s easy for you. We’re excited to see how we might be able to help your team!