Mr. Booster challenges Amazon recommendation system

Mr.Booster is the A.I. sales booster which personalizes the whole shop for the specific user.

What if someone asks you who is the most successful online retailer in the US, what will be your first answer? Probably, Amazon, and you’ll be right. Every 4th product on the Amazon is sold thanks to its recommendation system, and it is one of the reasons for its success.

But what about the shops not officially selling on Amazon? Do they have a chance to sell online as successful as Amazon does? Sure! Our team has created the A.I., based on a conscious personalization system, which allows to significantly increase the average size of the cart, customer loyalty, and conversion, for the shops, not officially present on Amazon.

Mr.Booster analyzes the historical data and customer behavior using the original machine learning and deep learning algorithms and recognizes the purchase goal. Then it personalizes the entire shop assortment for a specific user, and offers the right product to the right client.

Imagine, your customer has just purchased a suitcase and a hat, and he’s shopping again. Mr. Booster, based on the original machine learning algorithm, will detect the goal of the purchase/ and consider that most likely your customer is going on vacation. It will restructure the whole shop so that the user can see the most relevant product for him, like swimming suit or slippers of his favorite brand and price range.

Only in one short click Mr.Booster will raise the conversion in your shop up to 2.5 times, and the average check to 31% – that’s why it’s worth trying Mr.Booster.

eCommerce business dont be blind when you growth

The most successful online stores winning at e-commerce are doing so because they’ve become absolutely obsessed with metrics. They swim in data.

Every marketing and promotional decision is driving by the data. Because without data you have virtually no chance at making improvements. You don’t know what’s working, what’s failing, or even what success looks like.

Driving growth in your e-commerce business requires a few key components:

  • Setting measurable goals (key performance indicators)
  • Identifying the metrics necessary to track those KPIs
  • Monitoring performance and making adjustments as necessary

While there are numerous metrics that can be tracked, I’ve listed the ones most commonly tied to the growth of your store.

Segmented Conversion Rate

Your conversion rate is a pretty cut and dry metric. It’s the percentage of the visitors on your website who decide to make a purchase.

It’s calculated by taking the total number of website visitors who make a single purchase and dividing that number by the total number of people who visit your site.

For example; 14 customers made a purchase among 150 visitors, so the conversion rate (14 divided by 150) is 9.3%.

Your conversion rate is a good overall indicator of success, but don’t stop there.

If you break it down and segment your conversion rate you can get a lot more granular with the data, giving you tremendous insight into individual campaigns you’re using to grow your business.

A few ways to segment your conversion rates include:

1. Conversion by traffic source

Reviewing how customers convert based on the traffic source (Google, Bing, Facebook, Reddit, etc.) can tell you where you should be investing in driving traffic, or what channels to focus on improving the targeting or message you’re using for campaigns.

2. Conversion by device type

Mobile devices accounted for 19% of US retail e-commerce in 2014, and that’s expected to climb to 27% by the end of 2018.

The traffic coming from mobile is much higher. According to Yotpo, mobile accounts for more than half of all e-commerce traffic.

3. Conversion of new vs. returning visitors

Keep in mind that conversions for returning visitors are traditionally higher because those customers already know you, trust your brand, and are more willing to make a purchase.

For example; if your returning visitors are converting at ~7% but your new customers are converting at ~2% then the average is going to fall somewhere around 5%. If you use that average to calculate your max budget for acquisition campaigns that actually convert at ~2% you’re going to lose money.

Segmenting these conversions can help you more accurately calculate what you should be spending on your acquisition campaigns and how well they’re performing.

Conversion by Product

If you only have a handful of products in your online store this is likely less important. For e-commerce stores with a huge SKU inventory though, this is a necessary metric to pay attention to.

It’s a great metric for tracking the performance of individual products when you compare individual product conversions against product page traffic and those who added the product to a cart or wish list but abandoned the purchase.

Not only can this help you spot the popular or trending products, you can also find the under-performers.

Looking at the conversions by product can make it easy to look into individual barriers that could be impacting conversions (price, descriptions, product images, better benefit statements, etc.) and make strategic adjustments.

Funnel Abandonment

Cart abandonment is fairly common metric that’s tracked by online stores. E-commerce platforms are even designed to help you keep up on cart abandonment with built in autoresponders to help win back abandoned carts.

Pixels are even in place for many brands to setup ad retargeting for customers that bail on the checkout process. But are you looking at the rest of your funnel to see where customers are dropping out during the shopping experience?

This can be done manually by checking the visitor flow on your site, or you can setup a conversion funnel in Google Analytics to see where potential customers are bailing on you.

Percentage of Returning Customers

Returning visitors is a great metric to track for measuring customer loyalty, but it helps to know how those returning visits translate to revenue. That’s why you should track your percentage of returning customers – the people who come specifically to spend money.

A lot of e-commerce platforms provide customer reports with details on the number of returning customers.

Shopify provides detailed reports for first time vs. returning customers.

If you don’t have a way to access a report like this, you can export your total orders and scrub the data for duplicate emails/customer data to get a sense for repeat orders.
Percentage of returning customers is important to watch. It tells you where you stand with your customers on a number of key things:

  • Customer service
  • Price
  • Trust
  • Customer sentiment

Returning customers are highly profitable because you don’t have to pay those acquisition costs to get them back. If you see a decline in return customer rates then you need to look at overall customer delight and try to find what’s keeping customers from coming back.

It’s not just about return revenue though. The best marketers for your business are your happy, fully satisfied customers. Those are the customers who will talk you up and take the time to leave you reviews. Data shows that 55% of customers say that reading reviews online influences their decision to make a purchase.

Average Order Value

Your average order value (AOV) is the sum of the value of all of your orders (the total revenue for a period) divided by the total number of orders for that period.

Knowing your AOV is necessary to understand the lifetime value of your customer and helps you better align strategies for growth.

According to ConversionXL, there’s only three ways to grow an e-commerce business:

  • Add more customers
  • Get customers to make more repeat purchases
  • Increase the average order value

Increasing your AOV is the one that costs virtually nothing, so focus on that.

Optimizely offers some tried and true strategies for boosting AOV, such as:

  • Cross-selling (offer a product that is relevant to the product customers are interested in)
  • Upselling (offer an upgraded option, or premium product, for just a little more)
  • Volume discounts (offer a discount if a customer buys multiples of the same product)
  • Free shipping (offer free shipping when the customer hits a minimum dollar threshold)
  • Coupons (offer discounts/offers on the next purchase if they hit a minimum dollar threshold)

Lifetime Value of the Customer

Customer lifetime value (LTV) is arguably one of the most important metrics to track in e-commerce. This is the overall revenue you forecast a customer to bring you during their lifetime, or span of time as your customer.

In an earlier example calculating average order value I said the AOV was $33. If the average customer purchased 14 times at that AOV then the customer’s LTV would be $462.

This can be difficult to track for businesses with more sporadic returning customers because you have to know the lifetime of the customer, at what point they leave, the frequency and other variables.

Depending on your platform you may have built in reports to show you your top customers as well as the lifetime value of those customers (and overall customer LTV).


Putting this information together on a regular basis (at least monthly) is the secret to running a data-driven e-commerce business. One in which campaigns are launched, managed, and refined with a purpose.

Stay in tune with that data and at any time, at a glance, you’ll be able to pinpoint areas that require immediate attention and a change in strategy in order to see your business grow.

I am text block. Click edit button to change this text. Lorem ipsum dolor sit amet, consectetur adipiscing elit. Ut elit tellus, luctus nec ullamcorper mattis, pulvinar dapibus leo.

The Story of Mr. Booster recommendation system

One of the most important events in the development of recommender systems was the Netflix Prize, $1M data science competition held a decade ago. The goal was to predict the viewer’s rating for films based on previous ratings, without any other information about the customers or films. It was within the framework of this competition that the basic principles of creating effective recommendation systems were laid out. For many years the development of recommender systems was associated with minor modifications and improvements to the ideas expressed by the contestants. On one hand tremendous progress was made in the development of such systems, on the other hand for many years the experience of one narrow specific area of application was transferred to another fields without taking into account their specifics. Several paradigms are common for the existing commercial implementations of recommender systems:

  • The system needs to recommend something very similar to what the customer liked (or bought) earlier. For example, if the customer liked the first series of “Star Wars”, then it should recommend the sequel;
  • It is necessary to find a very similar customer (or customer group) by the type of purchases (preferences) and to transfer their preferences to the customer of interest. For example, if one chooses a hamburger and french fries then the systems look for another buyer with these two items in the receipt, find an additional item in there, e.g. Coke, and recommend it to the customer;
  • It is necessary to find popular combinations and in the case of the purchase of one of them to offer the others. For example, if laptops are frequently bought with mouses, then the system needs to recommend mouse to the laptop buyers;
  • If the customer browsed a specific product then they are likely interested in purchasing it, and the system should repetitively recommend it.

The last paradigm is used even by the world’s largest online platforms despite the large number of negative customer reviews. If one is planning a vacation and looking for e.g.  sunscreen and flippers then the recommendations of these goods will haunt the customer for months even if they were already purchased. The first three paradigms generally work but in a number of narrow fields with extremely low efficiency.

During the research and development of Mr. Booster recommender system, several interesting and valuable “artifacts” were produced that gave nontrivial solutions to the current challenges. In many ways, these solutions have become the key to the effectiveness of the Mr. Booster recommendation system.


Challenges and solutions


How to determine the time period to analyze customer data united by the single purchasing intent?


The first hypothesis was to find such a time interval during which most of the purchases grouped by the specific intent have to complete (remember the vacation example) and the new ones haven’t started yet. We started by aggregating purchases over the past 10 days. However, the tests showed the inconsistency of this approach: if the time interval is small enough then many purchasing goals will not be adequately identified. Yes, many can buy everything needed for vacation during the week, but often the planning can take from few weeks to several months. If we take a larger time interval instead, say 50 days, then many customers have time to make plans of totally different nature (e.g. to get apparels for the gym as well) and the original recommendations become irrelevant. We propose to search for the time intervals during which the client did not buy anything, and then if the purchases start to follow one another consider the purchasing goal to not be finished. For example, if during the last month purchases were coming with the intervals of 2-3 days, and before that there was a month without purchases, then we aggregate everything that was purchased before this pause. Such aggregation turned out to be more efficient than the previous version, however, there were problems. Some customers obviously aggregated purchases with a clearly different intent. An analysis shown in most cases the problem was that such clients had a recurring purchases (for instance socks). For such clients, purchases were aggregated over a long period of time.


How to identify the joint purchases of a group of products?


Most recommender systems do not identify frequently purchased groups of products and recommendations are based on to the approach “if goods X1 and X2 were bought, they recommend goods in addition to item X1, separately recommend goods in addition to item X2 and then merge these two lists according to some rules”. In Mr. Booster we use features of the joint group purchase. A surprisingly good solution came from a different area of Data Science – Natural Language Processing. We used a concept of N-gram, a sequence of multiple words that occur frequently in the text. We took into account that the order of goods in the recept do not affect the identification of the N-gram (in contrast, in a text analysis the order of words is important). We found that the N-grams were identified relatively rare in purchases, not because they were absent in sales but because many items have a large number of analogues and even items with the same name have a number of manufacturers. Therefore, in addition to N-gram by products we built N-grams by product categories (at the level of aggregation by the most popular international classification, S4). These allowed us to identify statistically significant product combinations (e.g. skis, poles, helmet). The absence of a product from a single category (e.g. helmets) in the combination could now be identified virtually unambiguously and Mr. Booster was giving a relevant recommendation.


How to solve a general problem of recommender systems – trivial recommendations?


For example the recommendation to watch Star trek on Netflix would be trivial. Even if there is no information on the purchase of this movie by a specific customer but there is information about their interest in the movies from this category, most likely the customer has already watched Star trek or heard about it and would not be interested in seeing this recommendation. It would be much more beneficial to recommend a relevant but less advertised movie. We studied several different approaches. The most effective direction turned out to be the analysis of how unique the recommendation the product is in general. If the product we recommend is highly specialized and rarely occurs in other recommendations it is an excellent candidate. If the product is recommended frequently it is better to limit the number of its recommendations to the instances when the system rating of its recommendation effectiveness is high. For this we developed the solution for numerical evaluation of recommendation effectiveness.